OpenAI at Cannes: ‘We Are an Advertising Business’

OpenAI used its first-ever Cannes Lions appearance, the week of June 22, 2026, to stop hedging about advertising. Chief Revenue Officer Denise Dresser told assembled brands and agencies the company is “clearly in the advertising business now” — the first time OpenAI has owned that label publicly at a major industry event. For media planners, the slogan is less important than what sits behind it: a company barely four months into selling ads, projecting $2.5 billion in revenue this year, asking to be treated as a gatekeeper.

This is not another ChatGPT-ads how-to. The mechanics of buying a chat card are now well documented elsewhere. What changed at Cannes is the posture. A company whose CEO called ads “a last resort” less than two years ago is now pitching a $100 billion ad business to the same room that built the Google-Meta duopoly. The interesting question for any media team is not how to run a ChatGPT campaign — it is whether a third gatekeeper is actually emerging, and what it means that this one arrived without the measurement infrastructure the other two spent a decade building.

This analysis walks through what OpenAI actually declared, the 21-month reversal that got it here, the third-gatekeeper math against Meta and Google, an honest read of the $100 billion projection, and the infrastructure gap that defines the real planning decision. Every figure is sourced and dated; where a number is vendor-stated, reported-but-unaudited, or from a third-party survey, we say so plainly rather than launder it through a confident sentence.

The headline line came from Denise Dresser, OpenAI's Chief Revenue Officer, at the company's first-ever Cannes Lions press briefing on June 22, 2026: OpenAI is “clearly in the advertising business now.” She paired it with a framing meant to distinguish ChatGPT from the platforms it is being measured against — a shift, in her telling, “from the attention economy to the intelligence economy,” where what consumers want is usefulness rather than capture.

Head of global ads solutions David Dugan, who joined OpenAI from Meta earlier this year after more than twelve years there as a VP of global clients and agencies, made the commercial case. His pitch rests on intent: people arrive at ChatGPT with a problem to solve and hand over context about it, which he frames as a near-ideal setting for a targeted ad. The product reflecting that pitch is still narrow. The only current ad format is the “chat card” — a sponsored card with a title, short description, image, and link, shown below the model's organic answer — and it appears only on the free and lower-cost ChatGPT Go tiers, with the paid Plus and Pro tiers kept permanently ad-free.

The Cannes declaration lands so hard because of what preceded it. In May 2024, Sam Altman described advertising as “a last resort” for OpenAI's business model and called the combination of ads and AI “uniquely unsettling.” ChatGPT ads launched in the US on February 9, 2026. The Cannes pitch followed in June. Measured from that 2024 fireside chat, the “last resort” became a declared core business in roughly 21 months.

What changed in between is not mysterious, and it is worth doing the arithmetic on rather than gesturing at a vibe shift. OpenAI is reported to have lost just under $8 billion in 2025 — a figure drawn from reported and leaked financials, not an audited disclosure; the company's confidential S-1, filed around June 8–9, 2026, will be the first time these numbers face public scrutiny. Subscription revenue alone does not close a loss that size at the growth rate investors are pricing in. Advertising is the obvious second engine, and the timing — ads launched February, IPO paperwork filed June, advertising pitched at Cannes the same month — is not a coincidence.

For more than a decade, digital media planning has operated inside a duopoly: Google for intent, Meta for attention. 2026 is the year that framing started to crack from two directions at once. eMarketer projects Meta will overtake Google in global digital ad revenue for the first time ever this year — roughly $243 billion to Google's $239 billion — with Google, Meta, and Amazon together still commanding around 62% of global digital ad spend. Into that reshuffling walks OpenAI, asking planners to add a third name to the shortlist.

The case for taking that seriously rests on reach and intent. ChatGPT is approaching a billion weekly users — Altman confirmed 800 million weekly actives in October 2025, and 2026 reporting puts the figure near 900 million and climbing, so “approaching a billion” is the honest way to frame a moving number. OpenAI says roughly 20% of ChatGPT queries carry direct commercial intent (its own figure, not independently verified). A billion-user surface where one in five queries is shopping-adjacent is, structurally, the kind of intent inventory that built Google. That is the argument for a third gatekeeper, and it is a real one.

The skeptical read is equally grounded, and a planner should hold both. A third gatekeeper is not a third gatekeeper until advertisers can plan, buy, and measure against it the way they can against the other two. OpenAI has the audience and, plausibly, the intent. What it does not yet have is the decade of attribution plumbing, audience verification, and brand-safety controls that turn an intent surface into a media channel a CMO will move real budget onto. The reach is the easy part; the rails are the question. For the wider competitive picture of how these surfaces are reshaping search and discovery, our overview of how AI search platforms are reshaping digital advertising maps the field.

The number OpenAI is putting in front of investors is the spine of the whole story. According to projections reported by Axios and corroborated by eMarketer, OpenAI is forecasting roughly $2.5 billion in ad revenue in 2026, scaling to $11 billion in 2027, $25 billion in 2028, and approximately $100 billion by 2030. These are vendor-stated projections shared with investors, not audited results — and that caveat does more work the further out the curve runs. The Information has reported the 2030 figure as $102 billion; OpenAI's own public framing uses the round $100 billion, which is what we cite.

The reality check is the historical comparison, and it is unflattering to the projection without being dismissive of the ambition. AdExchanger noted at Cannes that Meta — founded in 2004 — took until around 2021 to build a roughly $100 billion ad business, about 17 years. OpenAI began selling ads in February 2026 and is projecting the same milestone by 2030: about four years. The chart below puts the two side by side. It is not that OpenAI's target is impossible; it is that it implies compounding at a pace no advertising business has matched, on infrastructure that does not yet exist.

Strip away the Cannes theater and the planning question is concrete: can you run ChatGPT ads the way you run Google and Meta — plan against verified audiences, attribute conversions across channels, and apply the brand-safety controls a regulated advertiser needs? Today, not fully. ChatGPT ads still lack standardized viewability metrics, independent audience verification, and mature cross-channel attribution. OpenAI is closing that gap by borrowing, not building: an integration with Criteo (which says over 2,000 brands now advertise on ChatGPT through its platform), a LiveRamp partnership announced around June 10 to enable measurement beyond browser-based tracking, and a CPM-to-CPC shift that hands advertisers a more familiar performance lever.

Those are sensible moves, but they are workarounds for infrastructure Google and Meta spent a decade compounding. The table below is the part of this analysis a planner should bookmark: a like-for-like read of where ChatGPT ads sit against the incumbents on the dimensions that actually decide a media plan — not a list of features, but a map of the gaps. The honest takeaway is that the inventory is novel and the rails are borrowed.

Read down the ChatGPT column and the pattern is consistent: the buying side has caught up fast (self-serve, no minimum, CPC), but the trust side — attribution, verification, brand safety — is still “emerging” or “via partners.” That is the single most important fact for a planner deciding budget. You can buy ChatGPT ads as easily as Google ads today. You cannot yet measure them with the same confidence. The Getty Images display agreement signed at Cannes adds licensed imagery to ChatGPT's surfaces, and the Japan/South Korea market activation announced that week extends reach — but neither closes the measurement gap, which is the one that decides whether spend scales past a test budget.

OpenAI's declaration is one move in a live experiment, and the cleanest way to read it is against the other two. Three leading AI labs have placed three opposite bets on the same question — can an ad-supported model coexist with user trust? OpenAI says yes and is building the business. Anthropic ran a Super Bowl LX ad in February 2026 with the tagline “Ads are coming to AI. But not to Claude,” making ad-free a brand position; reporting credited the spot with an 11% jump in Anthropic's daily active users post-game. Perplexity went furthest, abandoning advertising entirely in February 2026 to go subscription-only, and reported its ARR jumping 50% in the following month to around $450 million.

This is a testable hypothesis still running, not a settled debate, and it would be dishonest to call a winner today. There is third-party signal that the trust concern is real: an Ipsos survey in Q1 2026 reportedly found 63% of US adults say ads in AI search results make them trust the results less, regardless of labeling — though that is a secondary-sourced figure we have not independently verified, so treat it as directional color rather than proof. The genuinely interesting part is that the three labs disagree so completely about what users will tolerate, which means at least two of them are wrong, and the market is about to tell us which.

For a media team, the divergence is not trivia — it is a planning input. If you concentrate spend on ChatGPT and the trust thesis breaks against OpenAI, you have anchored budget to a deteriorating surface. If you avoid it on principle and the usefulness thesis holds, you cede a billion-user intent channel to faster movers. The defensible posture is neither faith nor abstention but a measured test with real attribution attached. Anthropic's contrarian stance is worth understanding in its own right; we cover it in our look at Anthropic's ad-free pledge, and the parallel content play in our piece on OpenAI's broader media strategy.

The strategic call for a media team is not whether OpenAI's $100 billion projection is right — it almost certainly will not land exactly there — but whether ChatGPT is a surface your customers now use with commercial intent. For a growing share of audiences, it is. That argues for a deliberate test, sized as a learning budget rather than a bet on the projection, and structured so the measurement gap is a known limitation rather than a surprise on the post-mortem.

In practice that means three things. Treat ChatGPT ads as incremental, not as a Google or Meta replacement, until attribution matures. Lean on the partner measurement stack — Criteo, LiveRamp, and your own first-party conversion data — to bridge the gap OpenAI has not closed natively. And hold OpenAI's own trust principles against your brand-safety standards: the company pledged at Cannes that ads stay visually separate from organic answers, that conversational data is never shared with advertisers, and that users keep control — useful commitments, but ones a careful advertiser should verify rather than assume.

Standing up that kind of test — channel selection, attribution that survives a measurement gap, and a routing model that does not over- index on any single platform's projections — is exactly the work our paid media management and AI and digital transformation engagements are built around. The goal is not to chase the loudest Cannes headline; it is to put a measurable, reversible toe in a genuinely new intent channel without mistaking a vendor projection for a media plan.