Four ad tech companies reported second-quarter 2026 results this week, and the numbers split cleanly into two groups that have nothing to do with revenue size. One group owns the pipe its ads travel through. The other group rents space in someone else’s. Owning the pipe is turning out to be the difference between raising guidance and explaining, on the record, why growth wasn’t enough.
The Same Earnings Story, Different Names
The Trade Desk, the largest independent demand-side platform, posted $715 million in second-quarter revenue, up just 3% year over year. Criteo, the commerce-media veteran, posted $428 million, down 11%. Both results came with unusually blunt language from the top. “This quarter did not meet the standard we set for ourselves,” said Jeff Green, CEO of The Trade Desk, in the company’s earnings release. Criteo CEO Michael Komasinski used a word executives rarely choose voluntarily: “While our second quarter top line performance was disappointing, our long-term strategy remains unchanged.”
Across the same week, two companies with a very different relationship to their own supply chain posted the opposite story. AppLovin, the mobile advertising platform, reported $1.92 billion in revenue, up 53% year over year, with adjusted EBITDA margins expanding to 84% and guidance for Q3 raised to 46-48% growth. Magnite, the sell-side platform, posted $192.8 million in revenue, up 11%, and raised its full-year guidance on every metric it tracks. “We significantly beat consensus expectations on both the top and bottom line in the second quarter.” That’s Magnite CEO Michael Barrett, and it is not a sentence Green or Komasinski got to say this quarter.
Four companies, one earnings week, two entirely different conversations with investors. The variable that lines up with which conversation a company had is not growth rate alone. It is whether the company owns the infrastructure its ads actually run through, or whether it is a layer sitting on top of someone else’s.
What Owning the Pipe Actually Buys You
AppLovin’s advertising business runs almost entirely inside its own network of mobile apps and games. When AppLovin sells an ad, it is selling inventory it controls end to end: the app, the user, the auction, and increasingly the measurement layer that tells the advertiser whether the ad worked. That closed loop is also what let AppLovin push net revenue per installation up 58% even as raw installation volume fell 2%, according to its own earnings release. The company is not dependent on a publisher deciding to keep working with it, or an advertiser trusting a third-party measurement vendor to referee the outcome. It owns every link in the chain it is selling.
Magnite’s version of the same idea is narrower but just as deliberate. The company has spent the past two years building out SpringServe, its own supply-side ad server for connected TV, rather than continuing to plug into ad servers other companies control. That bet paid off directly in the quarter: Magnite’s CTV contribution reached $97.1 million, up 36% year over year, and management credited the gain specifically to owning the decisioning layer rather than routing around it. A sell-side platform that controls its own ad server does not have to ask a publisher’s other vendor for permission to run an auction. It just runs it.
Why Ownership Changes the Auction Math
The mechanical reason this keeps happening comes down to how a programmatic auction actually works. Every ad impression that changes hands through the open web passes through a chain of intermediaries: a supply-side platform on the publisher’s end, an exchange in the middle, a demand-side platform on the advertiser’s end, and often a separate measurement vendor to verify what actually happened. Each link in that chain takes a cut, and each link is also a point where a company that does not own the infrastructure has to trust, or negotiate with, whoever does.
When AppLovin sells an impression inside its own app network, there is no separate supply-side platform to pay and no third party to convince that the impression was real; the company controls the inventory, the auction logic and the outcome data in one system. When Magnite routes a CTV impression through SpringServe instead of a publisher’s existing ad server, it collapses two links in that chain into one it owns, which is exactly why management pointed to SpringServe by name when explaining the quarter’s CTV outperformance. Owning more of the chain does not just capture a bigger share of the same dollar. It removes the negotiating leverage a partner further up or down the chain would otherwise have over the price and the terms.
What Renting the Layer Costs You
The Trade Desk does not own inventory. It is, by design, a neutral buying layer that sits between advertisers and every open-web publisher’s ad server, which is precisely what made it valuable when the pitch was independence from walled gardens. That same position now means The Trade Desk’s growth is hostage to how much money advertisers choose to route through the open web at all, a decision The Trade Desk does not control. Green’s own explanation leaned on the same idea from the other direction: “Marketers are navigating a complex environment,” he said, before describing a response built entirely around attaching The Trade Desk to other companies’ data infrastructure, including a launch partnership with Databricks and a new integration into Adobe’s Real-Time CDP.
Criteo’s version of the same move is a bet on OpenAI. The company became OpenAI’s first advertising technology partner in March 2026 and now has more than 2,000 brands running ads inside ChatGPT, alongside retail-media partnerships with Loblaw Advance, Monoprix, Druni and Olive Young. None of that changes what Criteo fundamentally is: a technology layer whose growth depends on being chosen by platforms and retailers that could, in principle, build the same capability themselves or choose a different partner next cycle. The company is not renting space in a stranger’s building anymore. It picked ChatGPT as its new landlord.
The forward guidance each company gave makes the split even sharper. The Trade Desk guided to at least $650 million in third-quarter revenue and roughly $160 million in adjusted EBITDA, both flat to slightly up from Q2. Criteo guided to $237 million to $241 million in Contribution ex-TAC, a projected decline of 14% to 15% year over year at constant currency. AppLovin, by contrast, guided to $2.055 billion to $2.085 billion in third-quarter revenue, growth of 46% to 48%, with adjusted EBITDA growth of 48% to 50%. Magnite raised every full-year target it publishes: Contribution ex-TAC growth to 13% to 14%, adjusted EBITDA growth to more than 20%, adjusted EBITDA margin to at least 37%, and free-cash-flow growth to the high 40% range. Guidance is a company telling investors what it believes about its own near-term control over the business. Two of these four companies believe that control is improving. Two believe it is still being negotiated.
The Risk on the Other Side
Owning the pipe is not a free strategy. A company that controls its own supply, auction and measurement in one closed system is also the only party responsible when something in that system goes wrong, whether that is a fraud allegation, a regulatory inquiry into self-dealing, or a customer discovering the referee and the player are the same company. Neutral, unbundled layers like The Trade Desk exist in the first place because advertisers spent years asking for exactly that separation of interests. The current earnings gap rewards vertical ownership. It does not settle the underlying argument for why the open, neutral layer was built in the first place, and a category that swings too far toward closed, owned supply tends to invite the regulatory and trust problems that created the demand for independent ad tech to begin with.
The Playbook Everyone Reaches for Anyway
What is notable is not that Green and Komasinski responded to a weak quarter. It is that both of them responded with the identical move: attach the company more tightly to somebody else’s proprietary infrastructure. A CDP integration. A retail-media partnership. A first-mover deal with the AI platform advertisers are starting to ask about. None of that is ownership. It is an attempt to make renting feel more permanent, and it is the same instinct that showed up earlier this month when mobile ad network Liftoff Mobile pushed its own ad business beyond gaming specifically to reduce its dependence on any single vertical’s ad budget. Diversifying who you depend on is a real strategy. It is just a different strategy from owning what you depend on, and investors priced the difference this earnings season with unusual clarity.
This is not a new pattern for ad tech. It shows up almost every earnings cycle in some form: independent, neutral-layer businesses growing revenue while getting punished for structural exposure, while businesses that control proprietary supply or data get rewarded even at a smaller scale. What is different this quarter is how explicitly the neutral-layer companies named the fix. Two years ago, The Trade Desk’s answer to competitive pressure was better bidding algorithms. This quarter, its answer was a partnership with a database company.
What It Means for the Marketing Leader
For a CMO or head of marketing operations evaluating vendors right now, the earnings language matters more than it usually does. A DSP or SSP that talks about its own proprietary supply or data asset is telling you where its negotiating leverage sits, and by extension where the pricing power in your contract with it actually lives. A vendor that talks about its integration partners is telling you it is managing dependency risk on your behalf, which is not the same as removing that risk from your stack. Both can be legitimate vendors to work with. They are not equally exposed to the next quarter’s version of this same story, and a media plan built assuming otherwise is making a bet the vendor itself is actively hedging against.
The practical tell is simple: ask any programmatic or CDP vendor what happens to their product if their single largest data or infrastructure partner walked away in twelve months. A company like Magnite or AppLovin has an answer rooted in what it owns. A company leaning on a Databricks integration or an OpenAI partnership has an answer rooted in a relationship it does not fully control. Neither answer should disqualify a vendor on its own. But only one of those answers is the kind that shows up as a raised guidance number instead of an earnings call explaining a miss.
Source: The Trade Desk, Q2 2026 Form 8-K Exhibit 99.1 (SEC EDGAR)
