Full Report

The numbers behind The Trade Desk, Inc.: as-reported financial statements and company metrics for FY2021–FY2025, traced to the source filings, opened with the share-price history those statements have to justify. Every linked figure opens the exact page of the filing it was printed on, with the statement row highlighted. Amounts in US$ thousands unless noted.

Reading notes: All figures are in US$ thousands, the unit line printed on every Trade Desk statement page ("In thousands, except per share amounts"). Per-share rows are as printed. The Trade Desk operates one reportable segment (advertising technology platform), so there is no segment profit disclosure. The revenue disaggregation the company reports is by principal geographic area, based on the address of the client or client affiliate. Revenue by geographic area was first disclosed in the FY2025 Form 10-K (Note 12), for 2025, 2024 and 2023; that filing states the 2024 and 2023 amounts were recast to the current-year presentation. FY2021 and FY2022 revenue was never disaggregated geographically, so those cells are blank rather than estimated. The FY2021-FY2024 Form 10-Ks disclosed Gross Billings by geography instead of revenue by geography; the FY2025 Form 10-K dropped that table, so FY2025 Gross Billings is not disclosed and is left blank.

Share Price — Full Available History — 10 Years

The stock closed at $17.29 on Jul 24, 2026 — up 474% over the window shown (+19.4% a year), trading between $2.28 and $139.51. At that close the stock trades at 19× FY2025 diluted EPS as reported below.

Source: market price feed, weekly closes, sampled from 2,474 source observations, Sep 2016–Jul 2026. Price return only, excludes dividends.

Market capitalization $316mn.

Market cap = 18.3M shares outstanding × the Jul 24, 2026 close of $17.29. Market-derived, shown without filing links.

FY2025 at a Glance

Source: FY2025 consolidated statements [1] [2] [3] [4]. Click any linked figure to open the filing page with the row highlighted.

Revenue by Geographic Area

Revenue by Geographic Area	FY2021	FY2022	FY2023	FY2024	FY2025
United States	—	—	1,696,911	2,133,502	2,476,683
International	—	—	249,209	311,329	419,601
Total	—	—	1,946,120	2,444,831	2,896,284

Source: Form 10-K Note 12 Segment and Geographic Information. The Trade Desk operates a single reportable segment; revenue by principal geographic area (based on the address of the client or client affiliate) was first disclosed in the FY2025 Form 10-K, for 2025, 2024 and 2023. FY2021 and FY2022 revenue was not disaggregated geographically in the filings. [5]. Click any linked figure to open the filing page with the row highlighted.

Income Statement

Source: Consolidated Statements of Operations [1] [2] [3] [4]. Click any linked figure to open the filing page with the row highlighted.

Columns marked E are consensus analyst estimates from S&P Capital IQ (CapIQ), shown alongside reported results for direct comparison; they are not company guidance.

Estimate source: S&P Capital IQ (CapIQ) consensus, as of 2026-07-30. Estimate figures are S&P Capital IQ consensus (vendor data — no filing page links). EPS and net income use the normalized (adjusted) consensus where the street reports it. Line-item analyst models (segments, drivers, KPIs) are in the Visible Alpha tab.

Balance Sheet

Source: Consolidated Balance Sheets [6] [7] [8] [9]. Click any linked figure to open the filing page with the row highlighted.

Cash Flow

Source: Consolidated Statements of Cash Flows [10] [11] [12] [13]. Click any linked figure to open the filing page with the row highlighted.

Gross Billings by Geographic Area

Gross Billings by Geographic Area	FY2021	FY2022	FY2023	FY2024	FY2025
United States	5,286,191	6,696,743	8,216,446	10,244,266	—
International	843,436	937,824	1,214,207	1,508,501	—
Total	6,129,627	7,634,567	9,430,653	11,752,767	—

Source: Form 10-K segment and geographic footnote. Gross Billings is the gross amount clients are billed for supplier features plus platform fees, net of allowances; revenue is reported net of amounts paid to suppliers. The FY2025 Form 10-K replaced this Gross Billings disaggregation with revenue by geography, so FY2025 is not disclosed. [14] [15] [16]. Click any linked figure to open the filing page with the row highlighted.

Platform Ecosystem and Client Loyalty

Platform Ecosystem and Client Loyalty	FY2021	FY2022	FY2023	FY2024	FY2025
Directly integrated ad exchanges, publishers and SSPs	105	100	140	220	430
Integrated third-party data vendors	200	200	250	350	370
Client retention rate (disclosed as over 95%)	95.0%	95.0%	95.0%	95.0%	95.0%

Source: company filings [17] [18] [19] [20]. Click any linked figure to open the filing page with the row highlighted.

Client and Supplier Concentration

Client and Supplier Concentration	FY2021	FY2022	FY2023	FY2024	FY2025
Agency holding companies above 10% of Gross Billings	2	1	1	1	2
Gross Billings from holding companies above 10%	—	11.0%	12.0%	14.0%	30.0%
Clients individually above 10% of accounts receivable	3	4	2	3	2
Accounts receivable held by those clients	41.0%	49.0%	31.0%	42.0%	30.0%
Suppliers individually above 10% of accounts payable	1	2	2	2	2
Accounts payable held by those suppliers	17.0%	25.0%	31.0%	36.0%	34.0%

Source: company filings [21] [22] [23] [24]. Click any linked figure to open the filing page with the row highlighted.

Non-GAAP Measures Management Reports

Non-GAAP Measures Management Reports	FY2021	FY2022	FY2023	FY2024	FY2025
Adjusted EBITDA	502,698	667,682	771,526	1,010,649	1,196,449
Adjusted EBITDA margin	42.0%	42.0%	40.0%	41.0%	—
Non-GAAP net income	455,556	522,032	628,099	832,303	873,080
Non-GAAP diluted earnings per share	0.91	1.04	1.26	1.66	1.77

Source: company filings [25] [26] [27] [28]. Click any linked figure to open the filing page with the row highlighted.

Workforce and Global Footprint

Workforce and Global Footprint	FY2021	FY2022	FY2023	FY2024	FY2025
Full-time employees	1,967	2,770	3,115	3,522	3,843
Countries with employees	19	19	19	20	21
North America share of workforce	65.0%	63.0%	64.0%	64.0%	63.0%
EMEA share of workforce	18.0%	19.0%	19.0%	19.0%	20.0%
APAC share of workforce	17.0%	18.0%	17.0%	17.0%	17.0%

Source: company filings [29] [30] [31] [32]. Click any linked figure to open the filing page with the row highlighted.

Capital Returns, Equity Overhang and Tax

Capital Returns, Equity Overhang and Tax	FY2021	FY2022	FY2023	FY2024	FY2025
Class A shares repurchased and retired (thousands)	—	—	10,000	2,500	26,200
Repurchase authorization remaining at year end	—	—	53,000	464,000	150,000
Shares available for grant under Incentive Award Plan (thousands)	56,400	69,000	81,200	94,900	106,800
Effective income tax rate	(13.0%)	58.0%	33.0%	23.0%	33.0%

Source: company filings [33] [34] [35] [36]. Click any linked figure to open the filing page with the row highlighted.

Long-Term Record

Fiscal year	Revenue	Income from operations	Net income	Earnings per share, diluted	Net cash provided by operating activities	Purchases of property and equipment	Total stockholders equity
FY2016	202,926	57,518	20,482	—	75,031	(6,884)	164,380
FY2017	308,217	69,356	50,798	—	31,224	(10,110)	245,583
FY2018	477,294	107,323	88,140	—	86,603	(19,795)	394,567
FY2019	661,058	112,196	108,318	0.23	60,205	(35,693)	612,517
FY2020	836,033	144,208	242,317	0.49	405,069	(74,061)	1,013,145
FY2021	1,196,467	124,817	137,762	0.28	378,513	(54,804)	1,527,306
FY2022	1,577,795	113,654	53,385	0.11	548,734	(84,160)	2,115,339
FY2023	1,946,120	200,480	178,940	0.36	598,322	(46,790)	2,164,219
FY2024	2,444,831	427,167	393,076	0.78	739,456	(98,238)	2,949,145
FY2025	2,896,284	589,321	443,304	0.90	992,721	(197,011)	2,484,391

Source: consolidated statements across filings; older years from the standardized feed [10] [1] [6] [11]. Click any linked figure to open the filing page with the row highlighted.

Operating KPIs

KPI	FY2021	FY2022	FY2023	FY2024	FY2025
Gross spend	6,172,000	7,741,000	9,611,000	12,040,872	13,394,683

Source: company-reported operating metrics [37]. Click any linked figure to open the filing page with the row highlighted.

Analyst Consensus

Street ratings: 11 strong buy, 2 buy, 19 hold, 1 sell, 3 strong sell. Consensus: Hold.

Estimate source: S&P Capital IQ (CapIQ) consensus, as of 2026-07-30. Estimate figures are S&P Capital IQ consensus (vendor data — no filing page links). EPS and net income use the normalized (adjusted) consensus where the street reports it. Line-item analyst models (segments, drivers, KPIs) are in the Visible Alpha tab.

Traceability

536 of 558 figures on this page (96%) link to the filing page where they are printed — click a linked figure to open the source PDF at that page with the row highlighted. Unlinked figures come from standardized data feeds or pre-filing years.

• All figures are in US$ thousands, the unit line printed on every Trade Desk statement page ("In thousands, except per share amounts"). Per-share rows are as printed.

• The Trade Desk operates one reportable segment (advertising technology platform), so there is no segment profit disclosure. The revenue disaggregation the company reports is by principal geographic area, based on the address of the client or client affiliate.

• Revenue by geographic area was first disclosed in the FY2025 Form 10-K (Note 12), for 2025, 2024 and 2023; that filing states the 2024 and 2023 amounts were recast to the current-year presentation. FY2021 and FY2022 revenue was never disaggregated geographically, so those cells are blank rather than estimated.

• The FY2021-FY2024 Form 10-Ks disclosed Gross Billings by geography instead of revenue by geography; the FY2025 Form 10-K dropped that table, so FY2025 Gross Billings is not disclosed and is left blank.

• FY2019 and FY2020 long-term figures are comparative columns of the FY2021 Form 10-K; FY2019 stockholders equity has no filing coverage in the corpus (the FY2021 balance sheet prints only 2021 and 2020) and is taken from the data feed, unlinked.

• FY2016-FY2018 long-term figures come from the standardized data feed (SEC XBRL) and are shown without page links; no 10-K older than FY2021 is in the corpus.

• Income-statement sign convention is as printed: "Total other expense (income), net" is negative when it is net income (FY2022-FY2025) and positive when it is net expense (FY2021). The FY2024 and FY2025 10-Ks label this row "Total other income, net".

• Quarterly income statements for Q4 FY25 come from the GAAP financial information table in the February 25, 2026 investor presentation, the only document in the corpus printing the three months ended December 31, 2025; Q4 FY24 comes from the February 12, 2025 Form 8-K earnings exhibit. All other quarters come from the Form 10-Q.

• Quarterly balance sheets at December 31, 2024 and December 31, 2025 are cited to the audited FY2024 and FY2025 Form 10-Ks rather than to a press release.

• Quarterly cash flows are derived from the printed year-to-date statements: Q1 as printed, later quarters as the current year-to-date figure less the prior quarter year-to-date figure, with Q4 taken as the full-year 10-K statement less the nine-month 10-Q. Every derived value reconciles exactly to the two printed figures and to the change in the quarter-end cash balance, and Q2 FY25, Q3 FY25 and Q1 FY26 match data/financials/cash_flow_quarterly.json exactly.

• Cross-checked against data/financials/*.json for FY2021-FY2025: revenue, operating income, net income, operating cash flow, total assets, total liabilities and stockholders equity agree with the filings to the dollar. No feed-versus-filing conflicts were found in the core statements.

• 2 figure(s) differed between the data feed and the filing; the filing value is shown (see the run's metrics/metrics_tab.json for the audit trail).

---

The Trade Desk, Inc.'s management explains the business in its own materials. The slides below do the most of that work, pulled from the documents preserved in Sources. Each source link opens the complete presentation at that slide in a new tab.

Investor Relations Presentation — Q1 2026 — Q1 2026

The current edition of management's standing investor deck: what the platform does, the market it sits in, and where growth comes from. · Open the full document →

Investor Relations Presentation — Q3 2025 — Q3 2025

The prior deck, kept for the teaching slides the 2026 editions dropped: the ecosystem map, media-plan economics and financial history. · Open the full document →

More from management

Investor Relations Presentation — FY2025 — FY2025 · 48 pages · The same deck built on full-year 2025 results, and it still carries the seven-point strategy roadmap the May 2026 edition dropped. · Open →

Investor Relations Presentation — FY2024 — FY2024 · 52 pages · Full-year 2024 results and the framing management used a year earlier, including ~88% of spend still in North America. · Open →

Investor Presentation — FY2023 — FY2023 · 51 pages · The FY2023 edition, when management sized the market at ~$830B rather than $1T+ — the earlier TAM framing. · Open →

Investor Presentation — Q2 2023 — Q2 2023 · 52 pages · The oldest deck here: profitability milestones since 2013 and the disclosure that 95%+ of spend runs through MSAs. · Open →

---

The Trade Desk, Inc.'s management answers for the business every quarter. These are the exchanges that explain it best — verbatim, from the call transcripts preserved in Sources. Each link opens the full transcript at that page in a new tab.

Q1 FY2026 Earnings Call — Q1 FY2026

The current state of the story: growth decelerating on macro, management drawing the line between cyclical and structural, and the clearest statement yet of why it refuses to own supply. · Open the full transcript →

The retail-data scale claim against Amazon, and what a flat-fee data product did to one campaign.

Jeffrey Terry Green (CEO and Co-Founder): Over the last five years or so, we have created the world’s largest and richest marketplace of retail data. Combined, we believe the retailers in our data marketplace represent more than 80% of sales from top U.S. retailers, compared to Amazon, who represents less than 15% of U.S. retail spend. This is a huge advantage for us. […] For example, a leading travel brand recently ran a test to evaluate campaign performance with and without activating our new product Audience Unlimited. The results across all KPIs were fantastic. Audience Unlimited delivered 30% lower CPMs on media, 38% lower data costs, a 75% more efficient CPA, and a 2.7x increase in conversion rate compared to the control group. Most importantly, Audience Unlimited increased campaign performance while simultaneously reducing manual effort in the audience selection process.

p. 5 · Read in context →

How the revenue actually splits — by channel, geography and vertical — in the most recent quarter.

Tahnil Davis (Interim Chief Financial Officer and Chief Accounting Officer): In Q1, we delivered revenue of $689 million, representing 12% year-over-year growth. We generated $206 million of adjusted EBITDA during the quarter, representing a 30% margin. Our growth in Q1 was driven by strong trends across CTV and audio. Video, which includes CTV, represented a low-50s percent of our business in Q1 and continues to grow as a percentage of our channel mix. Mobile represented a high-28s percent share of the business during the quarter, while display represented a low double-digit share. Audio represented around 6% of the business and grew year over year at a rate higher than any other channel in Q1. […] Geographically, the United States represented approximately 82% of our revenue in Q1, and international represented approximately 18%. Our strong momentum in both EMEA and APAC reflects the investments we have made in these regions over the last several years as well as momentum in CTV across these markets. Among verticals that represent at least 1% of our business, we saw particularly strong growth in medical health, automotive, and events. We continue to see some pressure in the home and garden and food and drink sectors as CPG brands navigate geopolitical uncertainty, consumer softness, and input cost inflation.

p. 6 · Read in context →

The margin commitment defended with softer revenue: how the 40% EBITDA target is meant to hold.

Justin Tyler Patterson (KeyBanc); Tahnil Davis (Interim Chief Financial Officer and Chief Accounting Officer): Great. Thanks. Good afternoon. I am curious to hear more about investment priorities against that 40% EBITDA margin target. Obviously, revenue and margins are both off to a softer start in the first half. I am curious how we should think about the levers to achieve that target. Thank you. […] As a company, we have always been very disciplined around hiring and reinvestment in the business. 2026 is a year of disciplined reinvestment for us. We expect our full-year adjusted EBITDA margin percentage to be at least 40%, approximately in line with last year. We again expect headcount growth to remain below revenue growth, reflecting continued operating discipline and increasing productivity across our business. At the same time, we will continue investing in areas where we see the highest long-term ROI, particularly around platform innovation, AI, retail media, and measurement. One advantage of our model is that we generate strong cash flow and can maintain significant flexibility in how we pace our investments and expenses, which allows us to maintain those high levels of profitability. So our focus is clear: maintain strong profitability, invest where ROI is the highest, and continue positioning the business for greater leverage over the long term.

p. 9 · Read in context →

What the platform actually does per second, and why management thinks agentic AI reinforces rather than bypasses it.

Jessica Reif Ehrlich (Bank of America); Jeffrey Terry Green (CEO and Co-Founder): Thank you. Jeff, you said early in the call in your prepared remarks you mentioned the partnership with Stagwell. It just seems like you would not have brought that up if it was not important. I know it is early days, but when do you think agentic trading will become the dominant dynamic in programmatic media, and how will The Trade Desk, Inc. be impacted by this? […] We fundamentally believe that we will lead the agentic revolution in programmatic advertising. I have said on a number of stages in our industry recently that I do not think there is an industry in the world that is better suited to be upgraded from agentic AI than programmatic advertising. I do think programmatic will benefit tremendously from agentic. […] Most companies that are focused on agentic in our space are just talking about plugging into these tiny pools of inventory—one advertiser connects to one publisher. In doing so, you more or less create another ad network, where you have hundreds of thousands of ad networks, because of the combination of advertiser to one publisher and agents talking to each other, and it gets rid of the opportunity for you to look at everything at once and then make holistic decisions and compare all of those. That is part of the reason also in the prepared remarks that we talked about why it is so important to look at all of the QPS that we do and to maintain decisioning so that you can look at those—currently 20 million ad opportunities every single second—and then choose carefully the 300 or 400 the biggest brands in the world should be buying. […] One simple way to explain agentic AI is that it is a layer on top of the API that can reason—or, said simply, an API that can reason—while simultaneously creating productivity. What we started with Stagwell is the ability to create and edit campaigns in the most basic form. That will, of course, evolve into optimizations.

p. 12 · Read in context →

Q4 and Full Year FY2025 Earnings Call — Q4 FY2025

The best single explanation of the model: where the take rate comes from, what OpenPath charges, why CPG and auto cost five points of growth, and what the reorganization was for. · Open the full transcript →

A head-to-head against the Amazon DSP, quantified — the concrete case for not owning inventory.

Jeff Green (CEO): One of the world's leading appliance manufacturers recently ran a test between The Trade Desk and the Amazon DSP, focusing on CTV ad performance in one of their most important markets. They found that, with The Trade Desk, they were able to reach 70% more unique households because we gave them access to a much wider range of relevant touch points with those consumers. With The Trade Desk, they were able to reach those consumers at 30% lower total cost, so significantly better reach for meaningfully lower cost. And the kicker is The Trade Desk platform performed six times better in terms of delivering their campaign goals. All of this happened because we provided the client with objective decisioning across the open Internet. We didn't prioritize our own impressions because we don't own any.

p. 2 · Read in context →

The guidance philosophy behind a weak Q1 margin: a timing effect, with the data-center transition named.

Tahnil Davis (CFO): So regarding our Q1 EBITDA guide, thanks for the question. So first, in Q1, I would characterize this as primarily a timing thing. We continue to expect full-year adjusted EBITDA margins to be approximately in line with 2025. The primary driver in Q1 is infrastructure investment. […] Our incremental investments are focused on infrastructure and talent. In particular, we're completing our transition to owned data centers and strengthening the AI and machine learning capabilities that power our platform. So the balance is clear: invest where the ROI is highest, maintain strong profitability, and position us for improved leverage beyond 2026.

p. 8 · Read in context →

Why management rejects the brand-versus-performance split, and what last-touch attribution breaks.

Jeff Green (CEO), replying to Matthew Swanson (RBC): There is this narrative on Wall Street that performance budgets are more DTC or they're more mid-market and then there's brand budgets that are separate from that. That paradigm, I just reject. I think everything is performance now; it's just a matter of where you are in the funnel. The problem with framing it that way is you reinforce a serious problem in the ecosystem, which is that all of the measurement frameworks that have existed to date just give credit to the last person who touched the ball before it went in the net. The rest of the team gets nothing. […] Nobody types in 'Buy Mercedes-Benz' into Google without seeing the commercial or hearing about the company before that. Giving all the credit to the last touch has been a serious mistake.

p. 13 · Read in context →

Q3 FY2025 Earnings Call — Q3 FY2025

The call where management finally put numbers on the Amazon question, defined the open internet from first principles, and answered what happens if a rival prices its DSP at zero. · Open the full transcript →

A first-principles definition of the open internet — the single most useful passage for a new reader.

Jeff Green (CEO): A reminder, the open Internet is the portion of the Internet where price discovery and competition exists. In the open Internet, every transaction is arm's length. Walled gardens are built around owned and operated inventory instead of third-party inventory. Price discovery comes when the buyer and seller are different entities.

p. 1 · Read in context →

Trial exhibits used to argue Google's DSP stopped buying the open internet: YouTube up ~800%, open internet flat.

Jeff Green (CEO): This was helpfully made public throughout the antitrust trial of the Department of Justice versus Google when they revealed numbers that Google normally does not report on. Exhibits and industry experts estimated that in 2019, the open Internet and owned and operated inventory on YouTube were equally split in share of wallet on DV360. However, between 2019 and today, roughly all of the incremental dollars and growth from DV360 has gone to YouTube. YouTube spend increased by about 800%, while Google's buying of the open Internet stayed essentially flat for the same period of time. During that time, the Trade Desk seems to have surpassed Google in the amount bought on the open Internet, again, according to others.

p. 2 · Read in context →

Where the growth is supposed to come from: consumer time on the open internet and 60% of TAM outside the US.

Jeff Green (CEO): Second, we're introducing trading modes. This is a bit like driving modes in a car where the user can decide how they would like to engage with the system. Would they prefer to have control where they have more decisions and a greater burden of work? Or would they prefer to simply optimize the performance and lean on the machine? In both cases, we're introducing Agentic AI as a copilot to ensure optimal campaign performance, but its role and engagement will differ based on the trading modes. […] Our research shows that the average consumer now spends two-thirds of their digital time on the open Internet, even though most budgets today still go through Facebook, Google, and TikTok. This imbalance will correct over time. Outside the U.S., our business is growing significantly faster than in the United States. Given that 60% of the TAM is outside of the U.S., this movement is in the right direction of capturing the TAM.

p. 5 · Read in context →

The account-coverage machinery quantified: live JBPs, pipeline, scorecards and the resulting CPM decline.

Jeff Green (CEO), replying to Justin Patterson (KeyBanc): Last quarter, we had over 180 live JBPs with some of our largest clients. We have an additional 80 JBPs in the pipeline right now worth billions of dollars in total. We've been driving individual contributor-level accountability with BD, AM, and trader scorecards. As a result of those accountabilities alone, we believe that's contributed to CPMs declining by up to 43% on average and resulting in meaningful return on ad spend improvements. […] Today, joint business plans now make up about half the business. As you can tell from that pipeline, we're in a great position for that to be even more of our business as we go into next year.

p. 9 · Read in context →

Asked if AI search shrinks publisher inventory; the answer sizes the funnel TTD actually selects from.

Jason Helfstein (Oppenheimer); Jeff Green (CEO): Jeff, are you seeing an impact from Agentic search on available publisher inventory? And how are you helping publishers navigate that and basically AI? […] So we look at roughly 20 million ad impression opportunities every single second. That's about $1.7 trillion every single day. That means we're doing more transactions than Visa, Mastercard, and American Express combined do in a year in less than 30 seconds. When you look at that many impressions, and just to be open, we buy a low single-digit percentage of that total, you'd have to when it's that big. So that means that if we take 20 million down to 15 million per second because of AI, there's not really much different about our business model, nothing at all.

p. 11 · Read in context →

Q2 FY2025 Earnings Call — Q2 FY2025

The call that set off the Amazon debate — "Amazon is not a competitor" in full context — plus the dual-class extension rationale and why a large-advertiser book behaves differently from an SMB one. · Open the full transcript →

Why tariff pressure hits TTD differently: the book is large global brands, not SMBs.

Vasily Karasyov (Cannonball Research); Jeffrey Terry Green (CEO): So Jeff, you work with nearly all of the world's biggest advertisers. And Laura, in her prepared remarks, mentioned the uncertainty because of the tariff situation. We also heard names like P&G, Kimberly-Clark, Ford, and Volkswagen talk about this uncertainty on their earnings calls. So my question is, how do you see that dynamic playing out in terms of ad spend in the remainder of the year? And how are you factoring that into your Q3 guidance? […] There is an important point that we haven't made enough, I think, in our prepared remarks, and I don't know that this is fully appreciated about the difference between us and many other businesses in digital advertising. Most others rely heavily on SMBs, and our platform is largely concentrated on the large global advertisers. So we see the effects that are directly impacting them. So I would argue that this is a short-term negative, which by the way, this fact that we concentrate on the large ones is not generally a negative. It is almost always a positive. But just in this moment, it's negative because of how uniquely they're being affected by the tariffs and related policies.

p. 9 · Read in context →

The hardest question on the call: if the open internet is winning, why are the walled gardens growing faster?

Jessica Jean Reif Ehrlich Cohen (Bank of America); Jeffrey Terry Green (CEO): Jeff, while you present a strong case for the open Internet, it seems to be losing market share when you consider the growth rates of major platforms like Meta, Amazon, and Google. How do you see the share shift between walled gardens and the open Internet in the coming years? Are Connected TV and retail media simply growing at a slower pace than walled gardens? Specifically for Trade Desk, when you speak about capturing market share, who are you taking it from? Is it other demand-side platforms or the walled gardens? […] While Facebook had a strong quarter and clearly understands the potential of AI, their situation allows for easier optimization du to their abundant supply and the consistent engagement on Instagram. In the short term, integrating AI into Facebook and Instagram is less complex than building better supply chains across the entire Internet. Our focus on enhancing the premium segments of the open Internet may take longer but offers greater potential. […] It's essential to recognize that we are engaged in a distinct and extended game, and the impact of AI provides us with a substantial advantage.

p. 11 · Read in context →

Q4 and Full Year FY2024 Earnings Call — Q4 FY2024

The landmark call: the first guidance miss in 33 quarters, owned as self-inflicted, with the reorganization, the Google-exit thesis and the TAM math laid out under direct questioning. · Open the full transcript →

The structural response: largest reorganization in company history, and engineering broken into scrum teams.

Jeff Green (CEO): First, we implemented the largest reorganization in company history in December. While we usually make structural changes at year-end to enhance our business, this one was larger than usual. We clarified roles and responsibilities for most employees, resulting in a change in reporting structures. Additionally, we streamlined client-facing teams, minimizing complexity and clarifying duties. Some teams now focus on brands, while others concentrate on agencies. Our commitment to agencies remains strong, while we expand direct relationships with brands, particularly through Joint Business Plans, which grow 50% faster than the rest of our business. […] Fourth, we revamped our product development process, returning to smaller, agile teams that provide weekly updates instead of relying on waterfall methods, which are less suitable for our fast-changing industry. Our engineering team is divided into nearly 100 scrum teams, enhancing collaboration with the business team on what has been accomplished and what’s upcoming.

p. 2 · Read in context →

The two bets underneath the thesis: that Google leaves the open internet, and that objectivity is the moat.

Jeff Green (CEO): Second, we are preparing for a world where Google distances itself from the open internet. I believe Google will eventually withdraw from the open internet, which would address many of its antitrust issues. […] Third, we will prioritize and safeguard our objectivity more than ever. Increasingly, the few competitors we face exhibit significant objectivity issues. Amazon, for instance, is soliciting ad budgets while competing against numerous Fortune 500 companies across various sectors. Fifteen years ago, we argued that the objective, independent DSP should capture the majority of the market because it can be trusted.

p. 2 · Read in context →

The CFO takes ownership of the forecast failure, and confirms take rate held its historical range.

Laura Schenkein (CFO): However, for the first time, in our 8.5 years as a public company, excluding the first quarter of 2020, our results came in below our expectations. As a company, we take great pride in our ability to forecast accurately, and we take full ownership of this shortfall. Importantly, this miss was not due to lack of opportunity or increased competition, it was on us. […] As expected, our take rate in 2024 once again remained within a very consistent historical range. The shift of advertising dollars to CTV continues to be a core driver of our business.

p. 5 · Read in context →

The investment answer: deliberate deleverage, capital intensity near 5% of revenue, and a stated off-ramp.

Laura Schenkein (CFO): On the investments required for 2025, first just looking back at 2024, we delivered an incredibly strong year in terms of profitability and cash flow generation. And we exited the year with a strong balance sheet. So as I mentioned in the script, we anticipate a modest increase in the growth rate of our operating expenses in 2025 compared to previous years. And as a result of that, we would expect some deleverage for the year. […] Our capital intensity also remains low. We expect CapEx to be approximately 5% of total revenue. And when I look across our growth drivers frankly, I believe nearly all of them are still in their early stages compared to where they will be in 5 to 10 years. So if we generate significant revenue gains, we'll continue investing. And if not or if the current environment significantly changes, we'll have the flexibility to adjust our investment pace accordingly.

p. 9 · Read in context →

Amazon separated into its three advertising businesses — the framing management reused for the next year.

Jason Helfstein (Oppenheimer); Jeff Green (CEO): Thanks for taking my question. So Jeff, I just wanted to ask a bit about Amazon. It's gotten a lot of investor attention, a lot of trade press as far as the company making improvements to their DSP, getting aggressive with Prime Video ads. Just how do you view them in the competitive landscape? Did you see any kind of change in the fourth quarter? And just, I guess, how do you think about them as a competitor going forward? Thank you. […] And I think it is really important that investors parse out the three roles that Amazon plays in advertising. The biggest one by far is that they are a search engine, competing with Google's core business if you will. And that is the biggest source of revenue for them in advertising. The second is probably Prime Video. And I think that one is very interesting because I think that the right way to look at them is somebody like Paramount or like Box. They are creating premium content, and they created a lot of ads as a result of that.

p. 10 · Read in context →

More calls

Q1 FY2025 Earnings Call — Q1 FY2025 · 12 pages · The first call after the miss, where management argues the fixes took hold — 25% growth — and restates the founding bet from the original business plan that there would be ten or fewer scaled DSPs, most of them conflicted. · Open →

Q3 FY2024 Earnings Call — Q3 FY2024 · 14 pages · The source of the "10 macro forces" framework management cited on the next several calls; go here for the pre-miss version of the bull case, set out one force at a time. · Open →

Q2 FY2024 Earnings Call — Q2 FY2024 · 13 pages · The operating model at its peak — 26% growth described as the continuation of a multi-year 20%-plus streak — useful as the baseline the later deceleration is measured against. · Open →

Q1 FY2024 Earnings Call — Q1 FY2024 · 13 pages · Growth accelerating to 28% with the heaviest UID2 and OpenPath discussion of any call, before the Kokai migration became a drag. · Open →

Q4 and Full Year FY2023 Earnings Call — Q4 FY2023 · 19 pages · The clearest statement of the take-rate model as designed at inception: hold take rate constant while raising platform value, alongside FY2023 spend of $9.6 billion. · Open →

Q1 FY2022 Earnings Call — Q1 FY2022 · 42 pages · The pre-Kokai era, for readers wanting the origin of UID2 and OpenPath and the 95%-plus customer retention figure management leaned on then. · Open →

---

The Trade Desk, Inc.'s annual reports contain management's most considered account of the business. These are the sections, passages and visual pages worth opening in the originals preserved in Sources.

The Trade Desk, Inc. — FY2025 Annual Report (Form 10-K) — FY2025

The latest 10-K, and the one that recasts the company as an AI-led advertising technology leader while revenue growth slowed to 18%. · Open the full document →

Item 1. Business — Overview and Our Industry — p. 7 · Read the full section →

Management's own definition of the business and the one sentence that explains how a platform fee on client spend becomes revenue.

How The Trade Desk describes itself and how it gets paid.

We are a global leader in advertising technology. We empower ad buyers to create, manage and optimize digital advertising campaigns across ad formats, channels and devices. […] Our clients are advertising agencies, advertisers and other service providers for agencies or advertisers, with whom we enter into ongoing master services agreements (“MSAs”). We generate revenue by charging our clients a platform fee generally based on a percentage of our clients’ total platform spend and from providing value-added services and data to support their advertising campaigns.

p. 7 · Read in context →

Item 1. Business — Our Clients; Our Advertising Inventory and Data Suppliers — p. 16 · Read the full section →

Where the spend comes from and on what terms: contracts carrying no commitment, and concentration at the holding-company level.

Contract terms and holding-company concentration in gross billings.

Our MSAs, some of which may include joint business plans and other incentive programs, do not contain any material commitments on behalf of clients to use our platform to purchase ad inventory, value-added services or data. Generally, these MSAs have one-year terms that renew automatically for additional one-year periods, unless earlier terminated, and are terminable at any time upon 60 days’ notice by either party. […] If all of our individual client contractual relationships were aggregated at the holding company level, two holding companies would have each represented more than 10% of our gross billings in 2025 and one holding company would have represented more than 10% of our gross billings in 2024.

p. 16 · Read in context →

Item 1A. Risk Factors — The loss of advertising agencies, advertisers or holding companies as clients — p. 24 · Read the full section →

Concentration risk stated plainly: no exclusivity, agencies own the advertiser relationship, and spend can move.

Why the agency layer sits between the platform and the advertiser.

Our client base consists primarily of advertising agencies and advertisers. We do not have exclusive relationships with advertising agencies or advertisers, and we depend on agencies to work with us to build and maintain advertiser relationships and execute advertising campaigns.

The loss of agencies or advertisers as clients could significantly harm our business, financial condition and results of operations. If we fail to maintain satisfactory relationships with an advertising agency, we risk losing business from the current and future advertisers represented by that agency.

p. 24 · Read in context →

Item 1A. Risk Factors — The market in which we participate is intensely competitive — p. 28 · Read the full section →

The structural asymmetry for an independent demand-side platform: rivals that own the inventory they sell.

The walled-garden problem stated in the company's own words.

Furthermore, our current and potential competitors may have significantly more financial, technical, marketing, and other resources than we have, which may allow them to devote greater resources to the development, promotion, sale and support of their products and services. They may also have more extensive advertiser bases and broader publisher relationships than we have, rich first party data sets, may be better positioned to execute on advertising conducted over certain channels, such as social media, mobile, and video and in the case of “walled garden” inventory providers, may exclusively sell their own inventory directly to advertisers, which prevents us from competing with them entirely for such inventory.

p. 28 · Read in context →

Item 1A. Risk Factors — Third parties control our access to unique identifiers — p. 38 · Read the full section →

The dependency that sets this company apart from other ad platforms: identifiers it does not own or control.

Browser cookie policy, including Google's April 2025 reversal, and the UID2 hedge.

Today, three major web browsers—Apple’s Safari, Mozilla’s Firefox and Microsoft’s Edge —block third-party cookies by default. […] However, on April 22, 2025, Google announced that it would maintain its current approach to offering users third-party cookie choice in Chrome (thus, presumably, ending its efforts to deprecate third-party cookies in Chrome), and will not be rolling out a new standalone prompt for third-party cookies in Chrome. […] Although we believe our platform is well-positioned to adapt to browsers’ blocking or limitation of some cookies, particularly with our Unified ID 2.0 offering, the impact of such changes — and broader scrutiny on the advertising technology ecosystem — remains uncertain and could be more disruptive than we anticipate, including to the display advertising ecosystem in particular, where such changes could adversely impact our growth in that channel.

p. 39 · Read in context →

Item 7. MD&A — Executive Summary — p. 77 · Read the full section →

The scoreboard management chose: revenue against gross spend, plus the five opportunities it is underwriting.

The objectivity claim and the enumerated growth opportunities.

Our platform delivers valuable insights and results to clients without the conflict of interest and lack of objectivity that come with also selling owned advertising inventory. […] We believe that our key opportunities include (i) our ongoing global expansion, (ii) continuing development of our omnichannel ad inventory (including in channels such as CTV and other video, mobile, audio and others), (iii) continuing development, optimization and adoption of the data usage, measurement and targeting capabilities provided by our platform, which create a natural flywheel in our business, (iv) the adoption and utilization of third-party data, in particular, retail data, and first-party data by our clients, and (v) continuing development and incorporation of AI in our platform and related offerings.

p. 77 · Read in context →

Item 7. MD&A — Components of Our Results of Operations: Revenue — p. 81 · Read the full section →

Explains why revenue is a fraction of the money flowing through the platform, and why the balance sheet looks oversized.

One segment, agent accounting, and the gross-billings effect on receivables and payables.

We have one primary business activity and one operating segment. […] We generate revenue from clients who enter into agreements with us to use our platform to purchase advertising inventory, value-added services and data. We charge our clients for total spend on our platform, which includes spend and fees on advertising inventory, value-added services and data to support those purchases, in addition to the platform fee that is generally based on a percentage of our clients’ total spend on the platform. Generally, we report revenue as an agent on a net basis, which represents gross billings net of amounts we pay suppliers for the cost of advertising inventory, supplier-provided components of value-added services and data (collectively, “Supplier Components”).

Accounts receivable is recorded at the amount of gross billings to clients, net of allowances, for the amounts we are responsible to collect; and our accounts payable are recorded at the amount payable to suppliers. Accordingly, both accounts receivable and accounts payable appear large in relation to revenue reported on a net basis.

p. 81 · Read in context →

Item 7. MD&A — Results of Operations for the Year Ended December 31, 2025, Compared with the Year Ended December 31, 2024 — p. 85 · Read the full section →

Where management separates volume from price: gross spend grew 11% while revenue grew 18%.

Management's attribution of the 18% revenue increase.

Revenue increased by $451 million, or 18%, for the year ended December 31, 2025, as compared to the year ended December 31, 2024. The overall increase was driven by an 11% increase in gross spend on our platform, which was primarily driven by more overall advertising campaigns executed by new and existing clients, increased application of and changes in the mix of revenuegenerating value-added services and data and higher spend per campaign. The increase in revenue was also driven by a higher proportion of revenue earned from client spend due to increased utilization of our value-added services and data; and higher platform fees. Enhancements to our platform and the value-added services and data available to clients in 2025, including from Kokai and othe features, and increased pricing associated with value-added services and data, enabled both our clients and us to capture increased value and drove higher utilization of our value-added services and data.

p. 87 · Read in context →

Item 7. MD&A — Critical Accounting Policies and Estimates — p. 97 · Read the full section →

The net-versus-gross judgment is the accounting policy that defines the reported size of this business.

The principal-versus-agent test applied to advertising inventory and data.

We believe that the assumptions and estimates associated with the evaluation of revenue recognition criteria, including the determination of revenue recognition as net versus gross in our revenue arrangements, stock-based compensation expense and income taxes have the greatest potential impact on our consolidated financial statements. […] Generally, we report revenue net of amounts we pay suppliers for Supplier Components. Judgment is required to determine whether we are the principal and report revenue on a gross basis for Supplier Components or the agent and report revenue on a net basis for the amount of fees charged to the client. In this assessment, we consider if we obtain control of the specified service before it is transferred to the client, as well as other indicators such as the party primarily responsible for fulfillment, inventory risk and discretion in establishing price.

p. 99 · Read in context →

The Trade Desk, Inc. — FY2024 Annual Report (Form 10-K) — FY2024

Included for one reason: it holds the pre-AI self-description and the 26% growth baseline that the FY2025 rewrite is measured against. · Open the full document →

Item 1. Business — Overview — p. 6 · Read the full section →

The prior self-description — an ad-buying platform, not an AI company — against which the FY2025 rewrite reads as a repositioning.

FY2024 opening definition of the business.

The Trade Desk, Inc. (the “Company,” “we,” “our,” or “The Trade Desk”) offers a self-service, cloud-based ad-buying platform that empowers our clients to plan, manage, optimize and measure more expressive data-driven digital advertising campaigns. Our platform allows clients to execute integrated campaigns across ad formats and channels, including connected television (“CTV”) and other video, display, audio, and native, on a multitude of devices, such as televisions, streaming devices, mobile devices, computers and digital-out-ofhome devices. Our platform’s integrations with major inventory, publisher and data partners provide ad buyers reach and decisioning capabilities, and our enterprise application programming interfaces (“APIs”) enable our clients to customize and expand platform functionality.

p. 6 · Read in context →

More annual reports

The Trade Desk, Inc. — FY2023 Annual Report (Form 10-K) — FY2023 · 101 pages · The Kokai launch year and the last 10-K filed as a Delaware corporation. · Open →

The Trade Desk, Inc. — FY2022 Annual Report (Form 10-K) — FY2022 · 94 pages · First 10-K to carry the CEO Performance Option expense at full run-rate and the initial share repurchase authorization. · Open →

The Trade Desk, Inc. — FY2021 Annual Report (Form 10-K) — FY2021 · 127 pages · The Solimar-era baseline: display-first origins, pre-CTV-dominance channel mix and the year the CEO Performance Option was granted. · Open →

---

Competitors describe The Trade Desk, Inc.'s market in their own filings and calls. These verified passages and visual pages show where their strategies meet, using source documents preserved in Sources.

Amazon.com, Inc. (AMZN)

Amazon Ads runs Amazon DSP, the demand-side platform that bids for the same connected-TV, video and open-web impressions The Trade Desk's clients buy, and it has spent the last two years signing the CTV supply — Roku, Disney, Netflix, Spotify, SiriusXM — that independent DSPs also need. Only the advertising discussion is used here; AWS, Stores and devices are out of scope.

Amazon's own description of Amazon DSP and the supply it has locked to it. The Roku deal is presented as giving advertisers 80 million connected TV households — which Amazon calls the largest authenticated CTV footprint in the US — 'exclusively through Amazon DSP,' alongside a direct pipe into Disney's ad exchange. Both the household count and the 'largest' claim are Amazon's and unaudited, and exclusivity as described covers the Roku audience path rather than Roku inventory as a whole. Read against The Trade Desk, this is a competitor arguing that premium CTV reach is reachable on better terms inside its own buying platform.

Andrew R. Jassy, CEO, prepared remarks: Another area we're excited about is our demand-side platform, or Amazon DSP. Our DSP enables advertisers to plan, activate, and measure full-funnel investments. Our trillions of proprietary browsing, shopping, and streaming signals, paired with extensive supply-side relationships and our secure clean rooms, provide advertisers the ability to optimize advertising, deliver greater precision, and drive efficient and effective advertising outcomes. And in June, we announced a momentous partnership with Roku, giving advertisers access to 80 million connected TV households—the largest authenticated connected TV footprint in the U.S.—exclusively through Amazon DSP. It's a giant leap forward for advertisers, bringing best-in-class planning, audience precision, and performance to TV advertising. We also announced an integration between Disney's real-time ad exchange and Amazon DSP. This collaboration allows advertisers to gain direct access to Disney's premium inventory across platforms like Disney+, ESPN, and Hulu while allowing them to leverage insights from both companies.

p. 2 · Read in context →

Asked to break advertising growth into its parts, Amazon's CEO says the DSP gap-closing work is done: 'We have addressed customer feedback over the past 20 months and closed key gaps, making our DSP fully featured.' That is Amazon's self-assessment, not a third-party product review, and it is the single sentence most directly at odds with the argument that a full-featured independent DSP is hard to replicate. He pairs it with the Roku CTV position and Netflix, Spotify and SiriusXM inventory integrations.

Andrew Jassy, CEO, answering Colin Sebastian (Baird): Our demand-side platform, Amazon DSP, is also growing rapidly. We have addressed customer feedback over the past 20 months and closed key gaps, making our DSP fully featured. Our partnership with Roku provides the largest connected TV presence in the U.S. Furthermore, we have added integration opportunities with ad inventory from Netflix, Spotify, and SiriusXM for our DSP customers.

p. 10 · Read in context →

Scale and direction of travel for the advertising business Amazon points at the same budgets: $17.2bn of revenue in the quarter, up 22% year over year, with a cited Forrester ranking as a leader in omnichannel advertising platforms. The Netflix, Comcast local and Samsung interactive-video items show Amazon extending its demand into third-party CTV supply rather than only its own properties. The Forrester citation is Amazon's characterisation of a report not reproduced in the transcript.

Andrew R. Jassy, CEO, prepared remarks: Moving on to Amazon Ads. We continue working to be the best place for brands of all sizes to grow their businesses, and we are pleased with the continued strong growth across our full-funnel offerings, generating $17.2 billion of revenue in the quarter and up 22% year over year. Forrester recently recognized Amazon as a leader in omnichannel advertising platforms, with unmatched supply and insights for connected TV and commerce media. We deepened our Netflix partnership with Amazon Audiences, which enables advertisers to apply Amazon’s exclusive signals from shopping, browsing, and streaming to Netflix’s highly engaged viewers to reach the right audiences and drive even stronger performance. We also partner with Comcast to expand local advertising to thousands of brands, and expanded interactive video ad capabilities to partners starting with Samsung TVs.

p. 4 · Read in context →

Alphabet Inc. (GOOGL)

Google runs the other end-to-end alternative to an independent buying platform: Display & Video 360 on the demand side, Google Ad Manager on the supply side, and YouTube as the largest ad-supported living-room property. Its Google Network line is the closest public proxy for third-party open-web buying, and the DOJ ad tech remedies proceeding could restructure the supply chain The Trade Desk buys through. Search, Cloud and Other Bets are out of scope.

Google's stated CTV strategy: push the living room from a brand surface toward a measurable performance one, with checkout built into the TV screen via Buy with Google Pay. This is the same argument The Trade Desk makes for CTV budgets, advanced by the owner of the inventory rather than a neutral buyer. No reach or revenue figure is attached to the claim in the transcript.

Philipp Schindler, President & Chief Business Officer, prepared remarks: Looking at monetization across YouTube, we're driving sustained growth across our key priorities. In the living room, we see continued momentum across both brand and direct response. With the launch of Buy with Google Pay, viewers can complete purchases directly on their CTV, turning the TV screen into a stronger performance surface.

p. 4 · Read in context →

Google's own account of where the DOJ advertising-technology case stands: its advertiser tools and the DoubleClick and AdMeld deals were found not anticompetitive, but its publisher tools were held to have unfairly excluded rivals, with a remedies judgment pending and the DOJ seeking structural relief. This is the disclosure of the defendant, and it describes proposals rather than an outcome — but any structural change to Google's publisher stack reshapes the supply chain independent buyers route spend through.

Alphabet Inc. FY2025 Form 10-K, Risk Factors: In April 2025, the presiding judge issued a mixed decision in the DOJ case against us, ruling that neither our advertiser tools nor the DoubleClick and AdMeld acquisitions were anticompetitive, but that our publisher tools unfairly excluded rivals. A separate proceeding to determine remedies, the range of which varies widely, took place in September 2025 with the parties presenting differing remedy proposals. The DOJ's remedy proposal includes structural remedies that could harm our business. Closing arguments were held in November 2025, and we are awaiting a final judgment.

p. 20 · Read in context →

Viant Technology Inc. (DSP)

The closest like-for-like read on The Trade Desk: an independent, buy-side-only omnichannel DSP with a CTV emphasis, competing for the same agency and brand budgets. Viant names The Trade Desk in its 10-K competition section and returns to it repeatedly on calls, including a sustained argument that OpenPath compromises The Trade Desk's independence.

How a rival DSP defines the competitive set in its own filing: The Trade Desk as the public company 'exclusively serving our industry,' Yahoo DSP as the large private player, and Google and Amazon as divisions of larger firms. Viant characterises the market as fragmented but consolidating, with 'few scaled competitors' carrying self-service and autonomous-AI capability — a structural claim Viant makes for itself, unsupported by outside data in the filing.

Viant Technology Inc. FY2025 Form 10-K, Item 1 — Competition: Our industry is highly competitive and fragmented. We compete with large, privately-held companies, such as Yahoo DSP, with public companies exclusively serving our industry, such as The Trade Desk, and with divisions of large, well-established public companies such as Google and Amazon. The competitive landscape in recent years has been affected by consolidation and limited investment in new startups in our industry and there are currently few scaled competitors with self-service capabilities and AI-driven autonomous capabilities like those offered by ViantAI.

p. 14 · Read in context →

Asked how the DSP competitive environment is evolving, Viant's CEO puts The Trade Desk in the same frame as Google and Amazon: 'Google wants to sell you YouTube. Amazon wants to sell you Prime Video. And Trade Desk wants to redirect your spends through OpenPath, their own SSP where they are making incremental margins.' He then argues those moves make The Trade Desk 'no longer independent or objective when it comes to the pathways.' This is a competitor's characterisation of OpenPath's economics, not a documented fee analysis; the elision seams a garbled clause in the transcript.

Tim Vanderhook, CEO, answering Wyatt Swanson (D.A. Davidson): I mean, I view the competitive space as getting smaller and smaller. Trade Desk has made specific moves around OpenPath and charging for what used to be SSP territory. So we made in our prepared comments Google wants to sell you YouTube. Amazon wants to sell you Prime Video. And Trade Desk wants to redirect your spends through OpenPath, their own SSP where they are making incremental margins. Viant takes a different approach from that, and so we see less competition. You look at truly objective buy-side only platforms. Historically, there was The Trade Desk and ourselves. […] some of The Trade Desk's recent moves, that puts them more in the, I guess, no longer independent or objective when it comes to the pathways.

p. 9 · Read in context →

Viant's sizing of the independent buy side: after excluding Amazon, Google's DV360 and Yahoo DSP as platforms that also sell media, it counts the field of 'independent and objective enterprise-level buying platforms' at two — itself and The Trade Desk. The passage is Viant's positioning argument and the self-attribution charges against Amazon, Google and Yahoo are assertions without cited evidence; the count is useful mainly as a competitor's own definition of the category The Trade Desk sits in.

Chris Vanderhook, Co-Founder & COO, prepared remarks: Google's DV360 deploys the same self-attribution tactics as Amazon, claiming YouTube and Google Search are the only channels capable of driving sales on behalf of the advertiser. Yahoo DSP is no different, claiming the same self attribution across their publisher properties. Well, advertisers are finally waking up. They have learned you cannot trust the buying platform that also serves as a seller of ads because selling their own content will always take priority at the expense of the advertiser. We believe this leaves Viant and The Trade Desk as the two remaining independent and objective enterprise-level buying platforms in market

p. 4 · Read in context →

Nexxen International Ltd. (NEXN)

A partner and a competitor at once: Nexxen runs its own self-service DSP against The Trade Desk while supplying it with smart-TV inventory and ACR data. That dual role makes its disclosures unusually informative — it names The Trade Desk as the first DSP on its programmatic smart-TV home screen, and separately reports how an unnamed leading DSP customer's supply-path push cut into its own revenue.

Nexxen sizing a CTV surface that has not been programmatically buyable — the smart-TV home screen, on which it cites Nielsen for about ten minutes of viewing a day — and naming The Trade Desk as the first DSP to take it, routed into the Ventura ecosystem under a three-way agreement with Vidaa/V. Shown here as evidence of where new CTV supply is being created and who gets first access; the Nielsen figure and the 'first' claim are Nexxen's, and no spend or revenue is attached.

Ofer Druker, CEO, prepared remarks: According to Nielsen, viewers spend an average of about ten minutes per day on this screen deciding what to watch, making it a highly visible and valuable surface. Until now, advertising space on this page has been sold and managed through direct deals and ad servers. Our innovations transform this surface into a fully programmatic advertising opportunity. […] Vidaa, which rebranded as V, is a CTV operating system for Hisense and other OEM brands, and is our first OS partner to adopt this technology, which is now integrated across V-powered devices globally. As announced by The Trade Desk last week, we are pleased to welcome them as our first strategic DSP partner to adopt the solution following an agreement between V, The Trade Desk, and Nexxen International Ltd. to bring this inventory into The Trade Desk Ventura ecosystem. Together, we are collaborating to establish standardized DSP capabilities and drive industry awareness.

p. 9 · Read in context →

Nexxen cutting guidance and attributing part of it to 'a shift in our leading DSP customer reinforcing its SPO strategy' — supply-path optimisation pulling spend away from an intermediary. Nexxen does not name the customer anywhere in the transcript, and the identification should not be assumed — the only mention of The Trade Desk on this call comes from an analyst asking how the existing Trade Desk partnership is going, not from management. The response — doubling down on its own DSP and data to 'reduce third-party reliance' — is the competitive part: a supply partner building demand-side capability of its own.

Ofer Druker, CEO, prepared remarks: While we are encouraged by our momentum and strategic progress, we are disappointed to lower guidance due to near-term headwinds, including softness in select channels and a shift in our leading DSP customer reinforcing its SPO strategy.

That said, our platform's interconnected advanced technology solutions, TV data, and robust omnichannel media footprint give us confidence we can navigate these dynamics and emerge stronger in 2026 and beyond. Our strategy is evolving, not changing, as we are doubling down on our DSP, discovery, and broader data platform to drive enterprise adoption, strengthen end-to-end revenue opportunities, and reduce third-party reliance.

p. 10 · Read in context →

Criteo S.A. (CRTO)

Criteo competes for commerce and retail-media budgets from both sides: Commerce Max is a retail-media DSP and Commerce Grid an SSP that claims to be the programmatic route into retailer audiences for any DSP. It is also the peer furthest ahead on advertising inside an AI assistant, having been named OpenAI's first ad tech partner — a discovery surface that sits outside the open-internet auction.

Criteo's claim to first-mover position in advertising inside ChatGPT: OpenAI's first ad tech partner, with over 1,000 brands live. The brand count and the 'first' designation are Criteo's own and the revenue contribution is described elsewhere on the call as immaterial to guidance. For a demand-side platform, the relevance is the surface, not the size — ad inventory inside an assistant is not bought through the open real-time auction.

Michael Komasinski, CEO, prepared remarks: We entered 2026 with the ambition to lead in agentic AI, and we are already delivering on this ambition with discipline and focus. We became OpenAI's first ad tech partner, integrating our demand into ChatGPT's advertising offering with a focus on experiences that are relevant, additive, and built on user trust. This positions us at the forefront of a new high-intent Discovery Channel for our advertiser clients.

Momentum is building. We now have over 1,000 brands live with incremental budgets from both existing and new clients, strong agency traction, and early expansion across international markets.

p. 3 · Read in context →

A peer's sizing of the market both companies are chasing: AI-powered ad buying growing from roughly $35bn in 2025 to over $140bn by 2030, attributed to Madison and Wall. Criteo uses it to justify a self-service push into SMB budgets. The forecast is a third-party estimate quoted by an interested party, and 'AI-powered ad buying' is not defined in the transcript, so it is not directly comparable to programmatic spend measures.

Michael Komasinski, CEO, prepared remarks: Importantly, GO expands our addressable market, particularly among small and medium-sized businesses. This is supported by strong industry tailwinds with AI-powered ad buying expected to grow from approximately $35 billion in 2025 to over $140 billion by 2030, according to Madison and Wall.

We are already seeing strong interest and expect GO to be a multi-year growth driver. Clients running fully cross-channel campaigns are spending up to three times more, reinforcing the value of an integrated approach.

p. 5 · Read in context →

Criteo positioning Commerce Grid as 'the only programmatic path to Retail Media at scale,' open to any DSP, with over 30% of its commerce inventory buys already routed through it. The exclusivity claim is Criteo's and is contested by the retail-media offerings of other platforms; taken at face value it describes a toll position between advertisers' DSPs and retailer audiences — the same budgets The Trade Desk pursues through its own retail-media integrations.

Michael Komasinski, CEO, prepared remarks: A key differentiator for us is our Commerce Grid's supply side platform. It uniquely enables access to commerce audiences with full DSP interoperability and provides the only programmatic path to Retail Media at scale. It's a growing contributor to our business as agencies and brands ramp up investments to activate commerce audience deals and several leading retailers now use it to power offsite monetization. Today, over 30% of commerce growth inventory buys run through our SSP with meaningful upside as adoption continues to grow in the years ahead.

p. 4 · Read in context →

Adobe Inc. (ADBE)

Adobe Advertising is an enterprise demand-side platform that competed for the same video, display and search budgets, and Adobe's filings are the clearest public record of a software incumbent retreating from that fight: the DSP sits inside a segment Adobe labels legacy, shrinking and now dissolved. Creative Cloud and Digital Experience are out of scope.

Adobe classifies 'our Adobe Advertising offerings' — described in its FY2024 10-K as an end-to-end demand-side platform for video, display and search — among legacy solutions alongside eLearning and PostScript printing, and folds the segment that reported them into a single company-wide segment from Q1 FY2026. The filing gives no advertising-only revenue figure; the Publishing and Advertising segment as a whole was $256m of $23.8bn, down 7% year over year. The reclassification is disclosed as a management-reporting change, not as an exit.

Adobe Inc. FY2025 Form 10-K, Item 1 — Segments: Publishing and Advertising. Our Publishing and Advertising offerings contain legacy solutions including eLearning solutions, technical document publishing, web conferencing, document and forms platform, web App development, high-end printing through Adobe PostScript and Adobe PDF standards and our Adobe Advertising offerings.

Effective in the first quarter of fiscal 2026, we will combine our prior segments—Digital Media, Digital Experience and Publishing and Advertising—into a single operating and reportable segment due to changes in how management intends to evaluate results, allocate resources and execute the strategic opportunities outlined above. Accordingly, we will reflect this segment change in our Quarterly Report on Form 10-Q for the first quarter of fiscal 2026.

p. 18 · Read in context →

The write-down that accompanies the reclassification: a $70m non-cash goodwill impairment on the Publishing and Advertising reporting unit, the unit holding Adobe Advertising, in a quarter of record company revenue. Adobe does not attribute the impairment to the advertising business specifically, and the unit also contains eLearning, web conferencing and print technologies — so this bounds rather than proves the DSP's decline.

Steven Day, Interim CFO, prepared remarks: In Q2, Adobe achieved record revenue of $6.62 billion growing 13% year-over-year as reported and 11% in constant currency. Diluted earnings per share was $4.25 on a GAAP basis and $5.96 on a non-GAAP basis. Our GAAP results reflected a $70 million or $0.17 per share non-cash goodwill impairment charge related to our Publishing and Advertising reporting unit.

p. 6 · Read in context →

More peer documents

Q4_FY2025 — 15 pages · Needham's Laura Martin puts the three-way share question directly to Viant — DV360 third-party down 2%, The Trade Desk up 13%, Viant up 19% — and management answers with named head-to-head wins. · Open →

Q2_FY2025 — 12 pages · Viant's read on the mid-2025 Amazon-competition scare that hit The Trade Desk, including where it thinks Amazon's data advantage does and does not travel beyond CPG. · Open →

DSP_annual_report_FY2024 — 112 pages · Prior-year version of the same competition section, useful for testing whether a rival's description of the DSP competitive set has shifted year over year. · Open →

AMZN_annual_report_FY2025 — 80 pages · Amazon's advertising-services net sales line and its competition and risk language — the audited frame around the DSP claims made on the calls. · Open →

Q1_FY2026 — 12 pages · Google quantifies the divergence in one breath: YouTube ad revenue up 11% while Network advertising revenue is down 4% year on year. · Open →

Q1_FY2026 — 33 pages · Nexxen on the DSPs onboarding its home-screen inventory and its ACR data licensees, and on winning enterprise DSP clients — the partner-competitor boundary in its own words. · Open →

Q4_FY2025 — 16 pages · Criteo's two routes into offsite retail media — Commerce Max and opening retailer audiences to third-party DSPs — and how it expects the industry to focus there in 2026-27. · Open →

ADBE_annual_report_FY2024 — 100 pages · The FY2024 10-K wording that describes Adobe Advertising as "an end-to-end, demand-side platform," the baseline against which the FY2025 legacy classification reads. · Open →

---

Source: S&P Capital IQ consensus via Xpressfeed · Generated 2026-07-30.

FY2028 normalized EPS consensus down 23% over 180 days, revenue down 18% - but flat for a month

EPS has been cut harder than revenue at both years, so the street has taken down margin expectations as well as the top line. The 30-day columns are close to unchanged on all four rows, which is the first sign in this window that the downgrade run has stalled rather than reversed.

Currency: USD · Scale: money in millions, absolute · Point-in-time consensus; Δ90d is Now versus 90d.

Metric	FY	180d	90d	30d	Now	Δ90d
EPS (normalized)	FY2027	$2.46	$2.40	$2.15	$2.15	-10.4%
EPS (normalized)	FY2028	$2.92	$2.72	$2.24	$2.24	-17.7%
Revenue	FY2027	$3.87bn	$3.66bn	$3.50bn	$3.48bn	-4.8%
Revenue	FY2028	$4.51bn	$3.95bn	$3.76bn	$3.72bn	-5.8%

Revenue beat in seven of the last eight quarters, yet the latest EPS print missed by 12%

Current sequences by metric: Revenue: 5 consecutive beats; EPS (normalized): 1 consecutive miss.

Currency: USD · Scale: money in millions, absolute · Consensus is captured before each actual first became effective.

Quarter	Metric	Consensus	Actual	Surprise	Outcome
Q1 FY2026	Revenue	$678.67m	$688.86m	+1.5%	Beat
Q1 FY2026	EPS (normalized)	$0.32	$0.28	-12.4%	Miss
Q4 FY2025	Revenue	$840.68m	$846.79m	+0.7%	Beat
Q4 FY2025	EPS (normalized)	$0.58	$0.59	+1.7%	Beat
Q3 FY2025	Revenue	$719.34m	$739.43m	+2.8%	Beat
Q3 FY2025	EPS (normalized)	$0.44	$0.45	+1.8%	Beat
Q2 FY2025	Revenue	$686.03m	$694.04m	+1.2%	Beat
Q2 FY2025	EPS (normalized)	$0.41	$0.41	-0.9%	Miss
Q1 FY2025	Revenue	$575.28m	$616.02m	+7.1%	Beat
Q1 FY2025	EPS (normalized)	$0.25	$0.33	+33.2%	Beat
Q4 FY2024	Revenue	$759.56m	$741.01m	-2.4%	Miss
Q4 FY2024	EPS (normalized)	$0.57	$0.59	+3.6%	Beat
Q3 FY2024	Revenue	$620.46m	$628.02m	+1.2%	Beat
Q3 FY2024	EPS (normalized)	$0.39	$0.41	+4.3%	Beat
Q2 FY2024	Revenue	$578.12m	$584.55m	+1.1%	Beat
Q2 FY2024	EPS (normalized)	$0.36	$0.39	+9.5%	Beat

Roughly high-single-digit revenue growth, with gross margin sliding about five points to FY2029

Consensus gross margin falls about five points between FY2025 and FY2029, pausing only between FY2026 and FY2027, and EBITDA grows slower than revenue at the back of the window. Free cash flow is the odd row: consensus has it falling in FY2026 even as revenue rises about 10%, with mean capex higher in FY2026 than in FY2025.

Currency: USD · Scale: money in millions, absolute · YoY uses the prior fiscal year from the feed; analyst count and range use the first displayed period.

Metric	FY2025A	FY2026E	FY2027E	FY2028E	FY2029E	YoY	Analysts	Low / high
Revenue	$2.89bn	$3.18bn	$3.48bn	$3.72bn	$4.05bn	—	37	$2.85bn / $2.91bn
EBITDA	$1.17bn	$1.26bn	$1.38bn	$1.47bn	$1.55bn	—	36	$1.13bn / $1.24bn
Gross margin	78.7%	76.7%	76.8%	75.9%	73.5%	—	—	—
EPS (normalized)	$1.76	$1.85	$2.15	$2.24	$2.33	—	20	$1.68 / $1.84
Free cash flow	$823.05m	$780.00m	$883.32m	$966.97m	$1.10bn	—	—	—

FY2027 EBITDA still spans 821 to 1,514 with 34 analysts covering it

On FY2027 EBITDA and FY2028 revenue the median sits above the mean, so the width comes from a handful of low outliers rather than an evenly split street. The FY2028 normalized EPS row rests on 10 analysts - read its range as thin coverage, not as a settled disagreement.

Currency: USD · Scale: money in millions, absolute · Spread/mean is absolute high-low divided by absolute mean.

Metric	Period	Mean	Low–high	Spread/mean	Analysts
EBITDA	FY2027E	$1.38bn	$821.00m–$1.51bn	50.1%	34
Revenue	FY2028E	$3.72bn	$2.41bn–$4.05bn	44.1%	20
Net income (GAAP)	FY2027E	$577.89m	$279.32m–$699.60m	72.7%	25
EPS (normalized)	FY2028E	$2.24	$1.78–$2.57	35.2%	10

Street snapshot

Currency: USD · Scale: money in millions, absolute · Analyst counts shown explicitly.

Street view	Reading	Analysts
Recommendation mix	Buy 11, Outperform 2, Hold 19, Underperform 1, Sell 3	36
Consensus score	2.53	36
Target price	mean $24.32; median $24.50; high $38.00; low $11.00	30

Coverage caution: FY2029 rests on a handful of analysts

FY2029 revenue carries 7 estimates, EBITDA 6 and normalized EPS 4, against 35 for revenue and 22 for normalized EPS in FY2026. FY2028 normalized EPS rests on 10. Treat the FY2028-29 columns and their ranges as indicative rather than a consensus.

---

Visible Alpha broker models via S&P Xpressfeed · 31 brokers · 319 line items · freshest revision 2026-07-27.

Thirty-one brokers model The Trade Desk, and they agree tightly on the revenue line — about $3.2bn in FY-2026 and $3.5bn in FY-2027 — while disagreeing on how it is earned. The models take gross spend on the platform from roughly $14bn in FY-2025 to $18bn in FY-2028, with take rate supplying most of the FY-2026 lift and volume most of the FY-2027 lift. Underneath that, Visual-and-TV and audio spend account for the growth: mobile in-app and web spend falls in FY-2026 and never regains its FY-2025 level. The soft spots are delivery cost and cash conversion, where broker ranges are far wider than on revenue.

Take rate adds about a point in FY-2026; volume takes over in FY-2027 as gross spend grows about 11%

Revenue growth barely moves across the two years, so the change is all in the mix — price of the platform first, then throughput. The quarterly path shows the same reset: take rate builds to about 21.6% by 4QFY-2026, then steps back to 20.7% in 1QFY-2027.

Line	FY-2025A	FY-2026E	FY-2027E	FY-2028E	YoY	Brokers
Gross spend	$14.14bn	$14.91bn	$16.60bn	$17.86bn	+5.4%	22
Take rate(%)	20.5%	21.4%	21.1%	21.3%	+0.9pt	24
Revenue	$2.89bn	$3.18bn	$3.48bn	$3.74bn	+10.0%	31

Visual-and-TV and audio spend take all the growth; mobile in-app and web never regains its FY-2025 level

Line	FY-2025A	FY-2026E	FY-2027E	FY-2028E	YoY	Brokers
Growing	—	—	—	—	—	—
Gross spend - Visual and TV	$6.95bn	$7.67bn	$8.74bn	$9.77bn	+10.4%	15
Gross spend - Audio	$728.67m	$880.46m	$1.02bn	$1.21bn	+20.8%	14
Flat to down	—	—	—	—	—	—
Gross spend - Mobile In- App & Web	$4.84bn	$4.37bn	$4.57bn	$4.66bn	-9.7%	14
Gross spend - Display, Social, Native and others	$1.54bn	$1.70bn	$1.72bn	$1.97bn	+10.3%	15

Platform-operations cost runs well ahead of revenue in FY-2026; the sales and R&D lines do not

Line	FY-2025A	FY-2026E	FY-2027E	FY-2028E	YoY	Brokers
Cost lines	—	—	—	—	—	—
Platform operations	$625.99m	$735.24m	$809.11m	$881.53m	+17.5%	25
Sales and marketing	$641.46m	$696.29m	$750.93m	$796.77m	+8.5%	24
Technology and development	$538.52m	$572.54m	$611.24m	$649.13m	+6.3%	24
Margin	—	—	—	—	—	—
EBITDA margin(%)	40.6%	39.7%	39.9%	40.4%	-0.9pt	31
Cash	—	—	—	—	—	—
Capital expenditures	$216.32m	$285.15m	$255.06m	$280.98m	+31.8%	27
Free cash flow	$783.03m	$782.36m	$883.77m	$978.15m	-0.1%	27
Free cash flow per share($)	$1.58	$1.64	$1.85	$2.05	+4.1%	27

Brokers agree on revenue and split on take rate, buybacks and free cash flow

Line	Period	Median	Q1–Q3	Min–max	Brokers
Take rate(%)	FY-2027E	21.4%	20.8%–21.6%	18.8%–22.6%	24
Repurchases of common stock	FY-2026E	$163.51m	$163.51m–$613.51m	$159.01m–$1.04bn	17
Free cash flow	FY-2027E	$866.59m	$767.16m–$1.00bn	$453.56m–$1.22bn	27
Gross spend - Mobile In- App & Web	FY-2027E	$4.60bn	$4.37bn–$4.86bn	$3.53bn–$5.68bn	11
Gross spend - Visual and TV	FY-2028E	$9.58bn	$9.17bn–$10.40bn	$8.12bn–$11.58bn	12

The differentiated KPI lines rest on far fewer brokers than the P&L

Revenue and EBITDA carry 29 to 30 brokers through FY-2027 and 19 in FY-2028, but the channel splits carry 10 to 15, the US/international billings split 6 to 9, and total clients only 3 to 5. The one customer-retention line in the feed is a single broker's, last revised 2022-02-14 for FY-2027 and FY-2028 — one analyst's placeholder, not consensus.

Headline P&L consensus, momentum and beat/miss live in the CapIQ tab.

---

Source: S&P Capital IQ transcripts via Xpressfeed · latest indexed call 2026-05-07 · generated 2026-07-30.

Latest call digest

The Trade Desk, Inc., Q1 2026 Earnings Call, May 07, 2026 · 2026-05-07T21:00:00

Q1 2026 earnings call — May 07, 2026. Prepared remarks and Q&A pointed in opposite directions. Jeff Green spent the bulk of his script on the size and shape of the opportunity — a $1 trillion TAM, a record supply-demand imbalance he calls "the biggest buyers' market in the history of advertising," the case that measurement is broken and that fixing it favors the open internet, and a run through partners (Disney, Spotify, NBCU, Netflix) and products (Audience Unlimited, retail data, agentic AI with Stagwell, Lyft Ads, Dollar General). Commercial proof points were real: 45 JBPs signed in March, total JBP count up 55% year-over-year, new JBP deal spend up 40% excluding renewals, and a pharma account won back from Amazon with a 2026 JBP that lifts spend 114% year-over-year.

Interim CFO Tahnil Davis reported revenue of $689 million, up 12% year-over-year, with $206 million of adjusted EBITDA (30% margin). Video including CTV was a low-50s percent of the business; audio was around 6% and grew faster than any other channel; the U.S. was approximately 82% of revenue. Guidance given: Q2 revenue of at least $750 million, Q2 adjusted EBITDA of approximately $260 million, full-year 2026 adjusted EBITDA margin of at least 40% and approximately in line with 2025, and headcount growth below revenue growth.

The Q&A reality. Neither of the two topics that opened Q&A appeared anywhere in the prepared remarks: the Publicis negotiation and the departure of Chief Strategy Officer Samantha Jacobson to OpenAI, which the trade press had reported hours before the print. Four of the nine analyst questions pressed on the Q2 outlook, its causes, or the path back to faster growth, and three of those used the word deceleration. Green attributed the deceleration to macro pressure on Fortune 500 brands — geopolitical instability, tariffs, consumer pressure — and said the near-term answer is execution rather than reinvention. On Publicis he said negotiations are ongoing and that it is "probably not prudent" to say more. On the record March JBP signings he declined to say whether they relate to the agency dispute. He drew a firm line against ever moving to the sell side, and framed LLM and AI-search advertising as a real TAM unlock still in "the first inning." The most notable omission relative to prior calls: Kokai, UID2, and Deal Desk went essentially unmentioned.

Participant coverage from the latest call.

Group	Participants	Count
Management	Operator; Chris Toth — Vice President of Investor Relations, The Trade Desk, Inc.; Jeffrey Green — Co-Founder, CEO, President & Chairman, The Trade Desk, Inc.; Tahnil Davis — Chief Accounting Officer & Executive VP, The Trade Desk, Inc.	4
Analysts	Shyam Patil — Senior Analyst, Susquehanna Financial Group, LLLP, Research Division; Vasily Karasyov — Founder, Cannonball Research, LLC; Matthew Swanson — Analyst, RBC Capital Markets, Research Division; Justin Patterson — MD & Equity Research Analyst, KeyBanc Capital Markets Inc., Research Division; Mark Zgutowicz — Senior Equity Analyst, The Benchmark Company, LLC, Research Division; Youssef Squali — Head of Internet, Truist Securities, Inc., Research Division; Timothy Nollen — Research Analyst, SSR LLC; Jessica Reif Cohen — Managing Director in Equity Research, BofA Securities, Research Division; Jason Helfstein — MD & Senior Internet Analyst, Oppenheimer & Co. Inc., Research Division	9

Curated latest-call exchanges; one row per analyst topic.

Analyst	Firm	Topic	What changed in Q&A
Shyam Patil	Susquehanna	Publicis negotiation and the Q2 deceleration	Green said the conflict has been framed in the most conflict-rich language the press can provide and called it overdramatized; confirmed negotiations are ongoing and declined further detail. On Q2 he pointed to macro pressure on Fortune 500 brands rather than anything company-specific, and said several fast-growing verticals would grow faster absent tariffs and geopolitical uncertainty.
Vasily Karasyov	Cannonball Research	Chief Strategy Officer's departure to OpenAI	Karasyov noted the trade-press story landed the same day, before the print. Green confirmed the move, said Jacobson stays on the Board of Directors, and pivoted to senior hires he says have been assembled quietly. No successor or scope change was described.
Matthew Swanson	RBC Capital Markets	Cyclical versus structural drivers of re-acceleration	Green separated the two, said structural drivers are extremely strong and that re-acceleration is about executing against an expanding opportunity, not reinventing the company. He conceded the industry pressures do not show up in results today and that stabilizing macro would be a tailwind that is not present now. No timing was offered.
Justin Patterson	KeyBanc	Levers to the full-year EBITDA margin target after a softer first half	Patterson noted both revenue and margins started the year soft against the at-least-40% target. Davis reiterated headcount growth below revenue growth, investment concentrated in platform innovation, AI, retail media and measurement, and flexibility in pacing spend. Green added that 2026 is a year of disciplined reinvestment. No quantified bridge to the target was given.
Mark Zgutowicz	Benchmark	Whether agency weakness or one-time items explain the Q2 guide	The hardest exchange on the guide. Zgutowicz asked directly whether one-time or one-to-two-quarter items sit inside a guide he characterized as below industry growth expectations. Green said there is nothing incremental to add on the agency front and did not address the one-time-items question; he then agreed CPG and auto comps get easier and argued the discipline those categories have adopted matters more than the comps.
Youssef Squali	Truist Securities	LLM and AI-search advertising opportunity and gating factors	Green declined to discuss talks with the key players. He compared the chatbots to Netflix five to ten years ago, argued expensive content forces ad monetization, and said detailed prompts support formats beyond keywords, potentially including video. He placed the opportunity in the first inning, "a couple of pitches in, max." No P&L timing.
Timothy Nollen	SSR	Whether The Trade Desk should add sell-side services; OpenTTD	Green identified OpenTTD as the hub he referenced in prepared remarks, then ruled out yield management outright, saying serving two masters is the flaw of the ad-network model and that they will never do it. He said OpenPath exists to plug into publishers running their own yield tech, and tied supply-chain inefficiency to why 2026 is a reinvestment year.
Jessica Reif Cohen	BofA Securities	When agentic trading becomes the dominant model	Green rejected the framing of being impacted by AI and said The Trade Desk will lead the agentic shift. He criticized rivals for connecting single advertisers to single publishers, which he argues recreates ad networks and forfeits holistic decisioning. The Stagwell work starts with creating and editing campaigns and is expected to move to optimization. No adoption timeline.
Jason Helfstein	Oppenheimer	Agentic path — technology versus commercial terms; whether record JBPs relate to the agency dispute	Green said the optimization variables are more the problem being solved than commercial terms, with frameworks set in advance and repeated by agents at scale. On the yes-or-no JBP question he said he cannot comment on whether the March signings are relevant to the agency discussions.

Theme tracker

Themes are curator-classified across supplied calls.

Theme	Status	Quarters mentioned	Read-through
CTV as the largest and fastest-growing channel	persisted	Q2 2023, Q3 2023, Q4 2023, Q1 2024, Q2 2024, Q3 2024, Q4 2024, Q1 2025, Q2 2025, Q3 2025, Q4 2025, Q1 2026	Present on every call in the window, but the framing has thinned. Through 2024 CTV carried multi-paragraph treatment and explicit growth claims; by Q1 2026 it appears mainly as a bullet in a list of investment areas, with the channel disclosure given as video in a low-50s percent share. Q4 2025 was the last call to state plainly that CTV grew faster than the overall business.
Buyer's market and the supply-demand imbalance	persisted	Q4 2023, Q1 2024, Q2 2024, Q4 2024, Q1 2025, Q2 2025, Q3 2025, Q4 2025, Q1 2026	Management's central strategic argument, and it has strengthened rather than faded. Q1 2026 escalated it to the biggest buyers' market in the history of advertising. Worth noting the argument cuts both ways: an oversupplied market is also a market where inventory prices fall, and revenue growth has decelerated across the same stretch in which the imbalance widened.
Objectivity and not owning inventory as the core differentiator	persisted	Q2 2023, Q3 2023, Q4 2023, Q1 2024, Q2 2024, Q3 2024, Q4 2024, Q1 2025, Q2 2025, Q3 2025, Q4 2025, Q1 2026	The most stable message in the entire history. It is unchanged from the 2023 calls, which makes it a poor source of new information but a reasonable read on management's strategic conviction. The Q1 2026 twist is that objectivity is now framed as the enabler of AI decisioning rather than only as a trust argument.
Amazon and DSP competitive intensity as a Q&A topic	persisted	Q4 2023, Q1 2024, Q2 2024, Q3 2024, Q4 2024, Q1 2025, Q2 2025, Q3 2025, Q4 2025	Analysts raised Amazon, its DSP, DV360 or Prime Video in at least one question on nine consecutive calls from Q4 2023 through Q4 2025, and Green's answer barely varied: Amazon's DSP is a distant priority behind sponsored listings and Prime Video, so the two are not really competing. Q1 2026 is the first call in that run with no analyst question on it, with attention shifting to agencies and the guide instead.
Agentic AI	emerged	Q2 2025, Q3 2025, Q4 2025, Q1 2026	First surfaced as a passing aside in Q2 2025, formalized as a Kokai trading-modes copilot in Q3 2025, and by Q4 2025 and Q1 2026 it is the dominant narrative — the Stagwell partnership, an agentic framework for partners, and a rebuttal to the view that AI disintermediates platforms. It is the clearest case of a new theme displacing older product stories.
CPG and automotive vertical weakness	emerged	Q2 2025, Q3 2025, Q4 2025, Q1 2026	Introduced obliquely in Q2 2025 as tariff-driven volatility in auto and CPG, named as a tale of two cities in Q3 2025, and elevated in Q4 2025 to a headline explanation with the disclosure that the two categories are roughly a quarter of the business and that growth would have been at least 5% higher without them. Still present in Q1 2026 as pressure in Home & Garden and Food & Drink.
Organizational upgrade and senior leadership turnover	emerged	Q4 2024, Q1 2025, Q2 2025, Q3 2025, Q4 2025, Q1 2026	Began with the Q4 2024 miss and the largest reorganization in company history, and has been a standing agenda item since — a new COO, CFO and CRO in 2025, a second CFO change by Q4 2025 with an interim in the seat, and a Chief Strategy Officer departure in Q1 2026. Six consecutive calls of leadership change is itself the signal; the JBP and go-to-market metrics are management's evidence that it is working.
UID2 and the identity framework	dropped	Q2 2023, Q3 2023, Q4 2023, Q1 2024, Q2 2024, Q3 2024, Q4 2024, Q1 2025, Q2 2025, Q3 2025, Q4 2025	A dominant theme through 2023 and 2024, when it anchored whole sections of prepared remarks and the cookie-deprecation debate, appearing in nine separate components on the Q2 2023 and Q4 2023 calls. It thinned steadily through 2025, down to a single passing reference in Q4 2025, and is absent from Q1 2026 entirely. Management's account is that UID2 became ubiquitous and stopped needing airtime; either way, a former headline differentiator has left the narrative. The absence itself is only one call deep, so treat it as a fade rather than a reversal.
Kokai platform migration	dropped	Q2 2023, Q3 2023, Q4 2023, Q1 2024, Q2 2024, Q3 2024, Q4 2024, Q1 2025, Q2 2025, Q3 2025, Q4 2025	Discussed on eleven straight calls, peaking in Q1 and Q2 2025 when adoption percentages and performance deltas were the core of the growth case, and not mentioned at all in Q1 2026. The migration completing is the benign reading. The less benign one is that the performance case built on Kokai case studies has not translated into the revenue re-acceleration it was expected to fund. As with UID2, the absence is one call deep.
Google antitrust and Google exiting the open internet	dropped	Q2 2023, Q1 2024, Q2 2024, Q3 2024, Q4 2024, Q1 2025, Q2 2025, Q3 2025	A recurring bull argument that peaked in Q1 2025, when Green called two guilty verdicts a major victory for the open internet and said no company would benefit more. It was still live in Q3 2025 ahead of closing arguments, then vanished from both Q4 2025 and Q1 2026. A predicted tailwind that stopped being discussed rather than one that was declared realized, and the only theme here with two consecutive calls of absence.
Supply-chain efficiency through OpenPath and adjacent tools	persisted	Q2 2023, Q3 2023, Q4 2023, Q1 2024, Q2 2024, Q3 2024, Q4 2024, Q1 2025, Q2 2025, Q3 2025, Q4 2025, Q1 2026	Continuous across all twelve calls but with a changed tone. Through Q1 2025 the treatment was expansionary, with publisher fill-rate and revenue case studies and a prediction that 2025 would be OpenPath's steep S-curve year. By Q4 2025 Green was defending it against trade-press criticism, and in Q1 2026 it appears only as the reason the company will not move to the sell side.

Guidance ledger

Quotes, calls, and speakers are source-verified; outcomes are curator-classified.

Verbatim guidance	Call	Speaker	Curator outcome	Outcome note
“We estimate Q3 revenue to be at least $618 million, which would represent growth of 25% on a year-over-year basis.”	The Trade Desk, Inc., Q2 2024 Earnings Call, Aug 08, 2024 · 2024-08-08T21:00:00	Laura Schenkein	kept	The Q3 2024 call reported revenue of $628 million and growth of 27% year-over-year.
“We estimate Q4 revenue to be at least $756 million, which would represent growth of about 25% on a year-over-year basis.”	The Trade Desk, Inc., Q3 2024 Earnings Call, Nov 07, 2024 · 2024-11-07T22:00:00	Laura Schenkein	missed	Q4 2024 revenue came in at $741 million, up 22%. Management called it the first shortfall against its own expectations in 33 quarters as a public company and attributed it to execution missteps rather than market conditions.
“We estimate adjusted EBITDA to be approximately $363 million in Q4.”	The Trade Desk, Inc., Q3 2024 Earnings Call, Nov 07, 2024 · 2024-11-07T22:00:00	Laura Schenkein	missed	Q4 2024 adjusted EBITDA was reported at $350 million, a 47% margin, below the guided figure alongside the revenue shortfall.
“We expect revenue to be at least $575 million, reflecting 17% year-over-year growth.”	The Trade Desk, Inc., Q4 2024 Earnings Call, Feb 12, 2025 · 2025-02-12T22:00:00	Laura Schenkein	kept	Q1 2025 revenue was $616 million, up 25%, which management described as far surpassing its own expectations.
“I expect that all of our clients will be using Kokai exclusively.”	The Trade Desk, Inc., Q4 2024 Earnings Call, Feb 12, 2025 · 2025-02-12T22:00:00	Jeffrey Green	unknown	Green set this for well before the end of 2025. Q3 2025 reported nearly 85% using Kokai as their default experience and Q4 2025 said almost 100% of clients are running through Kokai, so the calls do not establish whether full exclusivity was reached.
“in Q2, we expect revenue to be at least $682 million, reflecting 17% year-over-year growth”	The Trade Desk, Inc., Q1 2025 Earnings Call, May 08, 2025 · 2025-05-08T21:00:00	Laura Schenkein	kept	Q2 2025 revenue was $694 million, up 19% year-over-year.
“we expect Q3 revenue to be at least $717 million, reflecting 14% year-over-year growth”	The Trade Desk, Inc., Q2 2025 Earnings Call, Aug 07, 2025 · 2025-08-07T21:00:00	Laura Schenkein	kept	Q3 2025 revenue was $739 million, up 18% year-over-year, or approximately 22% excluding prior-year political spend.
“For Q4, we expect revenue to be at least $840 million.”	The Trade Desk, Inc., Q3 2025 Earnings Call, Nov 06, 2025 · 2025-11-06T22:00:00	Alex Kayyal	kept	Q4 2025 revenue was $847 million, up 14%, or approximately 19% excluding prior-year political spend.
“For the first quarter, we expect revenue to be at least $678 million, representing 10% year-over-year growth.”	The Trade Desk, Inc., Q4 2025 Earnings Call, Feb 25, 2026 · 2026-02-25T22:00:00	Tahnil Davis	kept	Q1 2026 revenue was $689 million, up 12% year-over-year.
“We estimate adjusted EBITDA for Q1 to be approximately $195 million.”	The Trade Desk, Inc., Q4 2025 Earnings Call, Feb 25, 2026 · 2026-02-25T22:00:00	Tahnil Davis	kept	Q1 2026 adjusted EBITDA was $206 million, a 30% margin.
“For Q2, we expect revenue to be at least $750 million.”	The Trade Desk, Inc., Q1 2026 Earnings Call, May 07, 2026 · 2026-05-07T21:00:00	Tahnil Davis	pending	Guided on the most recent call; no subsequent call in the supplied history. Three analysts characterized the implied growth rate as a deceleration.
“we continue to expect our full year 2026 adjusted EBITDA margin percentage to be at least 40%, approximately in line with 2025”	The Trade Desk, Inc., Q1 2026 Earnings Call, May 07, 2026 · 2026-05-07T21:00:00	Tahnil Davis	pending	Reaffirms the Q4 2025 framing and adds the at-least-40% figure. Q1 2026 came in at a 30% margin, so the target implies materially higher margins across the remaining quarters.

Q&A pressure map

Question counts and firms are curator tallies; analyst coverage shown above.

Topic	Questions	Firms	Pressure / response
Growth deceleration and the credibility of the guide	9	Susquehanna, Cannonball Research, KeyBanc, Oppenheimer, RBC Capital Markets	The most persistent line of pressure, running from the Q4 2024 miss through Q1 2026. It began as forensic questions about what went wrong, became questions about whether a 20%-plus growth rate is sustainable, and by Q1 2026 had turned into three separate questions about the Q2 guide and the path back to faster growth. Management's answers have been consistent in substance and consistently free of timing.
Amazon and competitive intensity across the DSP landscape	13	Truist Securities, Cannonball Research, Macquarie, Wolfe Research, Oppenheimer, RBC Capital Markets, Stifel, Susquehanna, BofA Securities, KeyBanc	Counting analyst question turns across the last eight calls that raise Amazon, its DSP, DV360, Prime Video or the DSP competitive landscape. Asked on every call from Q2 2024 through Q4 2025, and by Q3 2025 Green said outright he had been hoping for the question. His framing has been stable: Amazon's advertising is overwhelmingly sponsored listings and Prime Video, its DSP is a distant priority, and objectivity conflicts prevent it from competing on decisioned open-internet buying. Analysts kept asking, which suggests the answer has not fully settled the issue.
AI and agentic disruption risk	7	KeyBanc, Oppenheimer, Truist Securities, BofA Securities	Evolved from questions about Kokai's AI returns in Q2 2025 to explicitly adversarial framings by Q4 2025, when Oppenheimer put the super-bear view that an agentic future makes brands irrelevant and asked whether The Trade Desk can scale AI against Amazon and DV360. Green engaged the brands argument directly and answered the scale question with trust and data access rather than compute.
Supply-chain positioning and OpenPath	6	BofA Securities, Wells Fargo Securities, SSR	Early questions were opportunity-framed, asking when OpenPath would scale. The tone changed in Q4 2025 when Wells Fargo cited press reports of discomfort over transparency and perceived conflicts of interest, and again in Q1 2026 when SSR asked whether independence still makes sense and whether sell-side services are coming. Green called some of the assertions ridiculous and drew a hard line against yield management.
Organizational change and leadership turnover	5	Susquehanna, KeyBanc, Cannonball Research, Evercore ISI	Recurring since the Q4 2024 reorganization, with analysts repeatedly asking what has actually changed and where the results are. Cannonball's Q4 2025 question explicitly held management to the prior year's reorganization claims. Answers have relied on JBP pipeline and count as the evidence of progress.
CPG and automotive softness	4	Cannonball Research, Susquehanna, Oppenheimer, Benchmark	Concentrated in the last four calls. Analysts have pushed past the category description toward the harder question — what happens to growth if the weakness persists. Green has answered mainly by arguing the pressure makes those brands more disciplined and therefore better customers, which does not directly address the modelling question being asked.
Agency relationships and the Publicis dispute	3	Susquehanna, Benchmark, Oppenheimer	New in Q1 2026 and the sharpest exchange of the call, with three analysts approaching it from different angles. All three were substantially deflected: negotiations described as ongoing with no further detail, nothing incremental offered on whether agency weakness sits in the Q2 guide, and no comment on whether the record March JBP signings connect to the dispute. Green said he hoped the call would end the public discussion, which is a statement of intent rather than an answer.

Language shifts

Only language evidence verified against the referenced component is shown.

Observation	Verbatim evidence	Call ID	Component
The opening characterization of the quarter has softened from strong to solid. Q3 2025 and Q2 2025 both opened with having again posted strong growth; Q4 2025 and Q1 2026 both open with a solid quarter.	“As you've seen from our press release, we delivered a solid quarter once again.”	1996530678	2
Tahnil Davis's closing sentence is nearly a verbatim repeat of the Q1 2025 version with one phrase removed. In Q1 2025 Laura Schenkein said the company was confident in its ability to outpace the market and capitalize on the opportunities ahead; in Q1 2026 the outpace-the-market clause is gone and only capitalizing on the opportunity remains. Market-share outperformance was the single most repeated claim of the 2024 calls.	“we remain confident in our ability to capitalize on the significant opportunities ahead of us”	1996530678	3
Disciplined reinvestment is new vocabulary that entered in Q4 2025 and was repeated three times in Q1 2026, by both the CEO and the interim CFO. It functions as advance framing for margin pressure in a year guided to a margin only approximately in line with the prior year.	“Looking ahead, it's important to think about 2026 as a year of disciplined reinvestment.”	1980876042	7
Visibility language appeared for the first time in Q4 2025, replacing the earlier construction in which guidance was conditioned on the macro remaining stable. The caution is now located inside the company's own forecasting confidence rather than in an external assumption.	“Our Q1 guidance reflects a prudent approach in an environment, where visibility remains somewhat lower, particularly in CPG and to a lesser extent, auto verticals.”	1980876042	3
Q1 2026 introduced explicit near-term headwind and cloudier-macro phrasing in the same breath as the long-term case. Earlier calls tended to reframe macro pressure as an opportunity to grab land or gain share; this construction concedes the near term before pivoting.	“And so while there are clearly near-term headwinds and a cloudier macro environment, we continue to believe that the long-term opportunity for our business remains extremely strong.”	1996530678	6
The CEO has begun conceding the quality of results directly rather than only in the context of a specific miss. This is a marked change from the 2023 and 2024 calls, where results commentary was uniformly superlative.	“Despite the fact that I don't think that this is our best earnings report ever, I hope you can hear it that I am as optimistic as ever”	1980876042	10

Twelve calls show a strategic story that has barely changed and a growth rate that has come down a long way inside it. That combination is the crux of the debate: management's explanation is macro and cyclical, centred on CPG and auto, and the record JBP signings support the view that demand is intact. The harder reading is that the product cycle that was supposed to fund re-acceleration has gone quiet in the prepared remarks, the market-share-outperformance claim has been dropped from the CFO's closing language, and the newest pressure — the agency relationship — is the one management is least willing to discuss.

---

Toll on Ad Spend

The Trade Desk sells no advertising inventory. It rents advertisers a buying platform and keeps a fee on whatever they spend through it — $2.90 billion of revenue in FY2025 on $13.39 billion of client spend, a 21.6% cut [1]. The business is profitable, debt-free and generates cash. The stock is down 87.6% from its December 2024 peak. The gap between those two sentences is what this report is about.

What the company does

Digital advertising is bought in two broadly different ways. Inside the "walled gardens" — Google, Meta, Amazon's own retail properties — an advertiser buys inventory the platform itself owns and sells. Everywhere else, on what the industry calls the open internet, inventory is auctioned. The Trade Desk is a demand-side platform, or DSP [2]: the software an ad buyer uses to bid in those auctions across connected TV, video, mobile, display and audio [3].

Its clients are advertising agencies and brands, contracted under master services agreements, and the revenue mechanic is simple: "a platform fee generally based on a percentage of our clients' total platform spend," plus fees for data and value-added services layered on top [4]. The company owns no inventory, which is the whole pitch — it argues it can pick the best impression for a buyer because it has no impressions of its own to favour [5].

The addressable pool is large and growing. Digital advertising is reported at over $700 billion of annual spend and more than 70% of the total advertising market; global advertising TAM passed $1 trillion for the first time in 2024; and viewing continues to migrate from linear television to connected TV [6]. That is the tailwind the equity story has always leaned on.

Source: Q1 2026 prepared remarks, CFO commentary — video "a low-50s percent," mobile "a high-28s percent," display "a low double-digit share," audio "around 6%"; midpoints plotted [7].

Two concentrations matter. Video, which includes CTV, is roughly half the business and still rising as a share of mix; and the United States was about 82% of Q1 FY2026 revenue, leaving international at 18% despite a decade of investment [8]. The company ended FY2025 with 3,843 full-time employees in 21 countries, and its own competition disclosure names Google and Amazon as the large, well-established rivals [9].

Client stickiness has historically been the franchise's best evidence. The FY2024 10-K reports a client retention rate above 95% in each of the last eleven years — though the same paragraph notes those master agreements typically run one year and are terminable on 60 days' notice, so retention is a behavioural fact rather than a contractual one [10].

Volume and the cut

Because the fee is a percentage of spend, the business has two independent drivers: how much money flows across the platform, and what share of it the company keeps. Management discloses the first as gross spend and has disclosed the second, historically, as take rate.

Through FY2024 those two moved together. Gross spend compounded at 24–25% a year for three consecutive years and the take rate sat in a narrow band around 20.3% [11][12][13]. FY2025 broke the pattern. Gross spend grew 11%. Revenue grew 18% [14].

Sources: FY2025 10-K [15]; FY2024 10-K [16]; FY2023 10-K [17]; FY2022 10-K [18]; FY2021 10-K [19].

The seven-point wedge between those two bars is the take rate rising, and the company says as much. FY2025 revenue growth was driven by the 11% spend increase plus "a higher proportion of revenue earned from client spend due to increased utilization of our value-added services and data; and higher platform fees," with Kokai — the platform's major upgrade — and "increased pricing associated with value-added services and data" named explicitly [20]. The Q1 FY2026 10-Q repeats the construction for the March quarter's 12% revenue increase, again crediting "increased pricing associated with value-added services" [21].

Source: derived from reported revenue and gross spend, FY2021–FY2025 10-Ks [22][23][24].

My read is that gross spend, not revenue, is the cleaner measure of whether this franchise is winning, and on that measure FY2025 was a step down of a different order than the reported 18% suggests. Volume growth halved while price did the work.

Two things cut against reading that as structural decay. First, FY2022 saw a comparable take-rate step — from 19.4% to 20.4% — and gross spend growth then held at 24–25% for two more years, so a rising cut is not by itself a distress signal [25][26]. Second, the mix genuinely is shifting toward higher-priced data and measurement services, which is a real product story and not only a pricing lever. What would change my read in the other direction is a further year of single-digit gross spend growth: pricing can substitute for volume once, not repeatedly.

One disclosure detail is worth recording: the gross spend footnote's standing sentence telling readers to expect the take rate to fluctuate, carried from FY2021 through FY2024 [27][28][29][30], is gone from the FY2025 10-K, the year the take rate moved 130 basis points, replaced by a cross-reference [31], a retirement documented in full across three calls and the FY2025 10-K in Pricing and Take Rate; gross spend itself is still disclosed annually and still not quantified in the quarterlies [32].

The deceleration, quarter by quarter

Revenue growth has fallen in every quarter since the start of FY2025, and the guided June 2026 quarter continues the sequence.

Sources: quarterly revenue as reported in each quarter's Form 10-Q [33][34][35][36], with Q4 FY2025 derived as the full year less the first three quarters [37]; 2Q26 from the "at least $750 million" guide given on the Q1 2026 call [38].

Q1 FY2026 delivered $689 million of revenue at 12% growth and $206 million of adjusted EBITDA, a 30% margin; management guided the June quarter to "at least $750 million" of revenue [39] and roughly $260 million of adjusted EBITDA [40]. Against the $694 million booked a year earlier, that guide implies about 8% growth. Management's account of the slowdown is macro and vertical-specific: geopolitical uncertainty, CPG brands facing consumer softness and input-cost inflation, and tariffs restraining automotive [41]. Full-year 2026 adjusted EBITDA margin is still guided to at least 40%, roughly level with 2025 [42].

Asked directly whether competitive pressure had increased, the CEO's answer was no — "I think Google was a far better competitor than Amazon is today or, frankly, will likely ever be" — arguing that Amazon's economics tie it to selling its own inventory [43] and that the market has become more fragmented and noisier rather than more hostile [44]. The company's own risk factors are less sanguine, noting that walled-garden inventory providers "may exclusively sell their own inventory directly to advertisers, which prevents us from competing with them entirely for such inventory" [45]. Both statements can be true; they do not settle whether the 11% is macro or share.

What the accounts look like

Sources: FY2025 revenue and operating income, FY2025 10-K [46]; free cash flow derived as FY2025 operating cash flow of $993 million [47] less $197 million of purchases of property and equipment [48]; net cash is cash, equivalents and short-term investments at March 31, 2026 with no debt outstanding [49].

FY2025 turned $2.90 billion of revenue into $589 million of operating income — a 20% GAAP operating margin, up from 17% — and $443 million of net income [50], or $0.90 per diluted share [51]. Operating cash flow was $993 million, up 34% [52]. There is no debt: the credit facility was undrawn at year-end with $445 million available, so the company carries no near-term funding requirement [53].

Three features of the accounts shape how the numbers should be read.

The balance sheet is mostly other people's money in transit. At December 2025, receivables were $3.77 billion against payables of $3.01 billion [54] — both far larger than annual revenue, because the company invoices clients for the full cost of the media and remits it to suppliers. Total assets of $6.15 billion are dominated by that float, not by operating capital.

Adjusted EBITDA and GAAP earnings diverge by roughly the cost of the workforce's equity. FY2025 adjusted EBITDA of $1.20 billion reconciles to $443 million of net income mainly by adding back $491 million of stock-based compensation, plus $116 million of depreciation and $215 million of tax [55]. Stock compensation exceeded net income. The same wedge appears quarterly: Q1 FY2026 GAAP earnings were $0.08 per diluted share against $0.28 adjusted [56]. Any multiple quoted on this company needs its basis stated.

Buybacks have been large and, so far, badly timed. FY2025 saw 26.2 million Class A shares retired for $1.4 billion — an average of roughly $53 a share against a July 2026 price of $17.29, so that outlay would buy the same shares today for about $453 million [57]. Repurchases continued at $164 million in Q1 FY2026 [58]. The buying reduced the share count, and it was funded from cash flow rather than borrowing; it was also done at three times the current price.

Control sits with insiders. Class B shares carry ten votes each, and Class B holders — officers, employees, directors and affiliates — held approximately 49.9% of total voting power at December 2025, with automatic conversion to Class A not scheduled until December 2035 [59]. Co-founder Jeff Green remains CEO. What that concentration is worth to an outside holder depends on the record of the decisions it has produced, which this chapter does not settle.

The price and the arithmetic

The de-rating runs to 87.6% from the December 2024 peak, and it has been continuous rather than a single gap.

Source: daily closing prices as reported, month-end observations; all-time closing high of $139.51 on December 4, 2024.

The shares closed at $17.29 on July 24, 2026, against an all-time closing high of $139.51 on December 4, 2024 — a decline of 87.6%. FY2024 ended at $117.53 and FY2025 at $37.96, so the fall spans two calendar years of continuous compression rather than one event. Over the same period revenue rose from $2.44 billion to $2.90 billion and net income from $393 million to $443 million [60].

At $17.29 on 471.0 million shares outstanding at March 31, 2026, the equity is worth about $8.14 billion; net of $1.41 billion of cash and short-term investments and with no debt, enterprise value is roughly $6.74 billion [61].

Sources: derived from the July 24, 2026 close of $17.29 and 471.0 million shares outstanding [62]; FY2025 operating income and net income [63] and diluted earnings per share [64]; adjusted EBITDA and stock compensation [65]; operating cash flow [66] and capital expenditure [67]; FY2026 consensus per analyst estimates.

The spread down that column is where the disagreement sits. On adjusted EBITDA the business changes hands at 5.6 times enterprise value. On GAAP operating income, which charges the $491 million of equity compensation that keeps 3,843 employees in place, it is 11.4 times. Neither number describes a company priced for the 20%-plus growth of FY2024, and the second is not obviously cheap for a business guiding to 8% revenue growth in the current quarter. These three rows are orientation only; the full eight-row ladder, including the after-stock-compensation measures, sits in Cash and Share Count.

Consensus has moved with the price rather than ahead of it. Sell-side estimates put FY2026 revenue near $3.18 billion and FY2027 near $3.48 billion — growth of roughly 10% in each year, against 18% delivered in FY2025 — and FY2026 adjusted EPS around $1.85, with downward revisions outnumbering upward ones by roughly nineteen to two over the preceding month. The average published target is $24.32, and the analyst body covering the name is split, with more holds than buys and five outright sell ratings.

What this report tests

The facts above describe a business and a price that have moved in opposite directions. The Trade Desk is a debt-free, cash-generative, founder-controlled percentage-fee intermediary in programmatic advertising that has lost 87.6% of its market value while its revenue and profits rose, and in the year the shares fell hardest gross spend grew 11% rather than 25%, with pricing making up the difference.

This report tests: is the halving of gross-spend growth on The Trade Desk's platform — from 24–25% a year through FY2024 to 11% in FY2025 and single-digit guided revenue growth by mid-2026 — a repairable macro and execution stumble, or the onset of durable share loss to the walled gardens; and does the price at $17.29, roughly 11 times GAAP operating profit with $1.4 billion of net cash behind it, already pay for the pessimistic answer?

The evidence needed to settle it is specific and mostly knowable: whether gross spend reaccelerates or stays in single digits; whether take rate keeps rising and whether clients tolerate it; whether the CTV and retail-data mix shift is winning budget or merely repricing the budget already there; whether the leadership turnover of 2025–26 reflects an upgrade or instability; and what a business growing at 10% with a 40% adjusted-EBITDA margin and $491 million of annual equity compensation is actually worth. Those are the threads the chapters that follow pick up.

---

The three-year record and what is expected next

Over FY2023–FY2025 The Trade Desk turned 49% revenue growth into a near-tripling of operating profit, lifting operating margin from 10.3% to 20.3% [1]. In the first quarter of FY2026 that operating leverage stalled: revenue rose 12% and adjusted EBITDA fell 1% [2]. Consensus now models roughly 10% revenue growth in each of FY2026 and FY2027, and management's own full-year margin commitment requires the second half to do what the first half did not.

What the last three years show

The income statement over FY2023–FY2025 is a study in operating leverage. Revenue compounded at 22.0% a year while total operating expenses grew at 15.0%, so the expense base fell from 89.7% of revenue to 79.7% and operating income nearly tripled [3].

Sources: FY2025 Form 10-K, Consolidated Statements of Operations [4]; Adjusted EBITDA reconciliation for FY2024–FY2025 [5], FY2023 derived on the company's stated definition from reported net income, depreciation, stock compensation, interest income and tax [6].

Three features of that table matter more than the totals.

The leverage came from general and administrative, not from the platform. General and administrative expense was $520.3 million in FY2023 and $518.5 million in FY2025 — flat in dollars while revenue grew 49%, so it fell from 26.7% to 17.9% of revenue [7]. The company attributes the FY2025 decline chiefly to a $61 million reduction in the CEO Performance Option charge under graded-vesting attribution, partly offset by higher legal and professional fees [8]. That option was fully expensed by the first quarter of 2026, which is the subject of Cash and Share Count; the point here is that roughly nine percentage points of the ten-point margin gain came from a line whose decline was driven by that charge, which will not recur.

Platform operations moved the other way. The cost of actually running the bidding infrastructure rose from 18.8% of revenue in FY2023 to 21.4% in FY2025, driven by a $123 million increase in hosting costs for query volume, feature use by its technical teams, and new data centres [9]. In the first quarter of FY2026 it reached 26% of revenue [10].

The tax line is not stable. The effective rate was 33.2% in FY2023, 22.5% in FY2024 and 32.7% in FY2025 [11]. The swing is largely the tax effect of employee equity awards, and it now runs against the company: in the first quarter of FY2026 the rate was 49.3% because vesting awards produced tax detriments rather than the benefits recorded a year earlier [12]. The restricted shares that vested in the quarter carried a weighted-average grant-date value of $66.07, and the 15.5 million still unvested at March 31, 2026 carried $52.57, against a $17.29 share price [13]. The detriment is a function of the drawdown itself and persists while that gap does.

Cash followed earnings across the three years. Operating cash flow went $598 million, $739 million, $993 million; free cash flow on the company's own definition — operating cash flow less purchases of property and equipment and capitalised software — went $543 million, $632 million, $783 million [14]. Capital expenditure quadrupled over the same span, from $47 million to $197 million; what that buys, and how stock compensation and the buyback change the per-share arithmetic, is the subject of Cash and Share Count.

The balance sheet is the one part of the record that has not deteriorated. At March 31, 2026 the company held $878 million of cash and $528 million of short-term investments against no drawn debt, with $445 million available under an undrawn revolving facility and $2.0 billion of working capital [15] [16]. Total liabilities of $3.28 billion are almost entirely accounts payable to media suppliers ($2.63 billion), matched by $3.32 billion of receivables from advertisers [17]. There is no maturity wall, no covenant to breach and no refinancing to negotiate.

Where the operating leverage went

The clearest way to see the change is the incremental margin — what each additional dollar of revenue added to profit.

Source: derived from reported revenue, operating income and Adjusted EBITDA, FY2023–FY2025 Form 10-K [18] [19] and Q1 FY2026 Form 10-Q [20] [21].

In FY2024, 48 cents of every incremental revenue dollar reached adjusted EBITDA. In FY2025, 41 cents. In the first quarter of FY2026, revenue rose $72.8 million and adjusted EBITDA fell $1.8 million — an incremental margin of negative 2.5% [22]. The adjusted EBITDA margin fell from 33.7% to 29.9% year over year, and net income fell 21% on revenue that grew 12% [23] [24].

Management's account is that operating expenses excluding stock compensation rose 18% against 12% revenue growth, "particularly in areas like platform operations as we optimize our platform infrastructure and implement more AI-powered tools" [25]. The first quarter is also seasonally the weakest: adjusted EBITDA margin ran about 34% in the first quarter of FY2025 and about 47% in the fourth [26] [27]. Seasonality explains the level; it does not explain a 380 basis-point year-over-year decline in the same quarter.

Guidance against delivery

The company does not guide to a full year of revenue. It guides one quarter at a time, always as a floor — "at least" a stated figure — plus, since the fourth quarter of FY2025, an annual adjusted EBITDA margin commitment [28]. Nine consecutive quarters of that floor can be checked against what arrived.

Sources: quarterly revenue guidance from the Q4 FY2023 [29], Q1 FY2024 [30], Q2 FY2024 [31], Q3 FY2024 [32], Q4 FY2024 [33], Q1 FY2025 [34], Q2 FY2025 [35], Q3 FY2025 [36] and Q4 FY2025 [37] calls; reported quarterly revenue from the Form 10-Q series and, for fourth quarters, derived as the full year less the first three quarters [38].

Eight of the nine quarters cleared the floor, by a median of 1.7%. The exception was the fourth quarter of FY2024: guided to at least $756 million, delivered $741 million, and described on the call as the first shortfall "in our 8.5 years as a public company, excluding the first quarter of 2020" [39] [40]. What followed was a reset in how the floor is set: the very next guide, for the first quarter of FY2025, was cleared by 7.1%, and the adjusted EBITDA guide of approximately $145 million for that quarter produced $208 million — 43% above [41] [42].

Every adjusted EBITDA guide since has also been beaten: approximately $259 million guided and $271 million delivered in the second quarter of FY2025, $277 million and $317 million in the third, $375 million and $400 million in the fourth, $195 million and $206 million in the first quarter of FY2026 [43] [44] [45] [46] [47]. The practical reading of the second-quarter FY2026 guide of at least $750 million is therefore not 8.1% growth but something closer to 10%: applying the post-reset median beat of 1.7% gives roughly $763 million against $694 million a year earlier [48].

The 2026 margin commitment

On the first-quarter FY2026 call the company committed to a "full-year 2026 adjusted EBITDA margin percentage to be at least 40%, approximately in line with 2025" [49]. FY2025 came in at 41.3% [50] [51], so "at least 40%" already contains a step down of about 130 basis points, and the two phrases are only consistent at one decimal place of rounding.

The arithmetic of getting there is tight. The first half of FY2026 is now largely known: $689 million of reported revenue plus a $750 million floor, and $206 million of reported adjusted EBITDA plus a $260 million guide — a 32.4% first-half margin against 36.6% a year earlier [52] [53] [54] [55]. On consensus revenue of $3.18 billion for the year, the remaining half carries about $1.74 billion of revenue and, to reach a 40% full-year margin, needs $805 million of adjusted EBITDA — a 46.3% second-half margin.

Source: derived from reported and guided figures — Q1 FY2026 Form 10-Q [56], Q1–Q4 FY2025 and Q1 FY2026 earnings calls [57] [58] [59] [60] [61] [62]; FY2026 revenue on consensus estimates.

The second half of FY2025 delivered 45.2%. So the commitment asks the second half of FY2026 to expand its margin by roughly 110 basis points year over year, immediately after a first half that contracted by 420. Repeating last year's second-half margin exactly puts the full year at 39.4% — below the floor. Adding the guidance-beat pattern back in, with the second quarter clearing its EBITDA guide by the 5.7% it managed in the first, brings the year to 39.9%. On the evidence available, "at least 40%" is achievable and has essentially no cushion in it.

Two things could widen that cushion, and both are visible in the record. Headcount growth is planned to stay below revenue growth for the remainder of the year, a discipline the company also applied in 2025 [63]; and revenue above the consensus line lifts the margin denominator and numerator together, since the incremental cost of an additional impression is largely hosting.

What the estimates say

Consensus has settled on a business that grows at roughly the rate it is currently guiding, rather than one that reaccelerates or one that breaks.

Sources: FY2025 revenue per the Form 10-K [64]; forward figures are consensus estimates as of late July 2026, no filing source.

Two features of the consensus set deserve attention.

The reset is recent and it has held. Consensus adjusted EPS for FY2026 stood at $2.07 ninety days before the reference date and at $1.85 now, a cut of 10.6%; FY2027 fell from $2.40 to $2.15, a cut of 10.5%. Almost the entire move happened between the ninety-day and sixty-day marks — the window containing the May 7, 2026 first-quarter report — and the average has moved less than 0.2% in the month since.

Source: consensus adjusted earnings estimates and their 30-, 60- and 90-day prior vintages, as of late July 2026; no filing source.

The dispersion is one-sided at the far end. For FY2026 the revenue range is narrow — a $3.05 billion low against a $3.21 billion high across 35 estimates, because three of four quarters are already guided or reported. For FY2027 the range opens to $2.63 billion low against $3.80 billion high across 36 estimates. The low is not a slow-growth case: at $2.63 billion, FY2027 revenue would be 9.2% below the $2.90 billion the company reported in FY2025. At least one analyst covering the name models absolute revenue decline two years out. The adjusted EPS range for FY2027 is narrower in relative terms, $1.79 to $2.58.

The published sell-side stance is a split with a long tail of caution: 2 strong buy and 11 buy against 18 hold, 4 sell and 1 strong sell, with an average target of $24.32 — 40.7% above the $17.29 price. A target 41% above the market, published by the same body of analysts that cut its FY2026 earnings estimate by 10.6% in May, is better read as the output of a model than as a forecast of where the shares trade.

Forward multiples

At $17.29 on 471.0 million shares the equity is $8.14 billion; net of $1.41 billion of cash and short-term investments with no drawn debt, the enterprise is $6.74 billion [65]. Against management's own FY2026 commitment — 40% of consensus revenue, or $1.27 billion of adjusted EBITDA — that is 5.3 times. Against consensus adjusted earnings the shares trade at 9.3 times FY2026 and 8.1 times FY2027.

Sources: net cash and share count from the Q1 FY2026 Form 10-Q [66]; adjusted EBITDA on the company's FY2026 margin commitment [67] applied to consensus revenue; earnings multiples on consensus adjusted EPS; price of $17.29 at July 24, 2026.

Those multiples carry a definitional warning. The consensus $1.85 is an adjusted number that excludes stock compensation; the statutory result is materially lower. In the first quarter of FY2026 the company reported $0.28 of adjusted diluted earnings against $0.08 on a GAAP basis — a gap of $0.20 a share in a single quarter [68]. Carrying that per-share gap across four quarters puts the FY2026 consensus at roughly $1.05 of GAAP earnings, or about 16 times the price rather than 9. The gap is an illustration, not a forecast: it moves with the tax treatment of vesting awards, which is exactly the line that swung the first-quarter effective rate to 49.3% [69]. The full ladder from adjusted to statutory to post-compensation cash is worked through in Cash and Share Count.

The read

The three-year record and the forward set point in opposite directions, and the pivot between them is narrow enough to date. Through FY2025 the company converted growth into profit at 41 to 48 cents on the incremental dollar and lifted operating margin ten points; in the first quarter of FY2026 the incremental margin was negative, and roughly nine of those ten points of margin gain traced to a general and administrative line held flat by an executive option charge that has now been fully expensed. What consensus and the company now describe is not a business in decline but a business growing near 10% with a margin that has stopped rising — priced at 5.3 times its own committed FY2026 adjusted EBITDA, with $1.41 billion of net cash, no debt and no maturity to refinance.

The strongest fact against reading the first quarter as an inflection is the guidance record itself: eight of nine revenue floors cleared and five consecutive adjusted EBITDA guides beaten by between 4.6% and 43%, which argues the second-quarter $750 million floor and the $260 million EBITDA guide are set to be exceeded, and that the second-half margin the 40% commitment requires is a number management already has line of sight to. The strongest fact for reading it as an inflection is that the cost line moving against the margin is platform operations — the hosting and data-centre cost of running the bidding platform — which rose from 18.8% of revenue in FY2023 to 26% in the first quarter of FY2026 and is tied to query volume rather than to a discretionary budget.

Two observable items would settle it inside two reporting cycles. The second-quarter adjusted EBITDA margin: at or above 34.7% on revenue above $750 million, the first quarter reads as timing; materially below, the leverage reversal is structural. And whether the "at least 40%" language survives the second-quarter call intact — a downgrade to "approximately 40%" or its removal would say more than any single quarter's number, because it is the only annual commitment management has made.

---

Growth across the field in 2025

Calendar 2025 tests whether the slowdown in spend crossing The Trade Desk's platform was the market's or the company's. Amazon's advertising revenue grew 22% and Alphabet's total advertising 11%, while Viant — the closest listed independent demand-side platform — grew contribution ex-TAC 18% and guided 17% for the first quarter of 2026 against a market it sizes at 13%. The competitive language new to TTD's FY2025 filing points at its agency clients as much as at Amazon.

The Trade Desk reports gross spend — the whole amount a client puts through the platform, of which revenue is roughly a fifth — and says plainly what the metric is for: "For internal management purposes, we utilize gross spend as a metric to assess our market share and scale" [1]. That market-share metric grew 11% in 2025, to $13.39 billion from $12.04 billion, while revenue grew 18% [2]. The gap between those two numbers is pricing, and the Toll on Ad Spend chapter has already set it out. The 11% is the volume figure, and by the company's own description it is the measure of market share and scale.

Sources: TTD FY2025 Form 10-K, Management Discussion and Analysis, Executive Summary [3]; Amazon FY2025 Form 10-K, disaggregated net sales [4]; Alphabet FY2025 Form 10-K, revenues by type [5]; Viant Q4 2025 earnings call [6]; Criteo Q4 2025 earnings call [7]. Percentages computed from the reported currency amounts.

Amazon's advertising services line went from $56.21 billion in 2024 to $68.64 billion in 2025, a 22% increase after a 20% increase the year before — an acceleration [8]. Alphabet's total advertising grew 11.4%, to $294.69 billion from $264.59 billion [9].

Inside that Alphabet figure sits the strongest evidence for the market explanation. Google Network — the part of Alphabet that monetises third-party publishers, the same open web TTD's clients buy — fell $567 million, to $29.79 billion, and the volume decline was steeper than the revenue decline: impressions dropped 7% while cost-per-impression rose 7% [10] [11]. The largest incumbent in third-party publisher monetisation shrank in 2025, on falling volume and rising prices. Against that benchmark, TTD's 11% volume growth is a gain.

Viant is the more exacting comparison, because it runs the same model: a buy-side platform, no owned inventory, revenue taken as a cut of client spend under master service agreements that charge "a platform fee that is primarily a percentage of spend" [12]. Its 2025 revenue rose 19% to $344.2 million and its contribution ex-TAC — the measure closest in character to TTD's revenue — rose 18% to $208.7 million [13]. Viant absorbed a hard political-advertising comparison in that year and still grew at that rate; TTD does not disclose a political headwind of its own in the FY2025 filing.

Not every independent grew. Criteo's contribution ex-TAC rose 3.5% at constant currency on $1.9 billion of revenue [14], and at Nexxen self-service contribution ex-TAC declined 5% in the fourth quarter while private-marketplace and display contribution ex-TAC each fell 9% [15], with its full-year contribution ex-TAC retention rate dropping to 92% from 102% [16]. Both are sell-side or retail-media businesses rather than pure demand-side platforms, so their weakness says more about supply-path and retargeting economics than about the DSP category. Independence alone neither guaranteed growth nor caused decline in 2025: the one company running TTD's exact model grew at 18%, and TTD's volume grew at 11%.

The forward line

The deceleration did not stop at the year end. On the quarterly series charted in Toll on Ad Spend, growth runs from 25.4% in the first quarter of 2025 to 8.1% implied for the second quarter of 2026.

First-quarter 2026 revenue rose 12%, which the 10-Q attributes to higher gross spend plus "increased pricing associated with value-added services, higher utilization of our value-added services and higher platform fees" [17]. Second-quarter guidance of at least $750 million [18] implies about 8% against the $694.0 million reported for the same quarter of 2025 [19]. Viant, on the same call cycle, guided first-quarter 2026 revenue up 20% and contribution ex-TAC up 17% at the midpoint, and told investors it expected to keep "outpacing the broader U.S. programmatic market, which is projected to grow approximately 13%" [20].

An analyst put the arithmetic to management directly on the Q1 2026 call: industry expectations for digital and video growth were above 8%, so the guide was "pointing to below industry growth" [21]. The reply did not contest the framing. It confirmed that continued weakness in consumer packaged goods and automotive would at least create easier comparisons later in the year, and added that "there is not really anything incremental to add on the agency front" [22]. Management's macro account of the slowdown is on the record and unchanged; what the peer set adds is that a macro account has to explain why the two largest platforms and the nearest independent all grew faster through the same quarters.

The Amazon explanation, tested

TTD's filings register a competitive shift, and the sequence is precise. The FY2023 10-K described its rivals as "divisions of large, well-established companies such as Google and Adobe" [23]. The FY2024 10-K replaced Adobe with Amazon in the same sentence [24], and the FY2025 10-K keeps that pairing [25].

The FY2025 filing then adds a paragraph absent from the four 10-Ks before it: "Historically, some of our competitors have sought to differentiate themselves to prospective customers primarily on the basis of artificially low prices, which are enabled by inherent conflicts of interest and a lack of objectivity" [26]. The same clause was inserted into the growth-drivers paragraph of the management discussion, where FY2024 had simply said growth "has been driven by expanding our share of spend by our existing clients and adding new clients" [27] and FY2025 says growth "has been largely driven by" that, before adding the ability to differentiate against "competitors' platforms that may offer artificially low prices" [28]. A company whose revenue grew 18% on 11% more client spend has, in the same document, written price competition into both its risk factors and its growth discussion. Neither passage names the competitor.

The evidence that the price pressure is converting into lost volume is weaker than the pressure itself. Amazon's $68.64 billion advertising line is described in its own filing as revenue "to sellers, vendors, publishers, authors, and others, through programs such as sponsored ads, display, and video advertising" [29] — a bundle dominated by sponsored placements on Amazon's own store, not by third-party open-internet buying. Asked in November 2025 whether Amazon's reported zero-percent DSP fees had raised competitive intensity, Viant's chief executive said no, that "most of their revenue is sponsored listings," that the DSP "is a very small portion," and that "we do not see them … in the competitive bake-off processes at the finish line" [30]. That is an interested witness, and a smaller one. But it is a witness competing for the same budgets, growing at 18%, and reporting no Amazon effect — which makes Amazon a poor sole explanation for a deceleration that shows up at TTD and not at Viant.

Where the concentration sits

Two agency holding companies accounted for 30% of gross billings in 2025, against one at 14% in 2024. No other client-facing disclosure in the FY2025 filing moved as far.

Sources: TTD FY2025 Form 10-K, Concentration of Risk [31]; TTD FY2023 Form 10-K, Concentration of Risk [32]. Two holding companies cleared the 10% threshold in 2021 and 2025; one did in 2022, 2023 and 2024, so the 2021 and 2025 bars aggregate two disclosed relationships and the middle bars one.

Aggregated to the holding-company level, one agency group accounted for 11% of gross billings in 2022, 12% in 2023 and 14% in 2024. In 2025 two groups together accounted for 30% [33] [34]. Receivables tell the same story from the other side: at December 2025 two clients accounted for 30% of consolidated accounts receivable [35]. The company does not say why the 2025 figure roughly doubled, and it does not name either group; the FY2021 through FY2023 filings did name the one above the threshold as Publicis Groupe [36], and the name disappears from the FY2024 and FY2025 10-Ks.

What sits behind that 30% is thin contractual paper. Clients hold master service agreements that "do not contain any material commitments on behalf of clients to use our platform," run one-year auto-renewing terms, and are "terminable at any time upon 60 days' notice by either party" [37]. Nearly a third of the volume the platform prices moves on 60 days' notice from two counterparties.

One of those relationships was in open renegotiation through the first half of 2026. Asked on the May 2026 call about "the Publicis discussions," the chief executive said the reported conflict had been "overdramatized," that TTD had done "billions of dollars of business with Publicis" since 2018, and that "our negotiations are ongoing," declining to say more [38]. Trade press reported in June 2026 that the two had settled privately and that the agency resumed recommending the platform after pausing client spend in March; no terms were disclosed, and no filing has since described the renewed agreement [39]. Three months before that call, defending OpenPath — the supply-path product for which TTD charges publishers 4.5% — the same executive said that "at a moment where many agencies are focused on principal-based buying, I think they're not doing as good of a job of representing their clients as they could" [40]. Principal-based buying is the practice of an agency buying media on its own book and reselling it to the client at a margin — a model that competes with a transparent percentage fee for the same dollar. TTD is arguing publicly against the economics of the channel its two largest billing relationships sit in.

That is a different competitive problem from the one usually named. Amazon can under-price a platform fee; an agency holding company that buys as principal removes the need for the platform fee to be quoted to the advertiser at all. The evidence for the second mechanism is on TTD's own pages — the concentration jump, the 60-day terms, the renegotiation, the public dispute over principal-based buying — while the evidence for the first is a risk-factor paragraph and a set of trade-press claims the corpus does not otherwise document.

What would change the read

The read here is that the 11% is more company-specific than the macro account allows, and that the sharpest identified pressure point is the agency channel rather than Amazon's platform. Three facts cut against that read.

The strongest is Google Network: the incumbent in third-party publisher monetisation shrank 1.9% on a 7% impression decline in the same year [41] [42]. If the open web is the relevant market, TTD grew and the incumbent contracted, and 11% is a share gain in a shrinking pool. Second, scale: Viant's $208.7 million of contribution ex-TAC is about a fourteenth of TTD's revenue, and a platform at that size can add a single flagship advertiser and move its growth rate by points [43]. Third, retention: TTD's customer retention rate exceeded 95% in each of the eleven years through 2024 [44], which is not what a departing client base looks like.

Three developments would settle it. A gross-spend growth rate for 2026 that recovers toward the low-to-mid teens while Viant's decelerates would vindicate the cyclical account. A concluded Publicis agreement disclosed on terms that hold the platform fee, or a 2026 concentration figure that falls back toward the mid-teens, would remove the channel risk that the FY2025 filing introduced. And a second consecutive year in which the nearest same-model competitor grows at roughly 18% while gross spend grows near 11% would make the share-loss reading hard to argue against, whatever the macro backdrop does.

The corpus cannot close two gaps. It contains no third-party measurement of open-internet or connected-TV spend growth for 2025 and 2026, so the 13% market figure quoted here is a competitor's estimate, not an independent one. And it carries no agency-side or advertiser-side account of the Publicis negotiation or of principal-based buying volumes — only TTD's and Viant's characterisations, plus a trade-press note of a settlement whose terms are undisclosed.

---

Pricing and Take Rate

Between $159 million and $177 million of The Trade Desk's $451.5 million FY2025 revenue increase — 35% to 39% of it — came from keeping a larger share of each dollar crossing the platform rather than from more dollars crossing it. The range is the price/volume interaction term, which the upper figure assigns to the rate leg (the rate change applied to FY2025 gross spend) and the lower to the volume leg (the rate change applied to FY2024 gross spend); the rest of this chapter uses the upper convention. The filings show why: the fee base was redescribed in FY2024, third-party data is moving from a client-billed pass-through to a company-priced product, and a 4.5% fee now sits on the publisher side. All three are real, all three are finite.

Source: derived from reported revenue and gross spend, FY2024 and FY2025 Forms 10-K [1]; [10].

What the fee is charged on

Through FY2023 the company described its charge in one sentence: "We charge our clients a platform fee, which is generally a percentage of the clients' purchases through the platform. In addition, we invoice our clients for the cost of advertising inventory purchased, plus data and any add-on features purchased through the platform." [2] The fee was a percentage of media. Data and add-ons were invoiced alongside it and largely passed through.

The FY2024 filing replaced that sentence. "We charge our clients for total spend on our platform, which includes spend and fees on advertising inventory, value-added services and data to support those purchases, in addition to the platform fee that is generally based on a percentage of our clients' total spend on the platform." [3] Two things changed inside one paragraph. Data and add-ons became "value-added services and data" — a category the company charges fees on, not merely invoices. And the percentage fee is now described as applying to total spend, which by the same sentence includes those service and data fees. The percentage sits on top of a wider base.

The gross-spend footnote was rewritten in the same filing, from "the amount of a client's purchases through our platform plus the platform fee" [4] to "the amount of a client's spend on our platform for advertising inventory, value-added services and data; plus the platform fee, which is generally based on a percentage of a client's total spend on our platform." [1]

These are descriptions of contracts, not an announced price change, and the take rate did not move in FY2024 — it moved the year after. What the language records is the architecture that the FY2025 step then used.

Where FY2025's revenue increase came from

Splitting each year's revenue increase into the part explained by more gross spend at the prior year's rate, and the part explained by the change in rate, separates volume from price.

Source: derived from reported revenue and gross spend, FY2021–FY2025 Forms 10-K [1]; [10]; [4]; [9].

Across FY2023 and FY2024 the rate contributed a net negative $6 million to two years of revenue increases totalling $867 million. In FY2025 it contributed $177 million. Held at the FY2024 rate of 20.30%, FY2025 revenue would have been about $2,720 million and growth 11.2% rather than the reported 18.5% — the same figure as gross-spend growth, which is what a fixed percentage fee produces.

The company's own attribution matches. For FY2021 through FY2024 the filings explained revenue growth by volume alone: "more advertisers and more campaigns executed by existing clients" (FY2022 [8] and FY2023 [7]), "more campaigns executed by existing clients, new clients and higher spend per campaign" (FY2024 [6]). No pricing clause appears in any of them.

The FY2025 paragraph adds one: revenue rose partly on "a higher proportion of revenue earned from client spend due to increased utilization of our value-added services and data; and higher platform fees," with "increased pricing associated with value-added services and data" named as a driver [5]. The March 2026 quarter repeats the construction and moves pricing to the front of the list [11].

Data moves from pass-through to product

The mechanism behind "increased pricing associated with value-added services and data" is visible in the product record. In the fourth quarter of 2025 the company overhauled its third-party data marketplace and launched Audience Unlimited, which "enables our users to use third-party data for a single fee" [13]. On the February 2026 call the chief executive described it as "a flat cost structure … for an all-in cost," adding that it is "completely optional" and that clients "can use it or continue to buy third-party data a la carte" [12].

The accounting consequence is set out in the critical accounting policies. Where the company buys data itself and supplies it "generally at no additional charge to our clients outside of our standard fees," the cost is "recorded in platform operations expense" rather than netted against revenue [14]. A la carte, the client buys a data segment, the company bills it and pays the supplier, and only the fee is revenue. Bundled at a price the company sets, the whole charge is revenue and the supplier cost is an operating expense. Same economics to the client; a materially higher reported take rate.

The FY2025 10-K flags the choice explicitly, in both directions. Revenue may fluctuate on "the amount of certain costs of supplier-provided components of value-added services and data recorded as reductions to revenue versus as expenses in platform operations" [15]; platform operations expense may vary for the mirror-image reason [16]. The same filing adds that the company will "continue to monitor changes in our platform and related offerings to assess whether the related third-party costs … should be recognized as reductions to revenue or expenses included in platform operations" [14]. None of that language appears in the FY2023 or FY2024 filings.

Whether the lever has actually been pulled shows in the platform-operations bridge.

Sources: FY2025 Form 10-K, results of operations [5]; Q1 FY2026 Form 10-Q, results of operations [11]. Data-related costs were not named in the FY2025 bridge.

For all of FY2025, the $147 million increase in platform operations was attributed to $123 million of hosting and $20 million of personnel [5]. Data-related costs do not appear. That is the strongest fact against reading the FY2025 take-rate step as an accounting reclassification: on the company's own bridge, it was not one.

In the March 2026 quarter data-related costs appear for the first time as a named driver — $11 million of a $39 million increase [11]. Against a revenue increase of $73 million in the same quarter, an $11 million absorbed data cost is 15% of the increment. That is the bundled model beginning to show up in the cost line, and it explains part of why platform operations reached 26% of revenue in the quarter against 23% a year earlier.

A fee on the sell side

OpenPath connects the platform directly to publishers. Asked in February 2026 about trade-press claims of a conflict of interest, the chief executive gave the price: "We plug in as directly as we possibly can to the seller or the publisher. We charge them 4.5%, which is meant to be nearly breakeven to slightly profitable" [17].

That fee is collected from publishers, so it does not enter gross spend, which measures what clients spend [1]. It reduces what the company remits to suppliers, and under net revenue recognition a smaller supplier payment on the same client billing is more revenue. It therefore lifts the measured take rate without any buyer seeing a higher fee.

Scale is not disclosed. Management said in August 2025 that "a material amount of spend on our platform is now flowing through OpenPath" [18] and in November 2025 that "OpenPath has grown by many multiples this year" [13]. Neither is a number. The bound is arithmetic: producing the entire $177 million rate contribution at 4.5% would require about $3.9 billion of spend routed through OpenPath, roughly 29% of FY2025 gross spend. The FY2025 filing attributes the step to services pricing and platform fees and does not mention OpenPath [5], so the plausible contribution is a fraction of that — but it is not zero, and it is not visible.

The tension worth naming is positional rather than accounting. The company's stated differentiator is that it "delivers valuable insights and results to clients without the conflict of interest and lack of objectivity that come with also selling owned advertising inventory" [1], and management states the principle plainly: "we will always only represent the buy side of digital advertising" [13]. Charging the sell side 4.5% for a supply integration is compatible with that on its own terms — the fee buys a connection, not representation, and publishers on the record report large gains from it. It is also the reason the objection exists, and the company's answer is a margin claim ("nearly breakeven") that no disclosure lets an outsider check.

What stopped being said

For three consecutive fourth-quarter calls the take rate came with an explicit reassurance. In February 2023: "for the ninth year in a row our take rate remained within a very consistent historical range" [19]. In February 2024: "our take rate in 2023 once again remained within a very consistent historical range. We continue to execute on the model set out at the company's inception of keeping take rate consistent while substantially increasing the value that our platform provides" [20]. In February 2025: "As expected, our take rate in 2024 once again remained within a very consistent historical range" [21].

On the February 2026 call, reporting the year in which the rate rose 132 basis points, the sentence is absent. What appears instead is a rebuttal of the opposite concern: "there's been a narrative that our margin or take rate must compress because other platforms offer lower upfront prices for non decisioned, non-data-driven buying. In reality, those business models deliver less value overall" [22]. The words "take rate" appear in the FY2021 through FY2024 Forms 10-K [9] [10] and nowhere in the FY2025 10-K or the Q1 FY2026 10-Q. A definition the company supplied for four years, and a commitment it restated on three calls, both lapsed in the year the number moved.

How much room the lever has left

Gross spend is disclosed annually, so FY2026's split between volume and rate cannot be measured until the FY2026 10-K. What can be framed is the trade-off. The table gives FY2026 revenue growth for combinations of gross-spend growth and take rate.

Source: derived from FY2025 revenue and gross spend as reported [1]. Cells show implied FY2026 revenue growth.

Consensus FY2026 revenue of $3,177 million needs 9.7% gross-spend growth at a flat rate, 7.2% with another 50 basis points of rate, or 4.9% with another 100. The consensus path (see Financials and Estimates) does not require volume to reaccelerate if the rate keeps climbing at the FY2025 pace. That is the load-bearing point for a reader watching gross spend as the health metric: for a second year, reported revenue can run ahead of the money actually crossing the platform.

The read here is that the FY2025 step was earned rather than engineered — it came from repricing and bundling a real product set, not from reclassification, and the company's own expense bridge supports that. But it is a lever with a visible end. Kokai, the platform release that carried the repricing, was already running about three-quarters of client spend by August 2025 [18], so adoption headroom is thin. Bundling raises the cost base as it raises revenue, which the March 2026 quarter already shows. Volume discounts remain a named driver of the rate [10] at a moment when two agency holding companies route 30% of gross billings on contracts terminable at 60 days (see Share or Cycle). And the counter-case is on the record: clients who moved most of their spend onto Kokai grew that spend more than 20% faster than those who had not [18], which is what a price increase paying for itself looks like.

What would change the read: gross spend growth in the FY2026 10-K landing close to revenue growth would say the lever is spent and the toll is back to tracking volume; a second year of a 100-plus basis point gap would say pricing is now carrying the growth rate, and would make client tolerance — not market share — the variable to watch.

Limitations. Gross spend is a management-defined metric disclosed once a year, so the volume/rate split cannot be computed for any quarter and the FY2026 split is unobservable until early 2027. The company publishes no revenue disaggregation, so platform fees, value-added services, data and OpenPath cannot be sized separately; the OpenPath figure above is an upper bound from stated pricing, not a disclosure. FY2021–FY2023 revenue and gross spend are reported in whole millions, which makes those years' take rates precise to roughly a basis point. The bundling mechanism described is inferred from the revenue-recognition policy and the product description; the company does not state which data agreements sit on which side of the line.

---

Where the growth is coming from

The deceleration is not spread evenly across the platform. In the March 2026 quarter, video — which includes connected TV — grew about 21% year over year on the company's own channel disclosures, while everything else grew about 3% and mobile, roughly 29% of the platform, shrank. Video is still compounding. The mix shift it produces, though, adds well under a point a year to total growth.

Nineteen quarters of channel mix

The Trade Desk publishes no channel split in its 10-K. It describes one in words on almost every earnings call — "a mid-40s percentage share", "a high-28s percent share" — and has done so consistently enough that the phrasing can be assembled into a series no filing contains.

In the June 2021 quarter, mobile was the largest channel at a low-40s percentage share and video, including CTV, was a high-30s share [1]. The two were level at about 40% each by the September 2021 quarter [2] and again in the March 2022 quarter [3]. By the March 2026 quarter, video was a low-50s percent of the business and mobile a high-28s percent [4].

Midpoints of the verbal bands given in the quarterly prepared remarks; the December 2021 quarter gave no numeric share for either channel and is omitted. Sources: Q2 FY2021 [5]; Q3 FY2021 [6]; Q1 FY2022 [7]; Q2 FY2022 [8]; Q3 FY2022 [9]; Q4 FY2022 [10]; Q1 FY2023 [11]; Q2 FY2023 [12]; Q3 FY2023 [13]; Q4 FY2023 [14]; Q1 FY2024 [15]; Q2 FY2024 [16]; Q3 FY2024 [17]; Q4 FY2024 [18]; Q1 FY2025 [19]; Q2 FY2025 [20]; Q3 FY2025 [21]; Q4 FY2025 [22]; Q1 FY2026 [23].

At the midpoints, the four named buckets — video, mobile, display and audio — account for about 99% of the platform in the March 2026 quarter, so they are being reported as mutually exclusive, leaving little room for digital-out-of-home and native [24]. The boundary between two of them is not defined anywhere in the filings: the same remarks that separate mobile from video have also described growth inside mobile as "solid across in-app and mobile video" [25].

What the shares imply about growth

Revenue in the March 2026 quarter was $688.9 million against $616.0 million, up 12% [26]. Applying the band midpoints to that base — video at 48% a year ago and 52% now, mobile at 35% and 28.7%, audio at 5% and 6% — gives the implied growth of each channel.

Derived: quarterly revenue from the Q1 FY2026 Form 10-Q [27], channel shares from the Q1 FY2025 [28] and Q1 FY2026 [29] prepared remarks, midpoints applied.

The decomposition reconciles: 48% of the prior-year base growing 21.1% plus 52% growing 3.2% produces 11.8%, the reported figure [30]. Taking the widest reading of both bands, video grew between 16% and 26%, mobile fell between 5% and 12%, and the non-video half of the platform lies between a 1% decline and 7% growth. The audio figure has an independent check: management said audio "grew year over year at a rate higher than any other channel in Q1", which is what a move from around 5% to around 6% of a business growing 12% implies [31].

The same exercise on the volume metric over the full year is directionally identical but less precise. Gross spend rose 11% to $13.39 billion in FY2025 from $12.04 billion [32]. Averaging the four quarterly midpoints puts video at 47.25% of spend in 2024 and 49.0% in 2025, which implies video gross spend of roughly $5.69 billion rising to $6.56 billion, up 15%, against 8% for everything else. Because the annual share moved less than two points, the band corners are wide enough that the two rates could be as close as 11% and 12%; the quarterly comparison, where the share moved four points and one bucket was quoted to a tenth, is the tighter test.

How the channel language changed

For three years the description of CTV was a superlative, repeated almost verbatim. It was "our fastest growing channel and it has rapidly become our largest" for the September 2022 quarter [33]; CTV "by a wide margin, led our growth again during the quarter" for the June 2024 quarter [34]; "our largest and fastest growing advertising channel" for the March 2025 quarter [35].

The wording then loosened, though not in a straight line. In the September 2025 quarter the two framings sat side by side: the CEO still called CTV the company's "largest and fastest-growing channel" [36], while the CFO's version of the same claim was relative rather than absolute — CTV "has been consistently growing at a faster rate than the overall business" [37]. For the December 2025 quarter the superlative itself was downgraded to "one of our fastest-growing channels", with audio named as the channel that grew faster than any other in the quarter [38] [39]. In the March 2026 quarter no superlative was attached to CTV at all: growth was "driven by strong trends across CTV and audio", and audio again grew faster than any other channel [40].

That sequence tracks the arithmetic rather than contradicting it. Video decelerating from a rate in the thirties to roughly 20% while the total decelerates to 12% is exactly the pattern that turns a superlative into a comparative. What the company has never disclosed, in any filing or on any call in the archive, is a consolidated CTV growth rate or a CTV dollar figure; the closest it has come is regional colour, as when CTV across EMEA and North Asia was described as "growing over 100% year-over-year in each region" in the September 2023 quarter [41]. The last platform-wide CTV operating metric was in August 2021: "nearly 10,000 CTV advertisers on our platform, up over 50% compared to last year" [42].

Sizing the tailwind

The FY2025 10-K opens its industry section with digital advertising at over $700 billion of annual spend and more than 70% of the total advertising market, a global advertising TAM reported past $1 trillion for the first time in 2024, and "Rapid Growth of CTV" as the first of six trends listed [43]. The promotion is new: the FY2024 filing led with media becoming increasingly digital and placed the emergence of CTV third [44].

Neither the company nor its filings size CTV itself. A competitor's does. Viant Technology, an independent demand-side platform that runs the same buy-side model and owns no inventory, quotes third-party forecasts in its own FY2025 10-K: US CTV advertising reaching $38.0 billion in 2026, continuing at double-digit annual growth rates, surpassing traditional TV advertising in 2029 at $52.5 billion; 89% of US CTV ad spend already transacted programmatically in 2025, rising to 93% by 2027; US programmatic advertising growing at a 12% compound rate from 2024 to 2027 to $225.3 billion, taking programmatic from 40% to 46% of total US media spend [45].

Third-party forecasts quoted in Viant Technology's FY2025 Form 10-K [46].

Two things follow. The first is that headroom is not the binding constraint: The Trade Desk bought roughly $6.6 billion of video inventory globally in FY2025 on the midpoint shares, against a US CTV market forecast at $38.0 billion for 2026 alone. The second is that video growing 15% to 21% against a market forecast to grow at a low-double-digit rate means the company is, at worst, holding its position in the channel the bull case is built on — the deceleration is happening somewhere else on the platform.

Three facts cut against reading that as a durable tailwind. With 89% of CTV spend already transacted programmatically, the conversion leg of the CTV story is close to finished inside the channel; what remains has to come from linear budgets migrating and from price, which is slower and more mechanical than a penetration curve [47]. YouTube, the largest single streaming advertising property, grew ad revenue 11.7% in 2025, to $40.4 billion from $36.1 billion — a global figure against a US market forecast, but a reminder that the shift to streaming was not delivering 20% growth to the biggest incumbent either [48]. And the supply side is not passive: Viant markets a programme built on "the removal of resellers from the digital supply chain" through direct partnerships with CTV publishers [49], while The Trade Desk's own risk factors note that "a few inventory suppliers hold a significant portion of the programmatic inventory" [50].

What the mix can and cannot do

Total growth is the weighted sum of two very different businesses. At the March 2026 base — video 52% of the platform, everything else 48% — the arithmetic is fully determined by the two growth rates.

Derived: total growth equals 0.52 times video growth plus 0.48 times growth in all other channels, using the March 2026 channel mix [51].

Consensus of roughly 9.7% revenue growth in FY2026 and 9.6% in FY2027 (Financials and Estimates) sits close to the cell where video compounds near 20% and the rest of the platform is flat to slightly down — that is, the street is underwriting the pattern the March quarter already showed rather than a repair of it.

The mix shift does work in the company's favour, but slowly. Holding video at 20% growth and the rest of the platform exactly flat, total growth rises from 10.4% in the first year to 11.3%, then 12.2%, then 13.0% in the fourth, as video's share climbs from 52% to about 69%. That is roughly 80 to 90 basis points of total growth a year purchased by arithmetic alone. A return to the high-teens rates of 2023 and 2024 needs something more than mix: either CTV re-accelerating well above 20%, or the 48% of the platform that is currently flat to shrinking finding a floor.

Limits of the disclosure

The shares are verbal bands offered on calls, not audited figures, and the midpoint convention can be off by a point in either direction per bucket — which is why the growth estimates above are given as ranges rather than points. The video bucket blends CTV with non-CTV online video, so a CTV-only growth rate is not observable from anything the company publishes; the risk factors state only that "the demand for CTV inventory on our platform has been a significant driver of growth" [52]. The shares are described as shares of spend, while the quarterly base they are applied to is revenue; differing take rates by channel would move the implied levels, though not the direction of the gap. The borrowed forecasts have a problem of their own: the segment breakdown further down the same page of that competitor filing puts US CTV advertising at $38.0 billion in 2026 growing to $42 billion by 2028, a 12% compound rate, which does not reconcile with the $52.5 billion for 2029 stated a few paragraphs earlier [53]. Either path still leaves the channel several times larger than the video volume the company currently buys. And the run's web-research access was unavailable, so no independent estimate of The Trade Desk's CTV share could be checked against any of these figures.

One initiative aimed squarely at CTV supply is worth marking for its disclosure status rather than its economics. The Ventura operating system for connected television was announced on the February 2025 call [54]; fifteen months later the CEO described "great partnership discussions with our operating system for CTV called Ventura" and said more would be heard "in the years to come" [55]. It is not named anywhere in the FY2025 10-K, which does name OpenPath, OpenAds, OpenSincera and PubDesk among the products built to improve supply-chain quality [56]. On the record available, Ventura carries no revenue, no partner count and no timetable.

The read

The evidence points to a company whose CTV franchise is intact and whose problem sits in the other half of the platform. Video grew about twice as fast as everything else across FY2025 on gross spend and roughly six times as fast in the March 2026 quarter, it is growing at or above the rate forecast for the programmatic market, and its share of spend has climbed from a high-30s to a low-50s percentage across the nineteen disclosed quarters without one reported decline. Against that, mobile — the second-largest channel — is now shrinking in absolute terms, and mix arithmetic alone buys back under a point of total growth a year.

The strongest fact against this reading is that it rests on management's own verbal bands, unaudited and unaccompanied by any CTV-specific figure, in a period when the company retired the superlative it had used for CTV on more than a dozen calls. Two observations would change it. If the video share stops rising while total growth stays in single digits, video growth has converged with the rest of the platform. If mobile's decline stops and display holds, the same 20% video growth that produces 10% total growth today would produce 12% to 13% — and the deceleration would look far more like the cyclical account management gives than a structural one.

---

Where the slowdown actually sits

The deceleration is a United States event. On the company's own recast geographic disclosure, US revenue grew 16.1% in FY2025 while international grew 34.8%; in the March 2026 quarter the US grew about 5% and international about 55% [1] [2]. International is 14.5% of revenue, too small to carry the company, and a billing-address reallocation inside two agency holding companies could account for much of the shift.

Source: revenue by principal geographic area, FY2025 Form 10-K, Note 12 [3].

The disclosure changed in the year the numbers did

Through the FY2024 filing, the geographic note reported Gross Billings by geography — a volume measure, disclosed back to 2019 [4] [5]. The FY2025 filing replaced it with Revenue by geography and recast 2024 and 2023 onto the new basis, with no reason given beyond the recast note itself [6]. No gross-billings-by-geography table appears anywhere in the FY2025 10-K, so the geographic volume series stops at FY2024.

That timing echoes the disclosure change documented in Pricing and Take Rate: the same filing that dropped the phrase "take rate" also dropped the geographic volume split. Both removals have the same consequence — within a geography, the split between more spend and a higher fee on the same spend is no longer computable.

The change is not, however, a cosmetic one that distorts the level. The two bases overlap in 2023 and 2024, and they nearly coincide: international was 12.88% and 12.84% of gross billings in those years, against 12.81% and 12.73% of revenue. The billings-to-revenue conversion is close enough between the two geographies that the series can be read across the seam.

Sources: gross-billings basis from the FY2021 and FY2024 Form 10-Ks, Note 12 [7] [8]; revenue basis recast for 2024 and 2023, FY2025 Form 10-K, Note 12 [9]. Shares computed from the reported dollar tables.

International was 14.7% of gross billings in 2019 and 12.8% in 2024 — seven years of disclosure, and it went backwards over five of them, bottoming at 12.3% in 2022 when international billings grew 11.2% against 26.7% in the US. The FY2025 step to 14.5% is the first genuine break in the series, and it happens in the year total growth halved.

Nine quarters, unchanged mix

Management has described international as outgrowing North America since the March 2023 quarter, counting the streak out loud on each call: "slightly outpaced" in Q1 2023 [10], through to "the ninth quarter in a row" in Q1 2025 [11]. Across those nine quarters the disclosed share barely moved: North America was about 88% of spend in Q1 2023 and about 88% in Q1 2025, having troughed at about 90% through 2022 [12]. An outperformance that leaves the mix unchanged for two years is, arithmetically, a rounding-scale outperformance.

The break comes later, and the calls show it in the same sentence where the vocabulary changes. Through Q3 2025 the disclosure was "North America represented 87% of our business" [13]. From Q4 2025 it became "the United States represented approximately 84% of our revenue" [14], and in Q1 2026 "approximately 82% of our revenue" [15]. Two things changed at once — North America to United States, and spend to revenue — so the apparent 87-to-82 slide across three quarters is not a like-for-like series. Canada sits inside "North America" on the old basis and inside "International" on the new one.

The Q1 FY2026 10-Q gives the geographic split as a percentage of revenue and recasts the year-ago quarter onto the identical basis: United States 82% in Q1 2026 against 87% in Q1 2025 [16]. Five points of mix in one year, measured consistently.

The March 2026 quarter, decomposed

Applying those recast percentages to reported revenue of $688.9 million against $616.0 million [17] gives US revenue of about $565 million against $536 million, and international of about $124 million against $80 million. That is US growth of 5.4% and international growth of 54.8%, which reconciles to the reported 11.8%.

The percentages are rounded to whole numbers, so the decomposition carries a band. At the corners of the rounding, US growth runs between 4.2% and 6.7%, and international between 45.0% and 65.5%. The conclusion survives the band on either side: the US business grew in the mid-single digits, and international grew at least 45%.

Sources: FY2024 and FY2025 revenue by principal geographic area, FY2025 Form 10-K, Note 12 [18]; 1Q26 derived from the recast percentage split [19] applied to reported quarterly revenue [20].

In FY2024 the two geographies grew within 80 basis points of each other — 25.7% and 24.9%. Whatever changed in FY2025 did not change them together. International contributed 24.0% of the FY2025 revenue increase from 12.7% of the prior-year base; a year earlier it had contributed 12.5% from a nearly identical weight.

What the consensus number implies for the United States

FY2026 consensus revenue of $3,177 million is 9.7% growth. With FY2025 weights of 85.5% US and 14.5% international, that headline resolves into a pair of geographic rates, and the pairs are not flattering to the domestic business.

Source: derived from FY2025 geographic revenue weights [21] and FY2026 consensus revenue as compiled in the estimates feed.

If international merely holds the FY2025 rate of 35%, consensus implies the US grows 5.4% — the rate it actually grew in the March quarter. If international holds the March-quarter rate near 55%, consensus implies a US business growing 2%. The forward number the market is underwriting is consistent with a domestic business that has stopped compounding and an international business doing the work.

The same arithmetic runs the other way and is where the constructive case lives. Holding the US flat and international at 45%, total growth is 6.5%; at a 5% US and 45% international, it is 10.8%, and because the fast segment gains weight each year, total growth then rises to 12.6%, 14.8% and 17.3% over the following three years as international's share climbs from 14.5% to 38%. That is the mix-drift mechanism described for channels in CTV and Channel Mix, operating on a wider spread — but it needs international to roughly quadruple to $1.9 billion, and nothing in the record establishes that a 45% rate survives that scale.

The capital is going the other way

The geographic note also splits property and equipment and operating lease assets. US balances rose from $366.2 million to $652.3 million during FY2025, up 78%, while international fell from $106.9 million to $86.6 million, down 19% [22]. The entire $265.8 million increase in the combined balance, and then some, was incurred in the United States.

That is the infrastructure build described in Cash and Share Count, located. It is defensible — data centres serve global query volume regardless of where the client is billed, and the filing states the platform "is used by clients globally in a similar manner across geographies" [23]. Headcount tells a milder version of the same story: 63% of the 3,843 employees sat in North America at the end of FY2025, with EMEA at 20% and APAC at 17% — roughly 37% of people supporting 14.5% of revenue [24]. Whatever operating leverage international eventually delivers, it is not delivering it yet.

What could undo this read

A billing-address reallocation is the strongest counter-fact, and it is arithmetically sufficient. Geographic attribution is "based on the address of the clients or client affiliates" [25] — not where the advertising runs. The FY2025 10-K discloses that two holding companies accounted for 30% of Gross Billings in 2025, against one at 14% in 2024, and states that the company contracts not with holding companies but "with various of their individual agencies" as separate clients [26]. Which agency entity signs therefore sets the geography. Five points of Q1 2026 revenue is about $34 million a quarter, roughly $138 million annualised — 4.8% of FY2025 revenue, or about $640 million of gross spend at the FY2025 take rate. That is comfortably inside one holding company's book. Nothing in the corpus rules this out, and the concentration step and the geographic step happened in the same year.

Volume and rate cannot be separated within a geography. The FY2025 take rate rose 132 basis points company-wide. Because the gross-billings-by-geography table was retired in the same filing, there is no way to tell whether international's 34.8% revenue growth is 34.8% more spend or a smaller volume gain carrying a larger fee increase — for instance from the value-added-services pricing that Pricing and Take Rate traces.

Currency is not the explanation. The functional currency of every subsidiary is the US dollar, so there is no translation adjustment inflating international revenue, and the majority of platform transactions are denominated in dollars [27] [28]. The FY2025 foreign currency exchange line was a $0.7 million gain against $2.9 billion of revenue [29].

Geographic concentration of the slowdown is not the same as absolution. The United States is where Amazon's DSP is scaled, where retail media is scaled, and where agency principal-based buying is concentrated — the pressures catalogued in Share or Cycle. A domestic-only deceleration is equally consistent with international being three or four years behind the same sequence. The company has been describing international expansion in the same words since at least FY2021 [30] [31], which is a long runway but also a long time to reach 14.5%.

The read that fits the evidence: the FY2025 and Q1 2026 deceleration is a US phenomenon, not a platform-wide one. US revenue grew 5.4% in the March quarter against international's 54.8%, and a global product losing to global competitors would not accelerate to 45%-plus in EMEA and APAC while stalling at home. The read is held loosely because a single agency-entity rebooking could produce the same table. Two disclosures would settle it: a second and third quarter of the recast 10-Q split holding at or above 18% international, which a one-off contract migration would not produce, and any reinstatement of gross billings by geography, which would separate volume from rate. A restoration of US growth to double digits without international decelerating would make the whole question moot; international slipping back toward 15% while the US stays near 5% would leave the company growing at low single digits with the mix argument gone.

---

What the cash actually leaves behind

The Trade Desk turns revenue into cash efficiently: $783 million of free cash flow in 2025 [1] on $2.9 billion of revenue [2]. Stock compensation of $491 million takes most of it. Three years and $2.27 billion of repurchases lowered the share count by 3.0%. The headline compensation charge has held flat for four years while the CEO Performance Option ran off, and that option finished amortising in March 2026.

Sources: FY2025 Annual Report (Form 10-K), Consolidated Statements of Cash Flows [3] and Consolidated Statements of Stockholders' Equity [4].

The cash ladder

Operating cash flow of $993 million in 2025 came from net income adjusted for non-cash items of $1.27 billion, less a $275 million net investment in working capital [5]. The add-backs are dominated by one line: $490.6 million of stock-based compensation, against $115.8 million of depreciation and amortisation and $167.7 million of deferred income taxes [6].

Free cash flow on the company's own definition — operating cash flow less capital expenditure and capitalised software — was $783 million. Subtracting stock compensation, which is a real cost settled in shares rather than cash, leaves $292 million.

Source: derived from Consolidated Statements of Cash Flows, FY2023 and FY2025 Annual Reports (Forms 10-K) [7] [8].

The series matters more than the 2025 point. Free cash flow after stock compensation was negative in 2021 and 2022, crossed zero in 2023, and reached $292 million in 2025 [9]. On a trailing-twelve-month basis through March 2026 it is $358 million, on $829 million of free cash flow and $471 million of stock compensation [10].

GAAP already carries this cost. Net income of $443 million and diluted EPS of $0.90 [11] are struck after the $491 million charge; adjusted EBITDA of $1,196 million and adjusted EPS are not. The reconciliation the company publishes adds back stock compensation as the single largest item — $490.6 million of the $753 million bridge from net income to adjusted EBITDA [12].

Stock compensation has been flat for a reason

Headline stock compensation has barely moved in four years: $498.6 million in 2022, $491.6 million in 2023, $494.7 million in 2024, $490.6 million in 2025, while revenue grew 84% [13] [14] [15] [16]. As a share of revenue the charge fell fast over that stretch: 31.6% in 2022 to 16.9% in 2025.

That ratio is most sensitive to a single award. Stock compensation other than the CEO Performance Option held a 14.6% to 15.1% band of revenue in all four years, so the movement in the ratio came from the option alone. The CEO Performance Option granted in October 2021 — 16 million target shares, up to 19.2 million, at a $68.29 exercise price, with a grant-date fair value of roughly $819 million — has been amortising on a graded-vesting schedule, and its annual charge fell from $262 million to $198 million to $128 million to $67 million [17] [18].

Source: derived from Note 10 — Stock-Based Compensation, FY2023 and FY2025 Annual Reports (Forms 10-K) [19] [20].

Everything other than that option has grown from $237 million to $424 million, tracking revenue almost exactly inside that band. The falling ratio was arithmetic from a single expiring grant, not operating leverage.

That grant is now finished. Stock compensation for the CEO Performance Option was $5 million in the first quarter of 2026, and the company states that as of March 31, 2026 the expense "had been fully recognized" [21]. From here the compensation line has no declining component to offset growth in the rest.

Two facts push in opposite directions on what comes next. Grant activity has accelerated at much lower prices: 5.65 million restricted shares were granted in the first quarter of 2026 at an average grant-date fair value of $33.26, against 7.93 million for the whole of 2025 at $54.26, taking unvested restricted stock from 11.6 million shares to 15.5 million in three months, with a further 2.68 million options granted at a $25.30 exercise price [22] [23] [24]. Delivering a fixed dollar of compensation at $17 rather than $54 requires roughly three times the shares. Against that, the $915 million of unrecognised compensation still to be expensed — $749 million on restricted stock over three years, $154 million on options, $12 million on the purchase plan — was struck at grant-date values far above the market, so the reported charge will overstate what employees are actually receiving [25]. The share count is where the two effects settle.

Below the executive tier, the same gap shows up in what employees are holding. The 15.5 million unvested restricted shares outstanding at March 2026 were booked at a weighted-average grant-date value of $52.57, about $816 million of accounting value against roughly $268 million at the $17.29 price [26]. The 12.6 million ordinary options outstanding carry a weighted-average exercise price of $42.32, and the tranche granted in the first quarter at $25.30 sits below that but still above the market, so none of the option pool has intrinsic value at the current price [27]. No option exchange, repricing or make-whole grant appears anywhere in the corpus, and the employee stock plan does not permit repricing without stockholder approval [28].

Holding staff whole in value therefore costs shares rather than dollars. Delivering $490 million of grant-date value, the level of each of the last four years, takes about 9 million shares at $54.26 and about 28 million at $17.29, against 471.0 million outstanding. The buyback cannot absorb that difference on what is left of its authorisation: the $327 million available at March 31, 2026 retires about 19 million shares at $17.29 [29], more than two years of the 6.04 million shares issued in 2025 but under a year of the 28 million a constant grant-date dollar would now require. Either the company issues materially more shares to hold employees at the same value, or it delivers less value at the same share count and carries the retention risk instead.

The share-count ledger

The repurchase programme, authorised in February 2023, is not framed by the company as a return of capital. Its stated design is "to help offset the impact of future share dilution from employee stock issuances" [30]. Judged on that objective, the record is checkable line by line.

Source: Consolidated Statements of Stockholders' Equity, FY2023 and FY2025 Annual Reports (Forms 10-K) and Q1 FY2026 Form 10-Q [31] [32] [33].

Across the three fiscal years 2023 to 2025 the company retired 38.84 million shares for $2,271.7 million, an average of $58.48 [34] [35]. Over the same period 24.30 million shares went out to employees and one small acquisition. The count fell from 490.5 million to 475.9 million — 14.5 million shares, or 3.0%. Including the first quarter of 2026, the totals are $2,445.8 million spent, 19.5 million net shares removed, and a 4.0% reduction from the end of 2022. That spending equals 30% of the company's market capitalisation at the July 24, 2026 close of $17.29.

The programme did what it was designed to do, and management has said so — on the third-quarter 2025 call the company described having "repurchased nearly $2 billion through our repurchase program, effectively offsetting dilution and reducing shares outstanding over that time" [36]. The claim is accurate. What it costs is the part worth holding onto: $156 of cash per net share removed over 2023 to 2025, against a share now quoted at $17.29.

Source: Consolidated Statements of Stockholders' Equity, FY2023 and FY2025 Annual Reports (Forms 10-K) [37] [38].

The shape of that line is the honest summary. Through 2024 the count was still above where it started in 2022, because option exercises and vesting outran a $236 million year of buying. The 2025 decline came from a $1.39 billion year of repurchase at an average of $52.96 — a price the stock has since fallen 67% below. The 2026 purchases at $24.51 are much better value on the same arithmetic, and the company has said it plans "to continue opportunistic share repurchases while also offsetting dilution from employee stock reissuances" [39]. Only $327 million of authorisation remained at March 31, 2026 [40].

One balance-sheet consequence is visible: because repurchases are charged against retained earnings, the company moved from $665.5 million of retained earnings at the end of 2022 to a $724.9 million accumulated deficit at March 2026, and book equity fell from $2,949 million to $2,453 million across five profitable quarters [41] [42].

Capital spending stepped up while cash capex understated it

Capital expenditure has quadrupled in two years — $46.8 million in 2023, $98.2 million in 2024, $197.0 million in 2025 — and the first quarter of 2026 alone consumed $112.7 million, 16% of quarterly revenue [43] [44]. Depreciation ran at $96 million in 2025, so the reported charge lags the spend, and $117.9 million sat in construction in progress at year end [45].

Source: derived from Consolidated Statements of Cash Flows, FY2023 and FY2025 Annual Reports (Forms 10-K) [46] [47].

The cash figure understates 2025's investment. Supplemental disclosure shows capitalised assets financed through accounts payable of $104.5 million at the end of 2025 against $20.5 million a year earlier [48]. Capital expenditure incurred was therefore about $281 million, not $197 million, and the unpaid $84 million sits inside the $291 million rise in accounts payable that helped operating cash flow. Gross property and equipment corroborates it, rising $234.7 million before disposals [49]. On an incurred basis, 2025 free cash flow was roughly $699 million rather than $783 million — a timing effect that reverses, not a permanent gap, but one that flatters the year the multiple is being struck on.

Working capital runs the wrong way, which helps in a slowdown

The Trade Desk bills clients for gross media spend and pays suppliers for it, so both sides of the balance sheet are far larger than revenue: $3.77 billion of receivables against $3.01 billion of payables at December 2025 [50]. The company states plainly in its liquidity discussion: "We typically pay suppliers in advance of collections from our clients" [51], and the reported cycle confirms it: days sales outstanding of 92 against days payable of 77 at the end of the third quarter of 2025 [52].

That is a negative float, and it means growth consumes cash rather than generating it. In each of 2023, 2024 and 2025 the increase in receivables exceeded the increase in payables — by $79 million, $175 million and $142 million respectively [53] [54]. The mechanism runs in reverse too. In the first quarter of 2026 receivables fell $429 million and payables fell $279 million, and operating cash flow of $392 million was collected on $40 million of net income [55]. A business whose volume growth is decelerating (Share or Cycle) will release working capital as it slows — which is one reason the cash position held at roughly $1.4 billion through the deceleration [56].

The same price against successively stricter cash

At $17.29 on 471.0 million shares, with $1,406 million of cash and short-term investments and no drawn debt, the enterprise value is about $6.74 billion [57] [58]. Applied to different definitions of what the business earns, that one number produces a wide range.

Source: derived from FY2025 Annual Report (Form 10-K) cash flow statement, statements of operations and Adjusted EBITDA reconciliation and Q1 FY2026 Form 10-Q [59] [60] [61] [62].

The GAAP price-to-earnings multiple of 19.2x on 2025 diluted EPS of $0.90 [63] sits in the same zone as the after-compensation cash multiples, because GAAP earnings already charge stock compensation. The gap between 5.6x and 23.1x turns on which measures charge stock compensation. Adjusted EBITDA adds the $491 million back, and the cash-flow measures never deduct it because it is settled in shares; GAAP operating income, adjusted EBITDA less stock compensation and free cash flow less stock compensation all carry it in full.

The read

On the evidence here, the business generates genuine cash but converts less of it to per-share value than the headline figures suggest. Stock compensation has held near 15% of revenue for four years once the expiring CEO option is stripped out; the buyback has absorbed $2.45 billion to lower the count 4.0%; and the price is not cheap on any measure struck after that cost. The multiples that look most attractive — 5.6 times adjusted EBITDA, 8.6 times free cash flow — are the ones that treat the compensation as free.

The strongest fact against that read is the direction of travel. Free cash flow after stock compensation went from negative $42 million in 2022 to $358 million on a trailing basis through March 2026, and management expects a 2026 adjusted EBITDA margin of at least 40% while holding headcount growth below revenue growth [64]. If share-based pay stays near $490 million in absolute dollars while revenue compounds, the burden falls mechanically, and the after-compensation multiple compresses fast.

What would settle it is observable in two lines of the next few filings. First, the stock compensation charge for full-year 2026: with the CEO option fully expensed, a figure that holds near $490 million means the underlying cost stopped growing, while a figure above $520 million means the run-rate has simply been unmasked. Second, the share count: shares outstanding of 471.0 million at March 2026 falling through 2026 on the remaining $327 million of authorisation would show repurchases finally outrunning issuance at a price where the arithmetic works; a count that drifts back up would show grant volumes at $17 doing what grant volumes at $54 could not.

---

What the founder owns

Jeff Green has increased his stake in The Trade Desk with his own cash, and he also has more control than his economics support. He owns 11.2% of the shares and 49.7% of the votes, and in March 2026 he bought $148.1 million of stock in the open market. Between September 2025 and May 2026, holders of the Class A shares voted against extending his control and against his pay, and lost both.

Beneficial ownership at March 6, 2026 splits cleanly. Green held 11,483,384 Class A shares — 2.6% of that class, including options exercisable within sixty days — and 42,071,879 Class B shares, or 97.6% of the supervoting class. Together that is 53.6 million shares out of 478.0 million outstanding, 11.2% of the equity, carrying 49.7% of the votes because each Class B share carries ten [1]. All eight current executive officers and directors together hold 49.8% of the votes on 3.0% of the Class A shares [2]. Green's own Schedule 13G/A, filed May 15, 2026, reports 11.3%, up from 9.7% in November 2024 [3].

The increase was bought, not granted. Across three sessions from March 2 to March 4, 2026, Green purchased 6,000,000 Class A shares in four open-market transactions — 527,324 at $23.49, 1,472,676 at $24.16, 1,685,696 at $24.97 and 2,314,304 at $25.08 — for $148.1 million at an average of $24.68 [4]. At $17.29 that block is worth $103.7 million, a paper loss of $44.4 million, or 30%. The company's insider trading policy bars pledging shares as collateral for a loan or holding them in a margin account, so the purchase was not levered against the stock [5].

Sources: 2026 Definitive Proxy Statement, beneficial ownership as of March 6, 2026 [6]; Form 4 filings, March 2026 [7].

The Form 4 disclosing the purchase was filed on March 4, 2026. The shares closed at $25.17 that day and $29.79 the next, an 18% move on 82.5 million shares against roughly 20 million in each of the preceding sessions. On March 5, Kathryn Falberg — appointed chair of the audit committee in April 2025 [8] — sold 152,828 shares at $30.45 and $30.48, raising $4.66 million at an average 23% above the price Green had just paid [9]. She resigned from the board eighteen days later.

The 2025–2026 vote record

Three ballots in the past year separate the founder's alignment from his accountability. Each was decided by the ten-vote shares, and on each the Class A holders can be counted precisely. The method is to assume the Class B shares present were voted in favour and to net their votes out of the disclosed totals; the residual then has to equal the Class A votes present, and on all three ballots it does, to the single share. An exact reconciliation is not proof, but it is hard to produce by accident.

At the special meeting of September 16, 2025, stockholders approved amending the articles of incorporation to move the date on which all Class B shares automatically convert to Class A. Of 356,794,733 shares present carrying 746,006,717 votes, 516,037,827 were cast for, 228,364,796 against and 1,604,094 abstained [10]. Solving the two-class arithmetic gives 43,245,776 Class B shares present, worth 432,457,760 votes; net those out and the Class A shares present voted 83,580,067 for and 228,364,796 against, a residual that with the 1,604,094 abstentions sums back to the 313,548,957 Class A votes present exactly. Roughly 73% of Class A votes cast opposed the extension. It carried with 69% of the votes cast.

At the annual meeting of May 4, 2026, the same decomposition holds. Say-on-pay drew 509,592,084 votes for and 172,498,666 against, with 66,028,847 broker non-votes; stripping out the 430,718,790 Class B votes present leaves Class A holders 78,873,294 for and 172,498,666 against — again reconciling exactly to the 252,104,037 Class A votes cast [11]. Some 69% of Class A votes rejected the compensation programme; it passed with 75% overall.

The third ballot needs no arithmetic. The articles reserve one board seat to be elected by Class A holders voting separately as a class while the dual-class structure stands [12]. Andrea Cunningham, the incumbent in that seat and a member of the committee that negotiated the control extension, drew 82,759,848 votes for and 169,344,189 withheld — 67% of the class withholding from the only director it elects. Under plurality voting she was elected [13].

Sources: derived from disclosed vote totals by netting out the Class B votes present — Form 8-K, September 17, 2025, reporting 516,037,827 votes for and 228,364,796 against [14]; Form 8-K, May 8, 2026 [15]. Percentages exclude abstentions; the Class A director seat is elected by Class A holders alone.

The Class A base is not hostile to Green himself. On the same ballot, his own re-election drew 200,811,565 Class A votes for against 51,292,472 withheld — 80% support [16]. Minority holders back the founder and decline to endorse the machinery that is supposed to price him.

Sources: derived from disclosed vote totals, Form 8-K September 17, 2025, reporting 516,037,827 votes for and 228,364,796 against [17] and Form 8-K May 8, 2026 [18]; both bases measured as votes for divided by votes for plus against.

The ten-year extension

The September 2025 vote mattered because the clock was about to run out. Under the articles then in force, every Class B share was to convert automatically into Class A on the first trading day on or after December 22, 2025 — three months after the meeting. The amendment moved that date to December 22, 2035, a ten-year extension, and added a waiver of the right to jury trial for internal actions under Nevada law [19].

The negotiation is on the record, meeting by meeting. A special committee of three independent directors — Lise Buyer as chair, Andrea Cunningham and Alex Kayyal [20] — engaged its own counsel and financial adviser [21]. Green opened on May 7, 2025 by proposing to eliminate the sunset entirely, soften the trigger tied to his own departure, and create a class of non-voting Class C shares to dividend out to existing holders. The committee countered on May 20 with a seven-year extension to December 22, 2032, plus two governance concessions: an annual advisory vote on executive pay, and a bylaw letting the lead independent director convene the independent directors [22]. Green came back on June 5 with ten years, and with the clarification that those special meetings of independent directors would be for discussion purposes only. The committee approved his terms five days later, and the board adopted them [23].

So the committee's opening number was seven years and the outcome was ten, on the counterparty's terms. What Class A holders received in exchange was the annual say-on-pay vote they then used, eight months later, to reject the pay programme by 69% — a vote the proxy correctly describes as non-binding.

Nasdaq took a view of its own. On December 9, 2025 the exchange sent a letter of reprimand finding that the amendment violated its voting-rights rules 5640 and IM-5640. The company recorded that it does not concur, did not appeal, and noted the dual-class structure would not change as a result [24]. The 8-K records that the Nasdaq staff concluded it was appropriate to close the matter with the letter, issued in accordance with Nasdaq Rule 5810(c)(4), with no further action to be taken on its part, and that the listing of the Class A shares is unaffected.

Litigation over the structure and over the 2024 move from Delaware to Nevada is still live. A books-and-records action, Scarantino v. The Trade Desk, went to trial in the Delaware Court of Chancery in July 2025; the Vice Chancellor denied the stockholder's exceptions in December 2025 and the appeal to the Delaware Supreme Court remains pending [25].

Pay through the drawdown

Green's summary-compensation total went from $6,756,299 in 2024 to $27,431,583 in 2025 — a fourfold rise in the year the shares fell. The 2025 figure comprises $1,350,000 of salary, $11,589,471 of stock awards, $11,590,864 of option awards, $2,804,775 of non-equity incentive pay and $96,473 of other compensation [26]. Against a median employee's total of $218,847, that is a ratio of 125 to 1 [27].

The company's own pay-versus-performance table sets the two series side by side. A $100 investment made at the end of 2020 was worth $146.73 at the end of 2024 and $47.39 at the end of 2025, while the peer group index rose to $137.99 [28].

Source: 2026 Definitive Proxy Statement, Pay Versus Performance Table, as calculated under Item 402(v) of Regulation S-K [29]. TSR indices show the value of $100 invested at December 31, 2020.

That table also carries the strongest fact against reading the pay rise as unearned. Compensation actually paid — the SEC's mark-to-market measure, which revalues outstanding awards each year — was negative $856.8 million for Green in 2025, and negative $620.3 million in 2022. Whatever the reported figure says, his paper wealth tracks the stock closely, and the 2025 decline is roughly 1.9 times the company's entire net income for that year. The measure is a valuation construct rather than cash, and the 2021 entries are dominated by the initial valuation of a single option grant, so the direction and size of the annual swings carry more information than any total.

Two features of the 2025 award decisions are worth naming. The first is the metric. The cash incentive plan ran on revenue alone, with a target of $2,935 million; actual revenue of $2,896 million produced $2,351,404 for Green against a $2,700,000 target, or 87% [30]. In October 2025, with the fourth quarter under way, the compensation committee adopted a Supplemental Executive Bonus Plan because, on its projections at the time, it "determined it was necessary to provide additional incentive for engagement and achievement" [31]. Fourth-quarter revenue of $847 million beat the supplemental threshold of $830 million [32] and added $453,371 [33]. A year that missed its revenue target by 1.3% paid the CEO 104% of his target bonus. The choice of metric also has a second edge: revenue rises with the company's cut of client spend as well as with the spend itself, so a plan keyed to revenue alone pays out on a higher take rate even in a year when the volume crossing the platform grows far more slowly.

Sources: 2026 Definitive Proxy Statement, Annual Cash Incentive Plan Formula [34] and Supplemental Incentive Plan Award Payments [35].

The second is the form of the equity. The board approved a $30 million target award for Green in April 2025, delivered as 235,367 restricted shares and 450,045 options with an aggregate grant-date fair value of $23,180,335. The restricted shares vest in sixteen equal quarterly instalments and the options monthly over four years; neither carries a performance condition [36]. The proxy is explicit about why these top-ups exist alongside the 2021 mega-grant: annual awards were made in 2023 and 2025 but not 2022 or 2024, another was approved for 2026, and the committee cites "the importance of maintaining a baseline annual compensation level sufficient to provide consistent motivation as market conditions vary from year to year" [37]. The 2026 grant landed on March 3, 2026 — 398,089 restricted shares and 737,028 options, the same week Green was buying in the market [38].

The 2021 grant itself is now far out of the money. It covers 16,000,000 target shares struck at $68.29, vesting in eight tranches on average closing prices from $90 to $340 [39]. Two tranches have vested, 4,800,000 shares in total, in 2021 and 2024; no other price target has been met [40]. At December 31, 2025 Green held 3,385,150 exercisable shares under it plus 14,400,000 unearned, all at $68.29 against a $37.96 close [41]. At $17.29 the next tranche needs an 8.4-fold move and the last a 19.7-fold move. The design has done what it was built to do — it pays nothing until holders are paid first — and the practical consequence is that the instrument meant to align the CEO now exerts almost no pull, which is precisely the gap the time-vesting top-ups fill. The 14,400,000 unearned shares stay tied to the $68.29 strike, while the 2025 and 2026 annual awards vest on the passage of time alone.

Oversight and turnover

The apparatus that sets and polices this pay has thinned faster than the pay has moved. The chief financial officer's chair has changed hands three times in under eleven months. Laura Schenkein left the role on August 21, 2025 and was succeeded by Alex Kayyal, a sitting director since February 2025; Kayyal ceased serving on January 24, 2026 [42]. Chief accounting officer Tahnil Davis held the post on an interim basis until Nate Olmstead took it on July 9, 2026 [43].

Kayyal's 2025 compensation, covering 156 days as CFO plus the earlier months as a non-employee director, totalled $12,940,803 [44] — including a $600,000 signing bonus and $11.4 million of stock and option awards — plus a separation package of $1,200,000 in salary and bonus, $400,000 to relocate to the United Kingdom, $39,717 of COBRA, $15,992 of accrued vacation, and twelve months of equity acceleration worth $2,607,132 at grant-date fair value [45]. Schenkein's exit added $1,000,000 of severance, $623,283 of incentive pay, $71,754 of her legal fees, $46,154 for unused sabbatical, and $5,878,074 of accelerated equity value recognised in 2025 [46].

Kayyal's appointment also removed one of the three independent directors negotiating opposite Green. He resigned from the special committee on becoming CFO on August 21, 2025 [47]. The committee had already delivered its recommendation on June 10; the shareholder vote was still four weeks away.

Four directors resigned over five weeks. Net of Andrew Vollero's appointment effective April 3, that took the board from eight members to five, rebuilt to seven by July 2026. Gokul Rajaram gave notice on March 3, 2026, effective April 3 [48]. Kayyal resigned from the board on March 19 and Falberg on March 23, both effective immediately [49]. Lise Buyer, who had chaired the special committee, gave notice on March 31, effective April 3 [50]. Each filing states the resignation was not the result of a disagreement with the company, and none gives a reason.

The audit committee that had been Falberg, Buyer and Rajaram at the 10-K filing was reduced to a single member [51]. By the April 2026 proxy, Andrew Vollero was "the sole member" and its chairperson [52]. On March 24 the company told Nasdaq it no longer complied with Listing Rules 5605(c)(2)(A) and 5605(d)(2)(A), which require three independent directors on the audit committee and two on the compensation committee, and received a notice of noncompliance with a cure period running to September 21, 2026 [53].

The rebuild has been quick. Vollero joined effective April 3, 2026 [54]; David Haddad joined the board and the audit committee effective June 11, taking the board from five to six [55]; Penry Price joined effective July 9 as an audit committee member and chair of the compensation committee, taking it to seven [56]. On the 8-Ks' own arithmetic the audit committee reached three members on July 9, ten weeks before the deadline. No filing in the corpus records Nasdaq confirming that compliance has been restored.

Sources: Forms 8-K filed March 2025 through July 2026 [57] [58] [59] [60] [61] [62]; 2026 Definitive Proxy Statement [63]; Form 4 filings [64].

One more piece of the record cuts against management. The consolidated securities class action's amended complaint adds a claim under Section 20A alleging that the chief executive, the then-chief financial officer and the chief strategy officer traded on inside information during a class period running from November 2023 to August 2025; the court denied the motion to dismiss on March 17, 2026 [65]. Surviving dismissal is not a finding of liability. Set against it, the proxy records that all Section 16 reports were filed on time in 2025, and that Green reimbursed the company roughly $628,000 for personal use of aircraft rather than taking it as a perquisite — the company, in turn, reimbursed him about $1,000,000 of legal fees for litigation naming him as CEO [66].

What would change the read

The evidence supports a specific and narrow conclusion: on this record, the founder's economic alignment is unusually strong and the independent check on him is unusually weak, and the two facts are separate rather than offsetting. A holder buying at $17.29 is buying alongside someone who paid $24.68 with his own cash four months earlier and who cannot exit without collapsing his own control premium. That holder is also buying into an entity where the Class A majority has been outvoted on the two questions it was allowed to answer, where the board lost four members in five weeks, and where the audit committee spent a quarter of 2026 below the exchange's minimum size. For an investor who prizes founder ownership, the first fact is the attraction and the second is its price — and the second is what determines whether a discount to intrinsic value can ever be closed by anyone other than Green.

Three things would move this assessment. Continued open-market buying by Green at prices near the current level would strengthen the alignment read materially, since the March purchase is a single event and Form 4 filings will show whether it repeats. A compensation committee that in 2026 either drops the mid-year supplemental mechanism or attaches a performance condition to the CEO's annual grant would show the new chair changing the substance rather than the composition. And a Nasdaq confirmation that both committees are compliant, together with two or more genuinely independent additions who are not recruited into management, would restore the check the September 2025 bargain was supposed to buy. Absent those, the arithmetic of the last three ballots is the best available guide to how the next contested decision will go: the Class A shares can register a view and cannot carry one, and the reader can weigh that alongside the cash generation and share count set out in Cash and Share Count.

---

Downside and the Floor

The Trade Desk has no funded debt, $1.41 billion of cash and securities, and a revolver its bank syndicate enlarged and repriced downward in April 2026. On a frozen cost base, revenue would have to fall 41% before adjusted EBITDA reached zero. The cost base has never been frozen: in the one quarter revenue ever fell, operating expenses rose 21%.

What has to go wrong, and by how much

FY2025 revenue was $2,896.3 million against total operating expenses of $2,307.0 million, leaving $589.3 million of operating income [1]. Of that expense base, $490.6 million was stock-based compensation [2] and $115.8 million was depreciation and amortisation [3]. Strip both out and the cash the company actually paid to run itself in FY2025 was $1,700.6 million — 59% of revenue.

That number sets the boundary conditions. Holding every FY2025 cost dollar flat, three thresholds follow arithmetically: GAAP operating income reaches zero at $2,307 million of revenue, 20.4% below FY2025; adjusted EBITDA covers FY2025's $209.8 million of capital expenditure and capitalised software at $1,910 million, 34.0% below [4]; and adjusted EBITDA reaches zero at $1,701 million, 41.3% below.

Sources: FY2025 cost base and capital expenditure per the FY2025 Form 10-K [5] [6]; thresholds derived; FY2026 and FY2027 estimates from consensus data, as reported.

The lowest FY2027 revenue estimate carried in the consensus set is $2,629 million — 9.2% below FY2025 actual and 17.3% below the FY2026 consensus of $3,177 million, against an FY2027 average of $3,482 million. The most pessimistic published forecast of this business therefore sits roughly half the distance to the GAAP breakeven line and less than a quarter of the way to the adjusted EBITDA line. A frozen cost base bounds the downside rather than forecasting it, and the bound is wide: 20.4% of revenue decline to GAAP breakeven, 41.3% to zero adjusted EBITDA.

The quarter revenue fell

There is one precedent in the public record. In the June 2020 quarter revenue fell to $139.4 million from $159.9 million, down 12.9%, and a $31.9 million operating profit became a $15.8 million operating loss. Operating expenses that quarter were $155.1 million against $128.0 million a year earlier — up 21% while revenue fell 13%.

Source: quarterly figures as reported in company filings; operating expenses derived as revenue less operating income. The annual context is in the FY2021 Form 10-K [7].

The recovery came from revenue, not from cost action. September-quarter revenue rebounded 32% year over year, and the full year still grew: revenue of $836.0 million, up 26%, and operating income of $144.2 million, up 29%, on operating expenses that themselves rose 26% [8]. The company later described the episode in its risk factors as "the initial decline in revenue in response to the COVID-19 pandemic" [9].

The counter-fact to reading that as a template: the 2020 company was roughly a fifth of today's size, the shock lasted a single quarter, and cutting into a two-month advertising pause would have been the wrong decision. What the episode does establish is that the observed elasticity of this cost base to a revenue decline, over one quarter, was zero.

Two years out, at the low end

The question the thresholds leave open is what management does with the cost base if revenue disappoints for longer. The grid below holds the FY2025 cost structure and grows it at three rates through FY2027, against three revenue outcomes.

Source: derived from the FY2025 cost base — total operating expenses of $2,307.0 million, of which $490.6 million stock compensation and $115.8 million depreciation and amortisation [10] [11] [12] — grown at the stated rates; revenue cases from consensus estimates, as reported. Figures in $ millions.

At the street's lowest FY2027 revenue estimate with costs frozen, the business still earns $928 million of adjusted EBITDA and $322 million of GAAP operating income; the $6.74 billion enterprise value at $17.29 — 471.0 million shares less $1.41 billion of net cash [13] — is 7.3 times that adjusted EBITDA. A cost base compounding at 12% a year instead produces $496 million of adjusted EBITDA and a $265 million GAAP operating loss on the same revenue. Adjusted EBITDA never turns negative anywhere in the grid; GAAP operating income turns negative in one cell and reaches breakeven in another. The difference between the two measures is stock compensation, which does not shrink when revenue does — and at $17.29 a share, holding a given team costs more shares than it did.

The current run rate sits at the wrong end of that range. First-quarter FY2026 operating expenses were $622 million, up 11%, and $513 million excluding stock compensation, up 18%, against revenue growth of 11.8% [14] [15]. Management's stated plan for the rest of the year is headcount growth below revenue growth, and the finance chief describes the model as one in which "we generate strong cash flow and can maintain significant flexibility in how we pace our investments and expenses" [16]. That flexibility is asserted rather than demonstrated; the FY2026 margin commitment examined in Financials and Estimates is the first place it becomes checkable.

The tax that has not been paid yet

Free cash flow is the denominator most of this report's valuation work rests on, and one input to it has been running below its own accrual for three years. FY2025 pretax income was $658.8 million and the total provision $215.5 million — a 32.7% effective rate — but the current portion of that provision was only $42.7 million, against $189.6 million in FY2024, with $172.7 million booked as deferred [17].

Source: FY2025 Form 10-K, Note 11 Income Taxes [18] [19].

Cash income taxes paid were $150.1 million in FY2025, $158.6 million in FY2024 and $151.9 million in FY2023 — essentially unchanged while pretax income went from $268 million to $659 million [20]. The cash tax rate on pretax income fell from 57% to 23% across those three years.

What financed the gap is now largely spent. Net deferred tax assets fell from $230.2 million to $55.7 million during FY2025, with the capitalised software development component collapsing from $181.9 million to $6.8 million [21], because the One Big Beautiful Bill Act allowed immediate deduction of previously capitalised domestic research costs; the company recognised $175 million as an income-tax receivable against that election [22]. Part of that receivable was collected in the March 2026 quarter, and the provision that quarter rose $14 million on tax detriments from vesting employee stock awards, taking the quarterly effective rate to 49% [23] [24].

FY2026 cash taxes are therefore flattered by a one-off refund, and the shield behind them has shrunk to $55.7 million. On FY2025's pretax income at FY2025's book rate, the tax bill is about $215 million rather than the $150 million paid — roughly $65 million a year of cash that the recent free-cash-flow record did not have to find. Against FY2025 free cash flow of $783 million on the company's own definition — operating cash flow of $992.7 million less $197.0 million of capital expenditure and $12.8 million of capitalised software [25] — that is an 8% haircut before any revenue assumption is made. The counter-fact is that the effective rate itself is inflated by non-deductible stock compensation, which added 7.1 points in FY2025 [26] — if the share price recovers, that penalty shrinks.

What does not flex

At March 31, 2026 non-cancelable contractual obligations totalled $1,115.9 million: $777.7 million of operating leases for offices and hosting facilities and $338.2 million of other commitments for hosting services, hardware, data providers and software, with $288.3 million of the total falling due in the remainder of 2026 [27]. At the December 2025 year end the lease ladder ran to $795.2 million undiscounted on a 5.5-year weighted-average term, of which $291.0 million related to leases not yet commenced [28]. In January 2026 the company added $92 million of non-cancelable cloud hosting and data commitments running to 2028 [29].

Total lease cost has risen from $61.5 million in FY2023 to $75.4 million in FY2024 to $94.4 million in FY2025 [30]. The data-centre build described in Cash and Share Count as a capital-expenditure step-up also arrives as a multi-year lease obligation, and leases signed for capacity not yet commenced are the part of the cost base least responsive to a revenue disappointment.

The floor

Sources: cash of $878.4 million and short-term investments of $527.5 million with no drawn debt at March 31, 2026 [31]; the $750 million facility less $5 million of outstanding letters of credit [32] [33]; breakeven derived.

On April 14, 2026 — with the shares roughly 87% below their December 2024 high — a syndicate led by JPMorgan replaced the $450 million revolver that was due to mature that June [34] with a $750 million facility running to April 14, 2031, plus the right to add a further $750 million on additional lender commitments [35] [36]. The terms moved in the borrower's favour on every disclosed dimension: the SOFR margin grid narrowed to 1.125%–1.500% from 1.25%–2.25%, and the undrawn commitment fee to 0.125%–0.200% from 0.200%–0.350% [37] [38]. The leverage test is now struck on funded debt net of up to $250 million of unrestricted cash, where the old one was gross [39] [40].

The executed agreement is in the filing record, and its negative covenants run to six sections — non-guarantor subsidiary indebtedness, sale and leaseback transactions, fundamental changes, use of proceeds, liens, and outbound investment rules — under a single financial covenant, a total leverage ratio of 3.50 to 1.00 [41] [42]. The 2021 agreement it replaces listed negative covenants running from Section 8.01 to Section 8.26 [43], among them Section 8.10, Dividends, Redemptions, Distributions, which barred the company and its subsidiaries from paying dividends or repurchasing their own equity except through an enumerated list of permitted exceptions [44]. With nothing drawn, the leverage test is not binding, and the restriction on repurchases has gone; $327 million of buyback authorisation remained at March 31, 2026 [45].

The honest qualification is that the old facility matured in June 2026, so a refinancing was scheduled rather than opportunistic. The evidence is in the terms rather than in the fact of the deal: a larger facility, a maturity running to 2031, a narrower SOFR margin grid and a lower undrawn commitment fee. A lending syndicate repriced this credit cheaper, longer and larger while the equity market was marking the enterprise at 5.6 times FY2025 adjusted EBITDA, and it retained a lien on substantially all assets with release on investment-grade ratings [46].

Where the balance-sheet risk actually sits

The exposure is in the receivable book. Accounts receivable were $3.32 billion at March 2026 against an allowance for credit losses of $16.2 million — 0.49% of the gross, up from 0.32% three months earlier [47]. At December 2025 two clients each accounted for at least 10% and together for 30% of consolidated receivables, and two suppliers for 34% of payables [48]. Under sequential liability an agency is not liable to the company if its advertiser does not pay, while the company is "contractually required to pay advertising inventory and data suppliers within a negotiated period of time, regardless of whether our clients pay us on time, or at all" [49].

That is the one event that could consume a material share of the $1.41 billion of net cash in a short period, and it is the concentration risk documented in Share or Cycle arriving through the balance sheet rather than the income statement. Two facts cut against reading it as acute: the concentration measured on receivables fell over the year — three clients were 42% of the book at December 2024 against two at 30% a year later [50] — and the counterparties concerned are the largest agency holding companies in the industry, not marginal credits. The reserve of 0.49% is the company's own assessment that the book is good; it has never been tested by a holding-company failure.

What would change the read

The judgment here is that at $17.29 the equity is priced against an earnings question, not a survival question, and the record supports that: no funded debt, $1.41 billion of net cash, a five-year facility signed at improved terms, and 41 points of revenue decline between FY2025 and an adjusted-EBITDA loss. The fact that argues hardest the other way is the elasticity evidence — in the only quarter revenue has ever fallen, costs rose 21%, and in the most recent quarter cash costs grew 18% against 12% revenue growth.

That judgment is about the balance sheet. The growth half of the report's opening question is answered across the other chapters, and the pieces do not point the same way: the deceleration is a United States phenomenon on the recast geography, where international revenue is still compounding at a multiple of the domestic rate (US and International); video is still compounding while mobile contracts (CTV and Channel Mix); the rate lever that carried reported revenue past gross spend in FY2025 is real but has a visible end (Pricing and Take Rate); and the cost base has never contracted with revenue. Against the $6.74 billion enterprise value, an FY2027 outcome anywhere inside the street's own range — $2,629 million at the low end, $3,482 million on average — is worth 7.3 times adjusted EBITDA at the low estimate with costs frozen and 3.8 times at the average, or 13.6 and 5.0 times if costs compound at 12% a year. What stands behind those multiples is arithmetic rather than judgment: more than 20 points of revenue decline separate FY2025 from a GAAP operating loss on a frozen cost base and more than 40 from an adjusted-EBITDA loss, so an outcome at the low end of the street range costs the holder multiple compression rather than a funding problem. What none of it settles is whether the volume line has stopped decelerating; the thresholds size the downside, they do not date the recovery.

Four lines in the filings will show whether that read holds:

Platform operations as a share of revenue, in the MD&A results table — 21.4% in FY2025 [51] and 26.4% in the March 2026 quarter [52]. A third consecutive quarter above 25% alongside revenue growth below 10% would mark the cost base as structurally heavier, not cyclically so.

Cash paid for income taxes, net of refunds, in the supplemental cash-flow disclosure — $150.1 million in FY2025 [53]. An FY2026 figure materially above $215 million confirms the step-up; one near FY2025's level means the shield ran longer than the deferred-asset balance suggests.

Operating lease commitments in the contractual-obligations table — $777.7 million at March 2026 [54]. Further increases lock in fixed cost ahead of revenue that has not arrived.

The allowance for credit losses on the balance sheet — $16.2 million at March 2026 [55]. A step change there is the earliest visible sign that the receivable concentration has stopped being theoretical.