The Trade Desk reported second quarter revenue of $715 million on August 6, 2026, a 3% increase that fell $37.6 million short of analyst forecasts, and guided third quarter revenue to at least $650 million, a floor roughly 12% below the same quarter a year earlier. Shares fell 24.22% in after-hours trading to $13.39, below the bottom of a 52-week range that had opened at $91.45.

The second quarter of 2026 produced the sharpest deceleration in The Trade Desk's history as a public company, and the guidance that accompanied it pointed to something the company has not previously had to describe: a quarter in which revenue is expected to be smaller than the equivalent quarter a year before.

Revenue of $715,057,000 for the three months ended June 30, 2026 compared with $694,039,000 a year earlier, an increase of $21 million, or 3%, according to the company's condensed consolidated statements of operations. For the six months ended June 30, revenue reached $1,403,914,000, up $94 million or 7% from $1,310,060,000. Both figures sit far below the trajectory the company established over a decade in public markets. On the earnings call held at 2:00 P.M. Pacific Time on August 6, Chief Executive Officer and co-founder Jeff Green opened by noting that the company will mark ten years as a public company next month, a period over which it compounded revenue at roughly a 34% annual rate.

"Our revenue growth is below our expectations and below the standard we hold ourselves to," Green said. "These numbers are not a reflection of our company or the long-term opportunity in front of us."

The market did not treat the gap between performance and stated potential as a distinction worth pricing in. The stock closed the regular session down 6.8% at $17.67, then fell a further $4.28, or 24.22%, to $13.39 after hours, leaving it roughly 29.3% below the prior day's close and beneath the $16.70 floor of its 52-week range. Market capitalisation stood at $8.91 billion.

The quarter in numbers

The distance between the earnings line and the revenue line was unusually wide. Adjusted diluted earnings per share of $0.34 exceeded the $0.18 analysts had modelled, an 88.9% surprise. Revenue of $715 million landed $37.61 million below the $752.61 million consensus, a 5% miss. For a company valued on its ability to compound share of a growing market rather than on quarterly profit conversion, the second of those two numbers carried the weight.

GAAP net income fell to $64,394,000 from $90,129,000, a 29% decline, with net income margin compressing to 9% from 13%. GAAP diluted earnings per share came in at $0.14 against $0.18. Adjusted EBITDA reached $241,279,000, down 11% from $270,755,000, with the margin narrowing to 34% from 39%. Non-GAAP net income was $157,554,000 against $203,074,000, and non-GAAP diluted earnings per share $0.34 against $0.41.

On a six-month basis the pattern held. Net income of $104,391,000 was 26% below the $140,807,000 recorded in the first half of 2025, and adjusted EBITDA of $447,345,000 was 7% below $478,630,000.

Where the margin went

Total operating expenses of $613,480,000 rose 6% year over year. Excluding stock-based compensation, they rose 12%, to $504 million. The gap between those two growth rates is the story of the quarter's margin compression: stock-based compensation fell to $109,556,000 from $128,885,000, masking cash expense growth that ran four times faster than revenue.

The largest single driver sat in platform operations, which increased $33 million, or 22%, to $184,333,000. That line moved from 22% of revenue in the second quarter of 2025 to 26% in 2026. According to the company's Form 10-Q, the increase broke down into $18 million in costs from supplier-provided components of value-added services and data, $8 million in hosting costs and $3 million in personnel costs.

The supplier component figure carries an accounting subtlety that materially affects how the revenue line reads. The Trade Desk generally reports revenue net of what it pays suppliers for advertising inventory, supplier-provided components of value-added services, and data. During the period, the company changed how certain of those components are treated, recognising some costs inside platform operations rather than as reductions to revenue. That change increases both reported revenue and reported platform operations expense without altering the underlying economics. The 10-Q lists it among the drivers of the reported revenue increase, alongside higher pricing on value-added services and more campaigns from new clients, partially offset by volume and other discounts tied to joint business plans and strategic partnerships.

Put differently: part of the 3% reported growth is a presentation effect rather than a spend effect. The company does not quantify how much.

Sales and marketing expense rose $13 million, or 8%, to $174,404,000, driven by $11 million in personnel costs and $3 million in marketing costs. The personnel increase reflected changes to incentive plan structure, merit and promotion cycles, and headcount added to support sales coverage. Technology and development rose $6 million, or 5%, to $140,742,000, with $4 million in personnel and $2 million in third-party costs. The 10-Q attributes the third-party increase specifically to greater use of AI software tools in the company's own development work.

General and administrative expense moved the other way, falling $17 million, or 13%, to $114,001,000. Almost all of that decline came from a $20 million drop in stock-based compensation, of which $19 million related to the CEO Performance Option. That grant was fully recognised by the end of the first quarter of 2026, and no expense relating to it was recorded in the second quarter, against $19 million in the comparable 2025 period. On a six-month basis the same item accounted for a $38 million reduction under the graded-vesting attribution method.

Total other income, net, fell to $11,514,000 from $16,424,000, reflecting lower interest income on a smaller invested balance at falling portfolio rates, plus foreign currency transaction losses partially offset by gains on forwards.

The tax line moved against the company despite lower pre-tax profitability. Provision for income taxes rose $6 million to $48,697,000 on income before taxes of $113,091,000, an effective rate above 43%. The driver was tax detriments associated with employee stock-based awards, where the prior-year period had recorded benefits. For the six months, the provision included detriments of $18 million against benefits of $14 million a year earlier, a $32 million swing on a comparable basis.

Cash, capital and the balance sheet

Net cash provided by operating activities was $153,594,000 in the second quarter, down 7% year over year from $165,013,000. Free cash flow, by contrast, rose 17% to $135,999,000 from $116,695,000. The divergence is explained almost entirely by capital expenditure timing: purchases of property and equipment fell 71% to $13,225,000 in the quarter, from $45,239,000 a year earlier, while capitalised software development costs rose 42% to $4,370,000.

That quarterly capex figure sits awkwardly against the six-month total. For the first half of 2026 the company spent $125,966,000 on property and equipment against $104,352,000 in the first half of 2025, meaning the first quarter carried $112,741,000 of that spend. The second quarter number is a trough rather than a run rate, which accounts for the free cash flow growth in the period.

Six-month operating cash flow reached $545,399,000 against $456,446,000. The increase was driven less by earnings than by working capital: accounts receivable fell $548 million and prepaid expenses fell $60 million, partially offset by a $428 million decrease in accounts payable. Both movements were attributed to the timing and seasonality of collections and supplier payments. Because the company bills clients gross and pays suppliers net, receivables of $3,200,824,000 and payables of $2,562,580,000 both appear large relative to net revenue.

The company ended the quarter with $1,122,979,000 in cash and cash equivalents and $362,354,000 in short-term investments, roughly $1.5 billion combined, against $658,175,000 and $644,882,000 respectively at December 31, 2025. Working capital stood at $2.0 billion, of which $92 million in cash sat with international subsidiaries. Total assets were $5,764,446,000 and total stockholders' equity $2,574,352,000, against an accumulated deficit of $719,488,000.

On April 14, 2026, the company entered into a restated loan and security agreement with a syndicate led by JPMorgan Chase Bank, expanding its revolving facility to $750 million from $450 million, with a $100 million letter of credit sublimit and a $75 million swingline sublimit, and the right under certain circumstances to increase the facility by a further $750 million. The facility matures on April 14, 2031. No debt was drawn at June 30, 2026, and availability stood at $745 million net of $5 million in outstanding letters of credit. The company was in compliance with all covenants.

Non-cancelable contractual obligations totalled $939,124,000 at quarter end, comprising $762,231,000 in operating lease commitments and $176,893,000 in other contractual commitments covering hosting services, hardware providers, data-related service providers and software as a service. Of that total, $97,937,000 falls due in the remainder of 2026.

Foreign exchange exposure is meaningful at this scale. An immediate 10% adverse move across foreign-denominated accounts would produce a currency loss of approximately $48 million, according to the company's market risk disclosures, with the principal exposures in Euro, British Pound, Australian Dollar, Canadian Dollar, Japanese Yen, Indian Rupee, South Korean Won, Swedish Krona, New Zealand Dollar, Indonesian Rupiah and Hong Kong Dollar.

A repurchase figure that does not reconcile cleanly

Two disclosures describe second quarter buyback activity differently. The press release states that the company used approximately $78 million of cash to repurchase Class A common stock during the quarter. The Form 10-Q states that during the three months ended June 30, 2026, the company repurchased and subsequently retired 3 million shares for an aggregate repurchase amount of $59 million.

The two are not the same measure. One describes cash outflow in the period, the other the aggregate value of shares retired in the period, and settlement timing separates them. Neither document reconciles the difference explicitly. For the six months, the 10-Q reports 10 million shares retired for $233 million, while the cash flow statement shows $241,331,000 paid for repurchases. Both aggregate figures include immaterial amounts relating to the 1% excise tax on net share repurchases introduced by the Inflation Reduction Act of 2022.

The board added $350 million to the authorisation in February 2026, bringing the total available to $500 million against the $150 million that remained at the end of 2025. As of June 30, 2026, $269 million remained authorised. Weighted-average diluted shares outstanding fell to 469,948,000 from 495,776,000, a 5.2% reduction that flattered per-share metrics against an already declining earnings base.

What the third quarter guidance implies

The company guided third quarter revenue to at least $650 million and adjusted EBITDA to approximately $160 million. Against third quarter 2025 revenue of $739 million, the guidance floor implies a decline of roughly 12%. The implied adjusted EBITDA margin of about 24.6% would be the lowest the company has posted in years, and compares with the 43% margin recorded in the third quarter of 2025.

The company also declined to provide a GAAP net income outlook or a reconciliation of the adjusted EBITDA guidance to net income, citing the unpredictability of stock-based compensation expense driven by share price movement. Given that the share price has fallen by roughly four fifths from its 52-week high, that variability is not a rhetorical concern.

Chief Financial Officer Nate Olmstead, who joined the company in July 2026 and was presenting his first quarterly results, addressed the guidance philosophy in response to a question from Dan Salmon of New Street Research.

"Really no change in the approach to guidance," Olmstead said. "It's very data-driven, I would say it just reflects the trends that we see in the business today. I would probably add that visibility is somewhat more limited than it has been in recent history. Given that we're not assuming any meaningful improvement in the environment during the quarter."

He added that the company does not consider the outlook either conservative or aggressive: "The goal is to be credible and I think grounded in the data. We'll call it like we see it, then we'll go execute with a lot of rigor and a lot of discipline."

Two analysts on the call, Vasily Karasyov of Cannonball Research and Youssef Squali of Truist, framed their questions on the explicit assumption that weakness implied by the third quarter guide would persist into the fourth quarter. Neither assumption was contradicted.

According to InvestingPro data cited in the transcript record of the call, 18 analysts had revised earnings estimates downward for the upcoming period ahead of the report. The same record puts trailing twelve-month revenue growth at 15.55% and trailing gross profit margin at 77.83%, figures that illustrate how abruptly the quarterly number diverged from the recent trend.

Two explanations, offered in a specific order

Green attributed the shortfall to two causes, and the order in which he presented them was itself a piece of information.

"We underperformed our own expectations for two main reasons," he said. "First, the macro conditions have made it more difficult for some of the world's largest brands to grow. Of course, this is bigger than advertising, and it's bigger than our company." The second cause followed: "Secondly, we didn't execute as well as we could have."

Placing execution second, and describing it in a single sentence against several minutes of macro analysis, set the frame for the rest of the call. Analysts returned to the split repeatedly. Squali asked directly whether management could quantify the two components. Jason Helfstein of Oppenheimer asked whether the impact could be unpacked into macro, agency-related risk, and a third category of client buying decisions.

Neither question produced a number. Green's response to Squali described what the company can control (products shipped, team composition, internal resource allocation) and what it cannot (the macro), without attaching magnitudes to either. The absence of an execution quantification, in a quarter where the company conceded execution as a cause, is the single largest gap between what management said and what investors were asking for.

The macro case

Green built the macro argument around the structure of the client base rather than around aggregate ad spend. Almost all spend on the platform, he said, comes from large Fortune 500 companies and their brands, a concentration he described as both a strength and a moat over the long term and a source of near-term exposure now.

Consumer packaged goods and automotive together account for around 25% of the business, according to Green, and both categories are described as overrepresented on the platform relative to the wider economy. The company's own investor materials support the concentration claim: in its 2025 spend-by-industry breakdown, Food and Drink represented 18% of spend, Automotive 13%, Medical Health 14%, Technology and Computing 9% and Shopping 9%.

Green cited external evidence for the pressure on those two categories. He referenced comments from Procter and Gamble describing the environment as volatile and challenging, and quoted the company as anticipating continued pressure from commodity and related costs tied to the crisis in the Middle East, with trade lane conditions dependent on whether the conflict eases and oil prices fall. He cited General Motors describing a multi-billion dollar impact from tariffs alongside plans to onshore production to reduce future tariff exposure. He referenced Ford and General Motors both highlighting the growing dependence of vehicle sales on affluent buyers, and cited Oxford Economics research placing the top 20% of households at more than 50% of new vehicle sales.

The mechanism Green described is not a simple budget cut. It is a shift in buying behaviour under pressure: "some affected advertisers have become more focused on buying cheap media rather than the best media." In some cases budgets have been temporarily reduced while brands formulate new go-to-market plans. In others, brands have moved toward what he called low-cost, low decisioning methods, naming Programmatic Guaranteed and fixed price arrangements specifically.

That characterisation matters commercially, because it describes buyers moving toward transaction types where a decisioning platform captures less value. Green's framing was blunt: buyers taking that route "will give away their decisioning in a great buyer's market to the sellers in exchange for lower cost of transactions." He called the approach "often deliberately short-sighted."

The tariff and consumer-bifurcation narrative is not unique to The Trade Desk. Pinterest faced a securities class action filed in March 2026 alleging its executives had concealed the likelihood of tariff-driven revenue declines across its retail and CPG advertiser base, a case that turns on precisely the same category exposure.

The execution admission, and what it did not cover

Green did not itemise execution failures. He instead described corrective measures: a platform usability release, two products in early distribution, an enterprise packaging motion, and a rebuilt leadership bench. Asked by Shyam Patil of Susquehanna to name the top priorities for stabilising the business, Green listed upgrading Kokai through the Zuma release, advancing the measurement product, ramping Audience Unlimited, expanding Enterprise Kokai, and giving newly hired leaders time to reach full effectiveness.

On the last of those, he was explicit about the timing cost: "in the numbers, especially in the forecast, we're trying to create room for them to get up to speed and make a meaningful contribution."

That is a material statement about the guidance. It indicates that the third quarter outlook embeds an assumption of continued sub-optimal commercial execution while a substantially new leadership team learns the business.

Where the growth actually is

Against the deceleration, management presented a set of metrics intended to isolate the pressured portion of the business from the rest.

Joint business plans were the centrepiece. The company had JBPs with 217 clients as of the second quarter, representing 38% growth year over year, and revenue under those agreements grew at a rate six times higher than overall revenue. Green described the ratio as "might be the most bullish number that we can share."

The arithmetic is worth stating plainly. Six times a 3% overall growth rate is roughly 18%. That is a real and meaningful figure, but it is also a multiple applied to a small base, and the company does not disclose what share of revenue sits under JBP structures. The 10-Q discloses the other side of the arrangement: volume and other discounts connected to joint business plans and strategic partnerships partially offset revenue growth in both the three- and six-month periods. JBP revenue grows faster in part because it is priced lower.

Other disclosed metrics:

  • The majority of the top 100 accounts grew at double-digit rates year over year.
  • Advertisers outside the top 500 grew over 50% year over year on a year-to-date basis, which Green characterised as green shoots among smaller, challenger and mid-market brands.
  • A dedicated growth team, part of which focuses on winning back lapsed customers, grew its book of business over 250% year over year and was described as the fastest-growing individual team in business development.
  • Customer retention remained above 95% during the quarter, as it has for over a decade.

Retention above 95% alongside 3% growth describes a specific condition: clients are not leaving, but incremental budget is going elsewhere. That reading is consistent with the William Blair survey of digital advertising buyers conducted during the first quarter of 2026, which found buyers reporting The Trade Desk losing share, and with Guideline data showing Amazon moving from under 10% to just under 20% of tracked programmatic spend in roughly fifteen monthswhile The Trade Desk grew at approximately the market rate.

Geography

International revenue reached approximately 17% of the total in the second quarter, up from 14% a year earlier. United States revenue fell correspondingly to 83% from 86%. The shift reflects both international growth and domestic weakness.

Olmstead reported that EMEA and APAC each delivered over 50% connected television growth year over year during the quarter. Green added that both regions grew almost 30% year to date, and that China grew over 100% year to date. He described the CTV performance in those regions as notable because both had historically been mobile-first markets.

The investor presentation frames the size of the remaining opportunity: approximately 86% of 2025 revenue came from North America, against roughly 40% of global advertising spend, citing Magna Global. The company operates in over 35 markets with more than 3,800 employees, with EMEA offices in Dubai, Hamburg, London, Madrid, Milan, Munich, Paris and Stockholm.

Channel mix

Video, which includes CTV, represented a low fifties percentage share of the business in the quarter. Mobile represented a high twenties percentage share. Display represented a low double-digit share. Audio represented around 7% and grew faster than any other channel, as it has for four consecutive quarters. Green tied the audio performance to an expanding partnership with Spotify.

CTV and audio both grew at double-digit rates in the quarter, according to Olmstead. Set against 3% total growth, that implies at least one large channel grew slowly or contracted.

Verticals

Among verticals representing at least 1% of the business, Olmstead reported strong growth in medical health, automotive and travel, and continued pressure in food and drink and home and garden as CPG brands navigate geopolitical uncertainty, consumer softness and input cost inflation.

The automotive characterisation deserves attention because it appears to cut against Green's framing. Olmstead described automotive as an area of strength overall, while adding that the business could be growing faster absent increased tariffs on the industry. Green, in prepared remarks, listed automotive alongside CPG as one of the two most overrepresented pressured categories. Both statements can hold simultaneously (a category can grow on the platform while growing below its potential) but the two executives placed the emphasis differently within the same call.

The company also benefited from political spending related to the United States midterm elections during the quarter, according to Olmstead, without quantifying the contribution. That is a non-recurring tailwind embedded in the 3% figure, and its absence in future comparable periods works against the growth rate rather than for it.

The product response

Green devoted a substantial portion of prepared remarks to three initiatives, each aimed at a different part of the value argument.

Measurement

Green identified measurement as "what might be the biggest problem in our industry" and tied it directly to the categories under pressure. "Real brand building, which is required for categories like autos and CPGs, cannot thrive while measurement standards are broken," he said. "As long as last click and last view are the standard of measurement, brands will struggle to understand what drives their growth, and the most premium parts of the open internet will always look expensive and ineffective."

That statement contains a commercial argument about the company's own position. If premium open internet inventory appears expensive under prevailing attribution models, and if The Trade Desk's fee sits on top of that inventory, then the measurement standard is a structural headwind to the company's pricing narrative before any competitor is considered.

The new measurement framework is in alpha, designed to assign value across the customer journey rather than at the final touch, and is being built in partnership with what Green described as some of the largest media companies, measurement companies and data companies. No partners were named, no timeline was given, and no beta date was disclosed.

Later in the call Green returned to the point with an unusual admission about the preceding period: "It's that last piece where I think we've missed a little bit in the last few years, and that's part of the reason why we're so excited about the products ahead."

Audience Unlimited

Audience Unlimited simplifies discovery and activation of third-party data, replacing segment-by-segment selection and pricing with AI models applied against an advertiser's own proprietary data. The pricing structure is the product. Green compared it to music streaming, describing an unlimited-access system priced as a simple subscription percentage rather than segment by segment.

The product launched in September 2025 with cost structures at 3.3% and 4.4% rates and has accumulated data partners since, including AudienceProject in February 2026 and Adsquare location data in June 2026.

Green disclosed a single anonymised result: a global advertiser reduced its cost per unique household and its data CPM by more than 25% each, compared with a prior campaign. The advertiser was not named, the campaign period was not given, and the baseline was described only as a prior campaign.

The product is moving to open beta. In the Q&A, Green noted that both Audience Unlimited and the measurement framework are "just in the hands of very few customers today."

Zuma

Later in August 2026 the company will ship Zuma, described as a significant upgrade focused on platform usability: navigation, workflow, troubleshooting and what Green called the dialogue between human and machine. It follows the May 2026 Kokai release that added AI optimisation controls, CTV pause ads, Sincera inventory signals and Deal Desk upgrades.

Usability is a recurring theme in the company's recent history rather than a new one. The migration from Solimar to Kokai encountered client resistance during 2025, and full migration took longer than the company planned.

Enterprise Kokai

Green introduced Enterprise Kokai as a motion in which large clients negotiate feature access upfront and then use the full product set, sometimes structured through JBPs. He described "massive growth rates" among companies on that arrangement without disclosing the number of such agreements or their revenue contribution.

The argument about cheap media

A through-line of the call was a defence of decisioned buying against transaction types that require less of it, and against competitors positioning on price.

"Some of our competitors, big and small, have been focused on producing products for Programmatic Guaranteed, fixed price transactions, and simple deals with approaches that do not leverage buyer decision," Green said. "Some are even wrapping these in agentic technologies. They argue that their cheap fee will compensate for their lack of objectivity and their decisions that favor publishers rather than the buyer who is footing the bill. These approaches look more like ad networks of 2006 than reflect the progress that our industry has made in the last 20 years."

Dan Salmon pressed the point directly, naming Amazon's exclusive live sports inventory and Google's buyer direct programme as concrete competitive threats. Green's answer had three parts.

First, scale. "We have by far the highest market share in the programmatic space, especially as we're looking at the open internet," he said, adding that owned-and-operated buying by walled gardens on their own inventory is unremarkable: "Amazon buys a lot of Amazon Prime and sponsored listings. Well, I would hope so. They're the only place where you can buy them."

Second, arithmetic on fees. Green argued that comparing platform rates in isolation misstates the economics, because most competitors bundle their fee with media. "You can wrongly conclude if The Trade Desk charges, let's say, 8% and our competitor charges 4%, then The Trade Desk has to be twice as good," he said. "If you look at it as, no, when you add that to the media, The Trade Desk buys an impression for $1.08, and they buy a piece of media for $1.04, where the underlying media in both cases cost $1. Did we buy a piece of media that was better for that brand? That comparison at $1.08 versus $1.04 is meaningfully different."

The illustrative 8% and 4% figures are notably below the ranges that have circulated in the trade press. Reporting has placed Amazon DSP at approximately 1 to 2% of media spend against 12 to 15% typically associated with The Trade Desk, with Amazon charging a 1% fee for open web publisher ads and no fee for programmatic guaranteed deals on its own media. The company does not disclose its take rate. If the real spread is closer to 1% against 12%, the $1.08 versus $1.04 comparison becomes $1.12 versus $1.01, and the incremental performance the platform must deliver to justify the difference rises accordingly.

Third, incentives. Green argued that a platform with near-zero cost of goods on its own inventory has a structural reason to buy that inventory: "When you have something like YouTube, where their cost of goods sold is almost zero. If they get $1, they can spend it on YouTube and keep the whole dollar. They can spend it on Disney+ and they keep $0.10." He continued: "None of these companies are in the community service business, if you will. They're not giving things away. When things come out as free, they are not really free. They're just moving the cost somewhere else."

He closed the answer with the most quotable line of the call: "if you buy media at random, you will get your ass kicked every single time."

Pricing and take rate

Justin Patterson of KeyBanc asked how pricing philosophy is changing in response to brand advertiser needs. Green's answer combined a defence of the existing model with an opening.

He described a decade-long consistency: "If you look at us as a publicly traded company and you just scrutinize the take rate more or less over the last decade, the take rate's gone up five of the 10 years, and it's gone down five of the 10 years." The reason it has stayed within a narrow band, he said, is a rule applied to every product: "we need to make certain that if our product costs 10%, it better add at least 10% more value than if it wasn't in the middle."

He then framed the founding principle: "It is not our goal to be the cheapest platform. It is our goal to be the best."

The opening came in what followed. "We've always looked at this as if we can grow faster or win more business by changing that price or changing the approach, we'll always look at it and always consider it." He identified simplification rather than reduction as the priority: "Where I do think is the biggest opportunity is for us to simplify the way that we price. I don't think that the net number has to change dramatically."

Green also noted that pricing has been broadly stable across successive platform generations, naming Next Wave, Solimar and Kokai, and across the introduction of UID2 and OpenPath.

Fee structure has been a live commercial issue for the company. Publicis advised clients to stop transacting on the platform in March 2026 after a FirmDecisions audit concluded the company had improperly applied its DSP fee to other fees and opted the agency and some clients into fee-based products without documented authorisation. The two companies issued a joint statement resolving the dispute on June 12, 2026, without disclosing whether the underlying billing question was resolved or absorbed into a commercial arrangement.

Asked by Squali about the agency relationship, Green addressed it directly: "there were some public disputes about what was essentially a negotiation, but that's behind us." He described all holding companies as being "in periods of transition right now" and said joint business plans with brands and strong agency partnership are not mutually exclusive, adding that many brand JBPs are developed in collaboration with agencies.

He also disclosed, without naming counterparties, that the company is working with agencies on white label versions of both Audience Unlimited and agentic AI products, and cautioned on timing: "It's going to take some time for those to pay off in meaningful ways."

The AI question

Matt Swanson of RBC asked what he framed as an existential question: what gives management confidence that the demand-side platform model remains relevant as AI reshapes digital advertising.

Green rejected the premise of disruption by redefining the category. "A DSP is a platform built to decide which of those impressions you buy and which you don't," he said. "I wouldn't say that the DSP model, if you will, is going to be disrupted by AI. It is AI."

He grounded the argument in scale, describing the platform as evaluating 20 million ad opportunities every second in what he called the ultimate buyer's market, with supply outpacing demand by more than ever.

The second half of his answer concerned data governance rather than decisioning. "You have to get the biggest brands in the world to trust you with their data and then reassure them that you are going to preserve their data so that their insights from buying are put to use for them and exclusively for them," Green said. "Most of the biggest platforms in the world are not doing that. They are asking the biggest brands in the world to give them their data, and then they use that for all of their clients, and especially their competitors."

On agentic technology specifically, Green called it "one of the biggest opportunities that advertising will ever see" while dismissing a subset of implementations: "Some are using agentic to just build yesterday's business model all over again, if you will. They're building ad networks out of agentic-like technologies."

That framing places the company in direct argument with a fast-moving field. The agentic buying layer has attracted substantial vendor activity through 2026, with Databricks, LiveRamp and others building competing structural answers to who controls the governed data layer through which AI agents act on first-party data.

Green also positioned AI search as a market expansion rather than a threat. "For years, search has largely been defined by a single dominant platform. That is beginning to change," he said, arguing that large language models reshaping consumer discovery would produce a more competitive landscape and new advertising surfaces, in the way connected television expanded premium inventory.

Partnerships disclosed alongside the results

The press release listed several commercial developments from the quarter and the period immediately following it.

Dentsu named The Trade Desk as the first DSP partner for a new end-to-end retail data offering from New Stream Media, according to the company. The arrangement is notable given that Dentsu and WPP quietly exited the company's OpenPath initiative in February 2026, citing transparency concerns and fees.

The company expanded its commerce media footprint through integrations with travel, hospitality and mobility brands including Booking.com, Agoda, Kayak, Priceline, Marriott, Uber and United Airlines, described as helping advertisers activate high-intent commerce and travel signals and unify activation, measurement and optimisation.

Databricks named The Trade Desk a launch partner for CustomerLake, connecting first-party data and agentic AI to media execution. CustomerLake launched on June 16, 2026 with Integral Ad Science as a partner, built on Unity Catalog governance and Lakehouse storage, with components covering ingestion, natural language querying, a real-time profile API and agentic identity resolution.

Adobe and The Trade Desk built an integration connecting paid media exposure data to first-party profiles in Adobe Real-Time CDP.

On the inventory side, Netflix joined the company's Sellers and Publishers 500+ marketplace, removing the spend minimums and deal-setup requirements that had previously gated access to its roughly 250 million monthly user base. Samsung Ads opened its home screen inventory to programmatic buyers with The Trade Desk among the first platforms granted access, an arrangement announced on June 10, 2026 alongside Google DV360 and powered by Magnite's SpringServe ad server, with global rollout beginning in the third quarter.

On the call, Green listed media partners including Disney, Netflix, NBCUniversal, Fox, Paramount, Spotify and News Corp, and infrastructure partners including Snowflake, Databricks, LiveRamp and Hightouch. He said participating retailers on the platform now represent more than 80% of United States retail sales, including a recently renewed partnership with Walmart.

Walmart has been distributing its first-party data more widely during 2026, reaching Yahoo DSP in May and Google DV360 for YouTube campaigns in June. Renewal secures continuity for The Trade Desk; exclusivity is not part of the arrangement.

A named customer result

The only fully specified campaign outcome disclosed on the call came from General Mills. The company ran a campaign for its Nature Valley brand in the United Kingdom, pairing retail data from Tesco, Sainsbury's and Ocado with in-flight Koa AI optimisations, predictive clearing and cross-device targeting. Over four months the campaign drove a 5X uplift in sales, a 92% lower CPM compared with the benchmark, and a 2X return on ad spend improvement versus running without retail data.

The 92% CPM reduction is a large figure and the benchmark against which it is measured was not defined.

Leadership rebuild and its cost

The company has replaced most of its senior team within roughly a year. The press release named recent appointments of Nate Olmstead as Chief Financial Officer, Sarah Gavin as Chief Marketing Officer and EVP, Kristi Argyilan as Chief Commercial Officer and EVP, Ron Lamprecht as Chief Business Development Officer and SVP, and Vinny Rinaldi as Vice President of Client Strategy and Growth. Green additionally named Vivek Kundra as Chief Operating Officer. The board added Penry Price and David Haddad.

Olmstead is the company's fourth chief financial officer in fourteen months. Laura Schenkein departed in August 2025 after more than a decade. Alex Kayyal joined on August 21, 2025 and left five months later, with the departure announced on January 26, 2026. Tahnil Davis served as interim CFO through the February 2026 results. Olmstead arrives from Penguin Solutions, with prior roles at Logitech and Hewlett Packard Enterprise.

Haddad's appointment on June 9, 2026 brought the board to six members against a Nasdaq cure deadline of September 21, 2026 for restored compliance with independent director requirements.

Tim Nollen of SSR asked what the new hires bring. Green highlighted three. Argyilan leads data partnerships and previously ran advertising at Uber, managing a go-to-market team he described as almost the same size as The Trade Desk's entire organisation, with earlier experience building some of the first retail media networks. Lamprecht joins in business development from a role at Amazon that Green said was broader than advertising, focused on constructing custom partnership frameworks across the company's assets. Rinaldi arrives from Hershey's, where Green said his analysis while leading media buying demonstrated that cheap reach and user-generated content did not move the needle for the brand.

Green also described recruiting hundreds of experienced general managers, vice presidents and customer-facing leaders across the commercial organisation over the past year, tying the build to marketing decisions moving higher inside client organisations.

The cost of that rebuild is embedded in the outlook. Green said the leaders are "giving them the room to get things up to speed" and that the forecast creates room for them to reach contribution. He returned to it in his final answer: "Most of them are new, we're just trying to give them enough runway to get up to speed and really make a substantial contribution."

The rebuild follows a period of sustained restructuring. The company conducted what Green called its biggest reorganisation in December 2024, and dismissed approximately 39 employees during an all-hands meeting on December 17, 2025, a year later.

Infrastructure: the data centre transition

One of the more concrete disclosures came from Olmstead on the cost side. "Over the past two years, we've transitioned critical workloads from third-party public cloud environments to owned data centers," he said. "This has strengthened our platform infrastructure, reduced our reliance on external cloud providers, and provides us with greater flexibility to support AI and machine learning workloads."

He acknowledged the near-term cost: "While this transition creates an increase in platform operations expense in 2026, it positions us to benefit from greater efficiency and operating leverage over time."

The 10-Q adds detail. Hosting cost increases were attributed to support costs from higher platform query and impression volumes running AI and machine learning workloads, investment in new data centres, and increased use of features by internal technical teams. The company also flagged an external constraint: due to high demand for hosting infrastructure components, prices have become increasingly inelastic and component costs have been rising.

Two offsetting items reduced the reported increase. The company recorded a gain on the sale of computing and networking equipment in connection with normal-course decommissioning of data centre assets, disclosed in the cash flow statement as a $13,772,000 gain against $15,513,000 in proceeds for the six-month period. Separately, an analysis completed during the second quarter changed the estimated useful lives of certain data centre computing and networking equipment from three years to four, effective April 1, 2026, on the basis of expected longer refresh cycles. For the full fiscal year from that date, the company expects the change to reduce depreciation expense in platform operations by approximately $14 million.

Both items flatter platform operations expense in the quarter without reflecting operating improvement. Reported depreciation and amortisation still rose to $30,140,000 from $26,704,000.

Tax and regulatory items

In June 2026 the State of California enacted Senate Bill 122, which extended annual limitations on research and development credits. The company disclosed that the change may limit or delay realisability of deferred tax assets related to California research and development credits, which stood at approximately $37 million for tax return purposes at June 30, 2026. No valuation allowance has been recorded against them.

Litigation overhang

The company carries a substantial and expanding litigation docket, most of it arising from the February 2025 disclosure that produced its first guidance miss after 33 consecutive quarters.

The consolidated securities class action, captioned In re The Trade Desk, Inc. Securities Litigation in the Central District of California, covers purchasers between November 15, 2023 and August 8, 2025 and alleges violations of Sections 10(b) and 20(a) of the Securities Exchange Act, plus a separate Section 20A insider trading claim against the chief executive, the then-chief financial officer and the chief strategy officer. The court denied the motion to dismiss on March 17, 2026. Defendants answered on April 30, 2026. A scheduling order entered June 29, 2026 set a trial date of May 16, 2028. Management believes the claims to be meritless and intends to defend against them.

Eight shareholder derivative complaints arising from the same allegations are pending across federal and Nevada state courts, most stayed pending the close of fact discovery in the securities action. On August 4, 2026, two days before the earnings release, the Nevada state court reassigned two of them to its Business Court.

A consolidated data privacy action in the Northern District of California, In re The Trade Desk, Inc. Data Privacy Litigation, advances wiretapping and privacy tort theories. The court denied the company's motion to dismiss on December 18, 2025 except as to declaratory relief. Discovery closes September 10, 2027.

Litigation over the 2024 reincorporation from Delaware to Nevada also continues. The Delaware Supreme Court affirmed lower court rulings on a books and records inspection on June 26, 2026. On July 27, 2026, a stockholder filed a class action complaint in the Gunderson action asserting breach of fiduciary duty claims against Jeff Green in his capacity as controlling stockholder and against the then-board. No briefing schedule is in place.

The company expects general and administrative expense to fluctuate based on litigation, regulatory and governance matters.

The market opportunity argument

Green returned repeatedly to a framing designed to place the quarter in proportion. "Today, we estimate the global advertising market approaches $1 trillion annually," he said. "Even after everything we've accomplished, we believe we participate in only about 1% of that opportunity."

The investor presentation breaks the market into open internet at approximately $280 billion, social at approximately $265 billion, search at approximately $245 billion and traditional media at approximately $200 billion, citing the Magna 2025 Global Ad Spend Forecast alongside company estimates.

A 1% participation rate against a $1 trillion market is a statement about gross spend rather than revenue. On roughly $13.4 billion of gross platform spend in 2025, the ratio holds. It also means that the metric will not move materially on any single quarter's execution, which limits what it can tell an investor about the near term.

Green closed prepared remarks with a commitment to narrower investment. "Over the rest of the year and into 2027, we're going to be more disciplined than ever about where we invest," he said. "We'll focus our resources on a small number of high-priority growth initiatives where we believe we can create the greatest long-term value for our clients and our shareholders. That means some teams will continue to grow while others will not."

The test he described: "Does it strengthen our ability to serve our customers and drive long-term growth? If the answer is yes, we'll invest aggressively. If not, we'll reduce those resources to higher impact opportunities."

Olmstead echoed the framing without attaching a target, telling Karasyov that the company would invest with conviction where returns are attractive and be "equally disciplined everywhere else," and promising an update on the long-term profitability framework and operating objectives in due course. The absence of a restated margin framework, at a quarter where the adjusted EBITDA margin fell five points and guidance implies a further ten-point compression, was itself notable.

Why this matters for buyers, publishers and the wider market

The quarter contains several developments that carry beyond one company's income statement.

The independent DSP thesis is being tested on price, not on capability. Green's core argument is that decisioned buying produces better outcomes than fixed-price transaction types, and that the fee gap between platforms is smaller than it appears once media cost is included. That argument depends on measurement that can demonstrate the incremental value, and the company acknowledged during the call that the measurement product capable of demonstrating it is in alpha. Until it ships, the value case rests on assertion in a market where buyers under budget pressure are reported to be choosing lower-cost paths. Amazon's rise from under 10% to nearly 20% of tracked programmatic spend in fifteen months happened while that gap was open.

Programmatic guaranteed is being reframed as a competitive category. Green's characterisation of Programmatic Guaranteed and fixed-price deals as low-decisioning transactions that transfer value from buyers to sellers is a direct commercial argument against a transaction type that has grown across the ecosystem. Google has been deprecating manual non-Programmatic Guaranteed deal creation in DV360, and premium publishers including Channel 4 have opened guaranteed and private marketplace inventory across five DSPs simultaneously. Buyers evaluating supply paths now face a live disagreement about whether guaranteed transactions represent efficiency or a transfer of decisioning value, with a major DSP arguing publicly for the latter.

Category concentration is a structural risk, not a cyclical one. A platform where two verticals supply roughly a quarter of spend inherits the balance sheet conditions of those verticals. Food and drink alone accounted for 18% of 2025 spend. That concentration was an asset when CPG budgets were expanding and is a liability while they are not. Advertisers and publishers assessing platform dependency have a comparable exposure in reverse: publishers whose programmatic revenue routes disproportionately through a single DSP inherit that DSP's client mix.

Retention above 95% no longer implies growth. The most useful number in the release, for anyone modelling the programmatic market, is the coexistence of decade-long retention above 95% with 3% growth. It indicates that the competitive contest has moved from platform displacement to incremental budget allocation. Clients are running multiple DSPs and deciding where the next dollar goes campaign by campaign, which makes share shifts faster and less visible than churn would be.

International is where the growth rate lives. EMEA and APAC growing near 30% year to date, China above 100%, and over 50% CTV growth in both regions, against a business still 83% United States, describes where the addressable expansion sits. It also describes a company whose growth engine is currently in its smallest segment. Sustained mix shift toward international would change both the growth profile and the currency exposure, which the company already models at approximately $48 million of loss on a 10% adverse move.

Data governance is becoming the differentiation argument. Green's answer on AI reduced the competitive question to whether a platform uses one advertiser's data to serve that advertiser's competitors. That is an argument about contract terms and technical architecture rather than about model quality, and it is testable. As agentic systems move decision-making upstream into the data layer, through products like CustomerLake and comparable offerings, the question of who governs first-party data at the point of activation becomes the substantive one for large advertisers.

What was not disclosed

Several items material to assessing the quarter were not provided. The company did not quantify the split between macro pressure and execution. It did not disclose gross platform spend for the quarter. It did not quantify the political spending contribution from the midterm elections. It did not size the accounting change that shifted supplier component costs from revenue reductions into platform operations, despite listing it as a driver of reported revenue growth. It did not name the measurement framework's partners or give a timeline. It did not restate its long-term profitability framework. It did not disclose take rate, consistent with prior practice.

Green's final remark on the call addressed the day rather than the numbers. "While days like today hurt, they're going to make the comeback story even more exciting and more impressive."

The third quarter guidance places the first measurable test of that proposition in early November 2026.

Timeline

Summary

Who: The Trade Desk, Inc. (NASDAQ: TTD), the Ventura, California-based demand-side platform, led by Chief Executive Officer and co-founder Jeff Green and Chief Financial Officer Nate Olmstead, who joined in July 2026 and presented his first quarterly results. Analysts questioning management on the call represented Susquehanna, RBC, Cannonball Research, KeyBanc, Truist, SSR, New Street Research and Oppenheimer. Chris Toth, Head of Investor Relations, hosted.

What: Second quarter 2026 revenue of $715,057,000, up 3% year over year and $37.61 million below the $752.61 million analyst consensus. GAAP net income of $64,394,000, down 29%, at a 9% margin. Adjusted EBITDA of $241,279,000 at a 34% margin, down from 39%. GAAP diluted earnings per share of $0.14; non-GAAP diluted earnings per share of $0.34 against a $0.18 forecast. Third quarter guidance of at least $650 million in revenue and approximately $160 million in adjusted EBITDA, implying a roughly 12% year-over-year revenue decline. Shares fell 6.8% in the regular session to $17.67 and a further 24.22% after hours to $13.39.

When: Results for the quarter ended June 30, 2026 were released on August 6, 2026, with the conference call held the same day at 2:00 P.M. Pacific Time. The Form 10-Q and Form 8-K were filed on the same date.

Where: The company operates from Ventura, California, with offices across North America, Europe and Asia Pacific in more than 35 markets, and more than 3,800 employees. United States revenue represented approximately 83% of the quarterly total; international revenue represented approximately 17%.

Why: Management attributed the shortfall to macroeconomic pressure on large advertisers, particularly in consumer packaged goods and automotive categories that together supply around 25% of platform spend, and to its own execution. Green described some pressured advertisers shifting toward cheaper media and lower-decisioning transaction types including Programmatic Guaranteed. The company presented joint business plans with 217 clients growing 38% year over year, revenue under those plans growing six times faster than overall revenue, retention above 95%, EMEA and APAC growth near 30% year to date and China above 100%, alongside product initiatives covering a measurement framework in alpha, the Audience Unlimited data product moving to open beta, and the Zuma usability release scheduled for later in August 2026. Guidance assumes no meaningful improvement in the macro environment during the third quarter and embeds time for recently appointed leaders to reach full contribution.