A day after Anthropic set its IPO for November and reported a $100 billion run rate, the numbers from the other side arrived. The Financial Times has seen an internal OpenAI presentation from July: negative free cash flow of $278 billion from 2026 through the end of 2030.
What the presentation says
The $278 billion is what OpenAI expects to spend beyond what it takes in over five years. The biggest line item: roughly $856 billion for compute and infrastructure by the end of 2030.
Against that sit revenue expectations just as steep. $36 billion this year, $350 billion in 2030, $840 billion cumulatively across the period.
Bloomberg confirms the figures and places the presentation: it was part of a July compute deal. A person with knowledge of that deal declined to discuss the financials.
Spending more, burning less
In February OpenAI told investors compute would run to about $600 billion. The July figure is 43 percent higher. Yet the expected cash burn dropped, from roughly $305 billion in May to today’s $278 billion.
That sounds contradictory and isn’t. Much of the build is financed by partners, so it never shows up in OpenAI’s own till. Nvidia, Oracle and others absorb the capital expenditure through contracts and leases. The data centres get more expensive; someone else fronts the money.
Which makes revenue the number everything hangs on. Burn is spending minus income, so if the curve from $36 billion to $350 billion falls short, the gap widens faster than any saving can close it.
The cash lasts until 2028
OpenAI raised $122 billion in March at an $852 billion valuation. On this plan that money runs out in 2028. The company filed confidentially for an IPO in June, then Sam Altman said it wouldn’t happen in 2026 on safety grounds. The FT has since reported talks at a valuation around $1.2 trillion.
Want the comparison? Anthropic goes public in November and expects to pass $100 billion annualised this year. OpenAI plans on $36 billion of revenue in 2026 and wants nearly ten times that by 2030.
Both bets can pay off. They just aren’t the same bet: one needs a market that grows, the other needs a market that multiplies by ten before the money runs out.