ChatGPT’s $102bn Ad Bet Needs Google to Stand Still

On April 14, Barclays published a research note projecting that ChatGPT advertising revenue will reach $2.4 billion in 2026 and $102 billion by 2030. Most coverage that followed compared those figures to Alphabet’s combined 2025 advertising revenue of $294.69 billion. The more load-bearing comparison is narrower: Google Search and other revenues came to $224.5 billion last year, per Alphabet’s 2025 annual report filed with the SEC, and it is the one line a conversational ad surface can plausibly contest.

The Barclays model, drawn on by EMARKETER in a published data table, reaches $102 billion through a specific price signal. ChatGPT’s revenue per query sits at $0.002 in 2026, representing 6% of Google Search’s current equivalent. By 2030, Barclays projects it at $0.041, a 2,050% increase across four years, at which point ChatGPT would represent 96% of where Google Search stands today.

The Barclays Model’s Four Moving Parts

The note, titled “Looking Steady Here, the AI Ads Are Coming,” builds its projection from four variables: advertiser-eligible user count, query volume per user, revenue per query, and annual revenue per user. Each carries distinct uncertainty over a four-year horizon, and the four interact in ways that compound.

User growth is among the model’s more defensible inputs. The model projects 700 million advertisable ChatGPT users in 2026, growing to 1.7 billion by 2030. Query volume follows the same trajectory, from 1,002 billion annually to 2,485 billion by 2030, with the sharpest single-year increase, at 34%, falling in 2027. A constant 4.0 queries per user per day is assumed throughout the projection period.

• $0.002 per query: ChatGPT’s revenue per query (RPQ) at launch, representing 6% of Google Search’s 2025 level
• $0.041 per query: projected RPQ by 2030, representing 96% of Google Search’s current level
• $3.50: annual revenue per user (ARPU) in 2026; the model projects $60 by 2030, a 1,614% increase
• 8% of OpenAI’s total revenue from advertising in 2026; the model projects 36% by 2030

The jump from $3.50 to $12 in ARPU between 2026 and 2027 (a 243% increase in a single year) assumes OpenAI will have built a competitive, auction-based advertising platform from a limited pilot in roughly 12 months. That auction depth matters: the $60 ARPU the model projects for 2030 requires the kind of advertiser competition that takes mature digital ad markets years to cultivate. Google’s search auction has 25 years of bidding history across hundreds of thousands of advertisers in every major commercial category; OpenAI opened its self-serve platform to all U.S. businesses only in May 2026.

Martin Grosse, Head of SEA and Programmatic Display at Suchmeisterei GmbH, put the skepticism plainly in a comment on the EMARKETER data table: “Revenue per query from .0002 to .041? Do they plan for a wall of ads?”

Google’s Three Revenue Lines

Alphabet’s $294.69 billion in 2025 advertising revenue comprises three businesses with distinct economic structures and separate competitive surfaces.

Google Search generates revenue because users type commercial queries and receive paid results alongside organic links, a surface built on 25 years of purchase-intent signal refinement. YouTube’s $40.4 billion in 2025 ad revenue comes from pre-roll and mid-roll formats against a content library created by third-party users; OpenAI has neither the video library nor the rights infrastructure to enter that market. Google Network’s $29.8 billion flows through a programmatic publisher-distribution system that took over a decade to build, and it is already contracting, down from $30.4 billion in 2024 as AI-mediated sessions reduce the traffic that reaches external publisher pages.

The 13.4% growth in Google Search revenues through 2025 happened alongside the introduction of AI-mediated features. Google’s AI Overviews, embedded directly into search results pages while preserving the advertising auction around them, kept commercial intent queries inside Google’s own monetization infrastructure as the AI transition accelerated.

ChatGPT’s conversational ad surface competes for commercial intent queries, the same queries that drive the search line. OpenAI’s 2030 target of $102 billion sits below that one line. YouTube’s economics and Google Network’s publisher infrastructure are outside the contested surface entirely.

The Assumption Google Stands Still

The Barclays model sets its convergence benchmark at Google Search’s 2025 revenue per query: roughly $0.043, derived from $224.5 billion in search revenue against the 5.25 trillion annual queries the note uses as its denominator. ChatGPT’s projected 2030 RPQ of $0.041 would, on that arithmetic, still sit below where Google stands today. Four years of compounded growth in Google’s own monetization rate does not appear in the model’s Google-side assumptions.

Google’s 2025 data makes the static-baseline assumption hard to defend. Search revenues increased $26.4 billion year-over-year, a 13.4% gain. In Alphabet’s Q4 2025 earnings release, Sundar Pichai, chief executive of Alphabet and Google, described search as seeing “more usage than ever before, with AI continuing to drive an expansionary moment,” framing generative AI as a force extending search monetization. Alphabet’s 2026 capital expenditure budget runs to $175 billion to $185 billion, and the company subsequently raised $85 billion in additional capital to fund the buildout behind it.

But 96% of RPQ of Google is only explicable if Google does nothing to increase its own RPQ, and that, my friend, is desperately unlikely.

Steven Plimmer, a digital marketing professional, wrote that in a LinkedIn comment responding to the EMARKETER table in April 2026.

The Gemini App, which Pichai confirmed had grown to over 750 million monthly active users by the Q4 2025 earnings, is Alphabet’s competing conversational surface. Google Network revenues are already down 3% in Q4 2025 and 4% in Q1 2026, a structural decline that conversational AI interfaces of all types are accelerating.

Four Months to a Self-Serve Platform

OpenAI moved from no advertising business to a self-serve infrastructure in roughly 16 weeks. The sequence is specific.

1. February 9, 2026: Pilot launches in the United States at a cost per thousand impressions (CPM) of $60, minimum advertiser commitment of $200,000 to $250,000
2. March 2, 2026: Criteo announced as the first formal ad tech partner
3. March 26, 2026: Annualized revenue crosses $100 million within six weeks; more than 600 advertisers in the program; expansion to Canada, Australia, and New Zealand announced
4. April 13, 2026: Self-serve Ads Manager confirmed; minimum spend lowered to $50,000
5. April 17, 2026: CPMs reported as low as $25, down from $60 at launch nine weeks earlier
6. May 5, 2026: Self-serve Ads Manager open to all U.S. businesses with no minimum spend; cost-per-click bidding and a Conversions API introduced simultaneously
7. May 14, 2026: Custom audience targeting added

In four months, OpenAI shipped a self-serve interface, programmatic partnerships, a Conversions API, CPC bidding, and custom audiences, each a piece of infrastructure Meta built into its advertising platform over several years. The CPM compression from $60 to $25 in nine weeks reflects deliberate threshold-cutting to widen advertiser participation, deepening the auction pool that the RPQ climb will eventually require.

What the Criteo Data Shows

Criteo, the Paris-based ad tech company that joined as OpenAI’s first formal ad tech partner in March, reported that ChatGPT-referred users converted at 1.5 times the rate of users arriving from other referral channels.

The mechanism is reasonably intuitive. A user who receives a specific product recommendation inside a conversation has passed through an intent-filtering step that a standard search click lacks; the recommendation already reflects the user’s stated need. A 1.5x conversion premium, if it holds at scale, would support higher CPMs and could enable a faster-than-expected RPQ ramp.

The data comes from a controlled pilot with fewer than 1,000 advertisers, in a period when competition for any given placement category was thin. ChatGPT’s targeting operates through context hints, phrases advertisers supply to describe the conversations where their products are relevant. There is no behavioral profile, no retargeting pixel, no third-party audience matching of the kind that anchors Google’s auction mechanics.

Traditional digital advertising relies on keyword-level reporting, click tracking, and pixel-based attribution. ChatGPT’s context-hint system doesn’t map onto those frameworks cleanly. An advertiser running Google Search campaigns gets exact-match keyword data linking spend to purchases; one running ChatGPT campaigns works through a Conversions API introduced in May 2026, calibrated against conversation-type descriptions as its targeting logic. Whether the 1.5x conversion rate persists as 50,000 advertisers bid for the same category query, against that measurement backdrop, is what the pilot data cannot settle.

The $14 Billion Gap

Per internal projections reported by The Information, OpenAI expects to lose $14 billion in 2026, roughly three times the estimated 2025 loss, with cumulative losses through 2029 projected at $44 billion. With over $20 billion in annualized revenue from subscriptions and API access, the company’s costs outpace revenue by a widening margin this year.

The $2.5 billion advertising target for 2026, cited in investor presentations and reported by Axios and Reuters on April 9, covers about 18% of that projected annual shortfall. The model shows advertising growing from 8% of OpenAI’s total revenue this year to 36% by 2030. OpenAI’s intermediate targets, per Axios, trace the path between those endpoints: $11 billion in advertising revenue in 2027, $25 billion in 2028, and $53 billion in 2029, each year requiring the user growth and RPQ trajectory to track closely.

Google Network revenues, down 3% in Q4 2025 and 4% in Q1 2026, confirm that conversational AI interfaces are diverting traffic from publisher display inventory. The revenue those publisher pages lose does not transfer to OpenAI; a user who gets an answer inside ChatGPT does not reach the publisher page where the display ad would have appeared. The two dynamics are distinct accounting entries for different parties in the same structural shift.

The Barclays model has ChatGPT’s revenue per query reaching 96% of Google’s 2025 level by 2030; the capital expenditure Alphabet commits between now and then will have set Google’s own benchmark somewhere else by the time ChatGPT arrives.