By OpenAI announced this week that its ChatGPT advertising pilot has surpassed $100 million in annualized revenue within six weeks of its U.S. launch. The company said it now has more than 600 advertisers on the platform and plans to launch self-serve access in April.
That's the press release version. Here's what actually happened: a company burning through an estimated $15 billion in cash just discovered the oldest business model on the internet, and Wall Street is acting like they invented it.
The numbers worth reading
$100 million in annualized revenue sounds impressive until you look at how they got there. According to Reuters, only about 20% of eligible users are seeing ads on any given day, even though roughly 85% of Free and Go tier users qualify. That means OpenAI is monetizing a sliver of its audience and still pulling nine figures.
The CPMs tell the real story. According to Adweek, advertisers are paying around $60 per thousand impressions with a minimum commitment of $200,000 to get in the door. For context, Google Search CPMs typically run between $1 and $5. Meta hovers around $8 to $12. ChatGPT is charging a premium that would make a luxury publisher blush.
Click-through rates, according to Storyboard18 citing Adweek data, are sitting at about 0.91%, compared to roughly 6.4% for Google Search. The relevance question is real: OpenAI says fewer than 7% of ads are rated "low relevance" by users, which they frame as a positive. Whether advertisers keep paying $60 CPMs for sub-1% click rates is the $100 million question.
Follow the money
The timing here is not subtle. OpenAI executives have told investors the company expects to generate more than $17 billion from ChatGPT consumers in 2026, with advertising representing what they call "a meaningful slice" of revenue from the free user base. An IPO is on the horizon. Ads are part of the profitability story they need to tell.
The company shelved its Sora video tool and paused plans for other experimental products. The focus has sharpened on things that generate revenue right now. Advertising, it turns out, is very good at that.
Meanwhile, OpenAI hired Dave Dugan, a former VP of global clients and agencies at Meta, to run ad sales. When you poach someone from Meta's ad operation, you're not experimenting anymore. You're building an ad business.
The early advertiser roster confirms the strategy. Target, Adobe, Williams-Sonoma, and Albertsons were among the first brands in the pilot when it launched on February 9. These are not scrappy DTC brands testing a new channel. These are enterprise advertisers with real media budgets, and they showed up fast.
Why the self-serve launch matters more than the $100M number
The managed pilot with 600+ advertisers and $200K minimums is nice, but it is a concierge business. The real inflection point is self-serve access in April. That is when the platform opens to the long tail: the mid-market brands, the performance marketers, the agencies running media for dozens of clients who don't want to call an OpenAI sales rep.
Anyone who was around for early Facebook advertising remembers this exact playbook. Facebook ran its own managed ad sales for big brands, then launched self-serve in 2007 and watched revenue compound. Google did the same thing with AdWords in the early 2000s. The pattern is always the same: prove demand with enterprise, then open the floodgates to everyone else.
OpenAI is also expanding geographically. The company said it will begin pilots in Canada, Australia, and New Zealand in the coming weeks, with "many more markets" planned later this year.
What this actually means for advertisers
Here is what nobody in OpenAI's press materials will say plainly: this is a bet on intent data that might be better than Google's.
When someone types a query into ChatGPT, they're often expressing a need in natural language with far more context than a three-word search query. "I'm renovating my kitchen and need affordable countertop options that are durable for a family with kids" gives an advertiser more to work with than "kitchen countertops." If OpenAI can match ads to that kind of signal without creeping users out, the $60 CPMs might actually hold.
But there are real risks. OpenAI insists ads will not influence ChatGPT's responses and that conversations remain private. That promise gets harder to keep as the ad business scales and revenue targets tighten. Every ad-supported platform in history has faced this tension, and every single one has eventually tilted toward the money.
The other risk is simpler: attention. ChatGPT users are there to get answers and leave. They're not scrolling a feed. They're not browsing. The 0.91% CTR suggests that ads placed at the bottom of an answer may get seen but not clicked. If performance doesn't improve as the platform scales, the premium CPMs will come under pressure.
What's next
Self-serve in April. Geographic expansion through the year. And a revenue trajectory that OpenAI will need to sustain through its IPO process.
The $100 million ARR number, generated from less than 20% of eligible users, is the number OpenAI wants investors to extrapolate from. If they can 5x the ad-exposed audience without tanking user satisfaction, the math gets very large very fast.
But extrapolation is not revenue. The ad industry has a long memory of platforms that launched with premium pricing and rich targeting, then watched CPMs collapse as supply flooded in. Twitter did it. Snapchat did it. The question for OpenAI is whether conversational AI ads are a genuinely new format with staying power, or another premium that erodes once the novelty wears off.
Right now, the money says it's real. Six hundred advertisers in six weeks, with enterprise brands paying $60 CPMs, is not a soft launch. It is a land grab. And with self-serve coming in April, the grab is about to get a lot bigger.
Zach El-Amin covers ad tech and the business of advertising for The Daily Vibe.
