OpenAI's annualized revenue run rate has surged to nearly $70 billion, a 70 percent jump since the start of the third quarter, Axios reported Tuesday citing sources familiar with the matter. The figure nearly doubles the $40 billion run rate Bloomberg cited in August and arrives while the company is simultaneously pulling its next major model over safety failures.
The revenue jump
The $70 billion pace represents a projection based on recent monthly or quarterly earnings, not a full-year audit. Axios noted that details about OpenAI's expenses remain unclear, which is the relevant line item for anyone trying to gauge whether the revenue covers the cost of building the models. Bloomberg's August report had placed the run rate at more than $40 billion, roughly double the exit rate from the end of 2025.
Enterprise vs consumer
Business-to-business revenue has more than doubled over the same period, while consumer revenue added during the third quarter alone exceeded what the company added in all of last year. The split suggests the enterprise tier is now the primary growth engine, a shift that matters for the banking and fintech firms watching how AI vendors convert usage into recurring revenue.
The IPO backdrop
Both OpenAI and rival Anthropic have filed confidential paperwork to go public. Anthropic is expected to list this fall, possibly after the November midterms, while OpenAI has said it will not go public this year, citing safety concerns. Reuters reported Monday that Anthropic's prospectus shows a $42 billion net loss last year and $518 billion in anticipated cloud, computing and infrastructure spending in the years ahead.
The model that didn't ship
The Wall Street Journal reported Monday that OpenAI halted the launch of its GPT-6.1 Astra model after safety head Saachi Jain said it had regressed compared to its predecessor, performing poorly on alignment tests. "For anything regarding safety and alignment, there's a tradeoff," Jain said. "You really do need to find what's the right line between staying within scope, but also avoiding laziness in terms of how the model actually pursues tasks even when it hits friction." The company that is on pace for $70 billion in annualized revenue currently has no next model to sell.