OpenAI Targets $100B Ad Revenue by 2030, Analysts Skeptical

OpenAI told investors it will book $100 billion in ad revenue by 2030. That requires growing ad sales more than 200% a year from a business that only started plugging in partners in February. Criteo (retargeting), StackAdapt (programmatic display), and Amazon's DSP (demand and commerce rails) are all in as pilots. The number is a fundraising slide, and everyone in the room knows it.

The actual question for operators is simpler: does a chat query surface with 600 million weekly users become a place your budgets have to go, and who owns the plumbing when it does?

What's actually been built

Not much. Nate Elliott of eMarketer lists team, technology, vendor deals, formats, and pricing as still missing. That's the whole operation. OpenAI is renting an ad business out of other companies' parts rather than building one. The Amazon pilot is the biggest named partner, which tells you how thin the roster is.

The structural problem: OpenAI's subscription business and its ad business are pulling in opposite directions. Every ad in ChatGPT is a reason a Plus subscriber asks why they're paying.

Who wins near-term

The verification players (DoubleVerify, IAS) get new contracts the moment agencies demand third-party checking on ChatGPT inventory, which they will. Criteo, StackAdapt, and Amazon's DSP collect margin as the rented rails. If OpenAI can't build its own stack, the transaction economics flow to whoever owns the plumbing. Right now, that's not OpenAI.

For operators

Nothing routes away from existing pipes yet. There are no formats, no pricing, no measurement. A buyer who tries to activate today hits a wall fast. The right move is a small, measured test through the Amazon-ChatGPT pilot with third-party verification attached, staffed by one person. Refuse to activate without conversion data. If you can't see the measurement, it's an option, not a channel.

Our call: OpenAI will not stand up its own end-to-end ad stack through the 2027 upfront season. It will keep renting demand and delivery from Amazon, Criteo, and StackAdapt rather than replacing them. Building an ad org from scratch takes years, and every move OpenAI has made says landlord. The intent-data thesis is real but unproven, and the near-term value goes to the partners it has to rent. Revisit by September 2027: if ChatGPT ad demand still routes primarily through outside DSPs heading into the upfronts, the call holds.

OpenAI is telling investors it will book $100 billion in ad revenue by 2030. That would mean growing ad sales more than 200% a year, every year, from a business that only started plugging in partners like Criteo and StackAdapt in February. Amazon just brought its DSP into ChatGPT as a pilot. The real question for ad-tech operators isn't the headline number. It's whether an AI query surface with 600 million weekly users becomes a place your budgets actually have to go, and what breaks in the current stack when they do.

How hard is this to undo? For OpenAI, easy. They can dial the ads business up or down. For operators deciding whether to build integrations and staff against ChatGPT inventory, it's a bit stickier, but a test budget is cheap to walk away from. Nobody is locked in yet.

What's actually being decided: Not "does OpenAI hit $100B" (it won't). It's whether ChatGPT's query stream becomes a new intent-targeting layer worth wiring into, and who owns the plumbing when it does.

What sets the deadline: Q4 is the loud season, but Elliott is right that it's too early to be a real test. The forcing event is whichever holdco moves a real, measured budget through ChatGPT and reports back.

---

The Market Analyst. Follow the money, not the number. $100 billion by 2030 is a slide for OpenAI's next raise, and everyone in the room knows it. The interesting part is who OpenAI picked to build with. Criteo for retargeting, StackAdapt for programmatic display, Amazon's DSP for demand and a commerce rail. None of those are OpenAI building an ad stack. They're renting one. For the public names, the read is defensive. If ChatGPT inventory ships without third-party checking, agencies balk, and that's a new contract for DoubleVerify and IAS. In plain terms: OpenAI is assembling an ad business out of other companies' parts, which tells you how far it still has to go.

The Skeptic. Two hundred percent a year, sustained for five years, from a standing start. That's not a forecast. Elliott listed what's missing: team, technology, vendor deals, ad formats, pricing. That's not a punch list, that's the whole business. And OpenAI's core model still runs on subscriptions that compete with ads for the same screen. Every ad in ChatGPT is a reason a Plus subscriber asks why they're paying. Calling the Amazon pilot the "biggest named partner" tells you the roster is thin. In plain terms: this is a number to defend a valuation, and the operational reality is a company that can't yet take an insertion order.

The Operator. Forget 2030. Can a campaign manager transact through ChatGPT next Tuesday? No. There are no formats, no pricing, no measurement. A buyer who tries to activate hits a wall in an hour. What you'll actually see is holdcos writing "test budgets" that look like commitment and are really just buying an option to be early. That's fine, that's cheap, and it commits nobody. Nothing routes away from your existing pipes yet. In plain terms: OpenAI is additive noise for now, and the smart move is a small test staffed by one person, not a roadmap line.

The Customer / End User. Two customers here, and they want opposite things. The advertiser wants intent signal, a query from someone actively asking what to buy, closed to a purchase through Amazon. That's a genuinely good funnel if it works, better than chasing people around the web with display. The ChatGPT user wants an answer, not an ad, and is already paying to avoid clutter in many cases. OpenAI has to sell to the first without alienating the second. In plain terms: the thing that makes ChatGPT ads valuable to advertisers, a trusted answer engine, is exactly the thing ads erode.

The CFO. The $100 billion is a distraction. The real cost is OpenAI's, and it's brutal. Standing up an ad sales org, verification, brand safety, and supply is years of spend before meaningful revenue. That's why they're renting Criteo and StackAdapt instead of building. For your P&L, the cost is smaller and clearer: a test budget and a couple of engineers to wire an integration that may not have measurement attached. Spend the option money, don't staff a team against it. In plain terms: this pays back for you as cheap optionality, and it won't pay back for OpenAI on anything like the timeline they're selling.

---

Where the council splits.

The Strategist read in the briefing window and the Skeptic disagree on what's actually being built. The Strategist says the moat is the query corpus, proprietary intent from hundreds of millions of users, and that's genuinely new. The Skeptic says none of that matters if OpenAI can't run an ad sales operation, and renting Criteo and StackAdapt proves it can't. Both can be right: the data is valuable and OpenAI is the wrong company to monetize it alone.

Second split: is this additive or a repricing event? The Operator says additive noise, nothing routes away yet. The Market Analyst says the sector reprices the moment intent-based targeting from an AI surface proves out, and that pressure lands on search-adjacent players first. The difference is timing, and timing is everything for whether you act now or wait.

Third: who captures the value? The intent signal lives with OpenAI, but the transaction rails belong to Amazon, Criteo, and StackAdapt. If OpenAI can't build its own stack, the margin flows to whoever owns the plumbing, and OpenAI ends up a supply source rather than an ad platform.

What it hinges on. Two things. First, whether OpenAI builds an ad org or stays a landlord renting other people's tech. Everything they've done so far says landlord. Second, whether intent from a chat query actually converts better than existing signals once you attach measurement to it. Nobody has that data yet. The council leans hard: the $100 billion is fiction, the intent-data thesis is real but unproven, and the near-term winners are the verification and DSP layers OpenAI has to rent, not OpenAI itself.

What to de-risk before committing. Run a small, measured test through the Amazon-ChatGPT pilot with third-party verification attached, and refuse to activate without measurement. If a buyer can't see conversion data, it's an option, not a channel. Staff it with one person.

---

Prediction: OpenAI will not stand up its own end-to-end ad stack by its next major funding milestone; through the 2027 upfront season it will keep renting demand and delivery from partners like Amazon's DSP, Criteo, and StackAdapt rather than replacing them.

Confidence: Medium — building an ad org from scratch takes years, and their moves say landlord, not builder.

Why: OpenAI started plugging in partners in February and has spent the months since renting the pieces of an ad business, Criteo for retargeting, StackAdapt for programmatic display, Amazon's DSP for demand, rather than building them. Nate Elliott of eMarketer lists team, technology, vendor deals, formats, and pricing as still missing, which is the entire operation, not a gap. Standing all of that up in-house is a multi-year build that competes for engineering attention with OpenAI's core model work, so the cheaper path is to stay the intent-and-inventory layer and let others handle demand and delivery. The opposite outcome, OpenAI displacing its own partners with a homegrown stack inside two years, would require it to out-execute companies that have spent a decade building exactly this, while its subscription business pulls in the other direction.

Revisit by 2027-09-30: We're right if OpenAI's ChatGPT ads still route demand and delivery primarily through outside DSPs and ad-tech partners heading into the 2027 upfronts. We're wrong if OpenAI has launched its own owned demand-side and delivery stack that materially replaces those partners.

The tail nobody is pricing: if the intent data proves out, the leverage sits with whoever owns the transaction rail, which right now is Amazon and the rented partners, not OpenAI.