OpenAI's revenue run rate jumped from $40 billion to near $70 billion in under seven weeks -- Anthropic has been leaking a similar curve all year

OpenAI's annualized revenue run rate is nearing $70 billion, Axios reported Sept. 29, citing people familiar with the company's financials -- up from the roughly $40 billion pace Bloomberg reported Aug. 13, and about $20 billion at the end of 2025, according to OpenAI's own chief financial officer. None of the three figures is an audited revenue number. All three are the same kind of estimate, run through the same kind of anonymous sourcing, at increasingly short intervals.

Axios's sourcing says OpenAI's business-to-business revenue more than doubled since July, and the company added more consumer revenue in the third quarter alone than it added in all of 2025 -- growth the report attributes largely to the spread of OpenAI's AI coding products. Bloomberg's mid-August figure had already described the run rate as roughly doubling from where OpenAI ended 2025, a pace co-founder Greg Brockman characterized internally in July as growing more than 20% month over month.

OpenAI's own finance chief has already tied the growth rate to the listing calendar. CFO Sarah Friar told staff at an internal all-hands in mid-August that OpenAI will be a public company in 2027 or sooner if the business keeps 'inflecting,' framing the eventual IPO as another fundraising step rather than a finish line -- a framing that turns each successive run-rate leak from a stray data point into a running scoreboard for exactly the decision Friar described. Framing growth as an *annualized run rate* rather than trailing revenue is standard practice for a fast-growing private company courting IPO investors: it captures the most recent month's pace rather than diluting it with a slower start to the year, which is part of why the number moves so much faster than an audited quarterly result would.

OpenAI is not the only lab leaking an accelerating number. Anthropic told investors its own annualized revenue run rate hit $65 billion in July, reported Aug. 17 -- more than sevenfold its pace at the end of 2025 -- and people close to the company have told reporters they expect it to finish 2026 between $100 billion and $120 billion. The two labs are now reporting run-rate figures on almost identical timelines, through almost identical anonymous sourcing, ahead of what would be two of the largest IPOs on record.

'More than doubled' is a base-rate question as much as a growth one, and Axios's report doesn't say what OpenAI's enterprise revenue doubled from. A number that goes from $2 billion to $5 billion and a number that goes from $20 billion to $50 billion both round to 'more than doubled,' and only one of those is the kind of growth that gets you to a $70 billion total run rate on its own. Neither Axios's sourcing nor OpenAI's own public statements break out what share of the run rate is enterprise contracts, ChatGPT subscriptions, API usage, or the advertising business the company has been quietly building -- which is the same gap that makes a leaked run-rate figure a headline rather than a financial statement.

Laid side by side, the gap that stands out isn't the growth -- it's the reporting lag. Anthropic's own number hasn't moved in the press for six weeks while OpenAI's has been updated twice in that span, which says as much about which company is currently more eager to have a favorable number in circulation as it does about which company is actually growing faster.

Scale matters for reading either number. DeepSeek's own revenue run rate reportedly crossed $1 billion this month -- treated as a milestone in its own right -- which puts OpenAI's and Anthropic's run rates at roughly 65 to 70 times that figure. The three companies aren't really competing on the same axis: DeepSeek is scaling a much cheaper model at a much smaller revenue base, while OpenAI and Anthropic are burning through record-setting private funding rounds to chase enterprise and consumer share simultaneously. A run-rate comparison that stops at 'bigger number wins' misses that the three labs are pricing three different bets.

Neither curve says anything about the other side of the ledger. A leaked OpenAI document projected $278 billion in cash burn through 2030, with the company not expected to turn cash-flow positive until 2030 even as its projected annual burn rises toward $63 billion in 2027. Anthropic has not disclosed a comparable long-range burn projection, but a company growing revenue sevenfold in eight months is also a company spending at a pace that growth requires -- fresh compute commitments, not fresh profit, are what a run-rate headline is actually funding.

Both companies have kept their IPO timing deliberately vague, and inconsistently so. Sam Altman ruled out an OpenAI listing in 2026 less than a month after his own CFO told staff a 2027-or-sooner listing was coming -- a gap between the public and internal messaging that the fresh revenue number does nothing to close. OpenAI's confidential S-1, filed in May, stays on file with the option to move faster regardless of what Altman has said publicly. Anthropic has given no public listing date either, though people close to the company have told reporters bankers are discussing a debut as soon as October -- at a valuation some reports put near $2 trillion, roughly double the $965 billion mark Anthropic's own funding round priced it at in May.