The Trade Desk told investors on Thursday to expect third-quarter revenue of “at least $650 million,” a floor that sits roughly 12% below what the company booked in the same quarter a year ago, and the stock fell 22.41% after the close to $13.71. A company whose entire investment case rests on taking share from an expanding digital ad market just guided to shrinking inside one.
The Guide, Not the Quarter, Did the Damage
The quarter itself was bad but survivable. The Trade Desk’s second-quarter release put revenue at $715 million, up 3% year over year, about $37.6 million short of the $751.39 million analysts carried. Net income came in at $64 million, adjusted EBITDA at $241 million, GAAP diluted earnings at $0.14 a share and non-GAAP at $0.34 against the $0.40 the street had modeled.
Then came the forward number. Wall Street was penciling in something near $805 million for the third quarter. Management said at least $650 million and pointed adjusted EBITDA at roughly $160 million, down from $241 million in the quarter it had just reported. As PPC Land calculated in its results coverage, that floor implies a 12% year-over-year contraction. Nothing in the release explains a $155 million gap between what analysts expected and what the company is willing to promise.
Retention Above 95% Is the Line That Should Worry People
Buried in the same release is the number that reframes everything: customer retention stayed above 95%, where it has sat for years.
That is normally a bragging stat. Here it is the diagnosis. If advertisers were walking out the door, The Trade Desk would have a competitive problem with a known playbook: win them back, cut price, ship features, hire sales. Retention above 95% alongside revenue heading backwards means something harder. The same clients are still there and they are routing less money through the platform.
Demand is not churning. It is migrating. Amazon has spent two years pushing its demand-side platform at advertisers with first-party retail purchase data attached and effectively no take rate to defend, because the ads are a lever on a commerce business rather than the business itself. Against that, an independent buy-side toll booth is arguing for its own margin. Management also flagged softness among large consumer-packaged-goods and auto advertisers, which is true and which is also the more comfortable half of the explanation.
Three Chairs, Three New Occupants
The executive turnover reads differently once you accept the migration thesis.
Nate Olmstead became chief financial officer effective July 9, arriving from Penguin Solutions after a stint as CFO of Logitech and 16 years across Hewlett Packard and Hewlett Packard Enterprise, according to the company’s own announcement. Sarah Gavin started as chief marketing officer and executive vice president on June 15, coming from Zendesk, where she had been chief communications officer and interim CMO. Commercial leadership turned over too, on top of a longer run of senior departures and board seats changing hands.
Refreshing three of your most visible seats inside eight weeks is a statement about diagnosis. It says the board concluded the problem is execution: the story was told badly, the sales motion was run badly, the numbers were framed badly. Chief executive Jeff Green told Adweek the 3% growth rate was “not a reflection” of the company, which is the same read in friendlier language. In the release he was more measured, saying the quarter “did not meet the standard we set for ourselves, but it has reinforced our belief that we are focused on the right opportunities for the future.”
If the diagnosis is right, the new bench fixes it over four quarters. If the demand is genuinely moving to platforms that own the transaction, then no CFO and no CMO changes the physics, and the company will have spent a year of credibility on the wrong hypothesis.
What the Buyback Was Actually Buying
The company repurchased $78 million of Class A stock during the quarter. It did that while the shares were completing a first half in which, as Yahoo Finance documented, they lost 52% of their value.
Buying your own equity into a decline is a specific claim: the multiple is wrong, the business is fine. Thursday’s guidance is the board’s own analysis contradicting that claim in public. Every dollar spent supporting the stock at $30 or $40 is a dollar not held against a year in which revenue may print negative and adjusted EBITDA compresses by a third. That is the awkward part of the follow-the-money read, and it is why the after-hours move was violent rather than orderly.
There is a version of this where the sell-off is the opportunity. A platform with 95% retention, real adjusted EBITDA and a functioning connected-TV franchise trading at a mid-teens share price is not obviously a zero. The bull case now requires believing that “at least $650 million” is a new management team setting a bar low enough to clear comfortably in its first full quarter, which is exactly what an incoming CFO is incentivized to do.
The Uncomfortable Comparison
Look across the same reporting week and the shape of the AI trade gets clearer. Companies that pay for the infrastructure got punished. Companies that collect a fee on what flows through somebody else’s infrastructure got rewarded, which is how Shopify turned a 34% revenue quarter into a 26% share move two days earlier.
The Trade Desk was supposed to be in the second category. It builds no models, owns no inventory and takes a cut of media that flows through its pipes. The problem is that the pipes are being rerouted. Amazon, Google and the retail media networks are not competing for a slice of the open internet’s ad budget so much as absorbing the budget into places where the buying tool comes free with the audience.
Watch two things into the third-quarter print. Whether connected-TV revenue holds its growth rate while the total shrinks, which would confirm that the decline is concentrated in open-web display. And whether Olmstead resets guidance methodology outright, because a new CFO who rebases the whole forecast is telling you the old numbers were never the issue.