OpenAI’s $278 Billion Bill: What Cash Burn Means – and What It Does Not

OpenAI expects to spend roughly $278 billion more than it takes in between 2026 and 2030. That is what the Financial Times reports, citing a recent presentation; Reuters and other outlets have picked up the figures. The headline sounds like a record loss, but the number means something else – and the real risks lie elsewhere.

Key takeaways

• According to the Financial Times, OpenAI expects negative free cash flow of $278 billion from 2026 to 2030; in May, the figure was still about $305 billion.
• About $856 billion is earmarked for computing power and infrastructure through the end of 2030 – more than the $840 billion in total revenue expected over the same period.
• Annual revenue is projected to rise from $36 billion (2026) to $350 billion (2030); the $122 billion March financing lasts until 2028, according to the FT.
• These are projections from a presentation, not audited figures; OpenAI did not comment at first, according to Reuters.

What the $278 billion means – and what it does not

Free cash flow is the money left after day-to-day operations and all investments. When it is negative, a company has to raise money from outside. It is not a “loss” in the accounting sense: Large purchases hit cash flow immediately, while the income statement spreads them over years as depreciation. Some headlines therefore speak loosely of billions in losses; the underlying reports are about cash flow.

By Runtimewire’s calculation, OpenAI burns about 33 cents per dollar of revenue over the period. The figure has also improved: In May the expectation was about $305 billion; the July version says $278 billion. According to Runtimewire’s reading of the FT, the reason is that annualized revenue rose by about 20 percent in July thanks to new model releases.

Why more spending does not mean more burn

There is a notable contradiction that is not one at first glance: Spending on computing power and infrastructure has risen, yet the cash burn has fallen. In February, OpenAI told investors about $600 billion through 2030, according to The Next Web citing CNBC; now the figure is about $856 billion. The site also points out that the numbers may not be fully comparable, because February referred to “compute” and July to “computing power and infrastructure.” Part of the gap could therefore be a matter of definition.

The likelier reason: Partners are carrying a large part of the buildout. Nvidia is reportedly in talks to guarantee $250 billion in data center debt. Oracle needed the asset manager PIMCO as an anchor investor for $10 billion of $16.3 billion in financing after U.S. banks pulled back. The energy provider SB Energy received warrants worth $5.5 billion for a 20-year lease. OpenAI has no investment-grade rating, meaning no credit score of the kind banks require for loans this large – so others step in with their balance sheets.

Where the risk actually lies

The $278 billion is not an independent forecast but the difference between planned spending and assumed revenue. If revenue falls short of the $350 billion projected for 2030 – a factor of almost ten in four years – the burn rises accordingly. The Next Web reminds readers that the projections have already been revised several times this year. Anyone who reads the figure as a certainty overestimates it.

More important than the sum is how the risk is distributed. According to the FT, Nvidia, Oracle and SoftBank’s data center arm depend heavily on contracts with OpenAI. If revenue comes in lower, it hits not only OpenAI but also the partners who are paying up front for the buildout. Price pressure adds to this: In July, OpenAI cut the price of a model in the GPT-5.6 line by about 80 percent, while rival Anthropic and open models put further pressure on margins. That computing power is already scarce was shown recently by the halt on new ChatGPT Pro sign-ups.

Outlook: Capital until 2028, no IPO in 2026

The March financing – $122 billion at a valuation of $852 billion – lasts only until 2028, according to the FT. Investors are now discussing a valuation of at least $1.2 trillion. According to Reuters, OpenAI confidentially filed for an IPO in June, but Sam Altman said there will be none in 2026; he cited AI safety concerns as the reason. For users, this mainly means that OpenAI still needs fresh capital, and the pressure to draw revenue from subscriptions, advertising and enterprise customers stays high. Whether the math works out will be decided less by the $278 billion figure than by whether demand grows as fast as the data centers.