OpenAI says enterprise revenue passed consumer at a $40bn run rate

OpenAI finance chief Sarah Friar told investors on 14 August that the company’s enterprise business now earns more revenue than the consumer side built around ChatGPT. She said it in the same week the executive who ran that revenue left.

Friar’s remarks were first reported by CNBC and written up by The Next Web. She had earlier guided investors to expect the two sides to reach parity at the end of 2026, so the crossover has landed ahead of her own forecast.

We entered the year at 60-40, but enterprise has accelerated much faster than expected and those lines have now crossed.

Sarah Friar, CFO, OpenAI, via Unite.AI

The supporting figures arrived with it. OpenAI’s annualised run rate has reached $40 billion, roughly double where it sat a year earlier, and Unite.AI reported the same month-on-month detail from the meeting.

What OpenAI disclosed	Figure
Annualised revenue run rate	$40bn
Revenue growth, July 2026	20% month on month
Business customer growth, July 2026	32%
Advertising run rate	Approaching $1bn
Consumer to enterprise split, January 2026	60-40

That last row is the one that carries weight, because it dates the shift. Enterprise didn’t overtake consumer gradually across several years. It crossed within the first eight months of the year, which is why Friar reached for the word accelerated rather than grown.

The people who built that number have been leaving, though. Brad Lightcap closed an eight-year run on 11 August to “start something new”, and chief revenue officer Denise Dresser followed days later, roughly eight months after she arrived from Salesforce. IBTimes reported Dresser was going “to pursue other opportunities”, with Dali Rajic, who comes from the security company Wiz, taking the revenue job.

Those weren’t the only exits, which is what makes the timing awkward. Product and business chief Fidji Simo stepped back in July after a medical leave and stayed on as an adviser. A tally by 36Kr adds the head of ethics, the head of safety systems and the chief futurist, all of whom went in July.

Safety got restructured rather than simply drained. Head of safety systems Johannes Heidecke announced his departure on 11 July as OpenAI folded safety back into research under Mia Glaese, whose newly created title is VP of Research and Safety. Chief research officer Mark Chen said it’s “important that our safety work is integrated with frontier-model development, with an earlier and more direct role in shaping key model, product and launch decisions.”

The Financial Times then reported that the preparedness team was dissolved at the end of July. That team’s job was assessing whether OpenAI’s own models posed serious or catastrophic risks, and its work went to senior staff on existing teams. The Preparedness Framework it operated under is still running, so what went was the standalone group rather than the policy, a pattern we’ve traced in how every AI capability spawns an industry to contain it.

All of this sits against a filing. OpenAI submitted its IPO paperwork confidentially to the SEC in June, which means investors are pricing a revenue mix whose architects have just walked. The counter-reading is that enterprise revenue, once contracted, tends to be stickier than consumer subscriptions, so the sales team’s exit may cost less than the headlines suggest.

What nobody outside the company can check is the shape underneath the run rate. None of the figures given to investors covered customer concentration, contract length or renewal rates, and an annualised run rate extrapolates from recent performance rather than reporting a booked year. That’s the gap where enterprise AI usually disappoints, as our piece on why enterprise AI fails at deployment rather than capability worked through.

So the figure to watch isn’t the run rate. It’s whether business customer growth holds near 32% through a quarter run by a new revenue chief, because a number assembled by a departed sales organisation is a different asset to one with its team still attached.