FT: Leaked OpenAI Projections Show $278B of Cumulative Cash Burn Through 2030

The Financial Times reports that an internal OpenAI presentation it obtained projects roughly $278 billion of cumulative cash burn between 2026 and 2030. The provenance is the first thing to get right: this is a leaked planning document, not a voluntary disclosure, and OpenAI has not published or confirmed the figure.

Three other large numbers travel with this story in coverage, and they are not comparable to each other or to OpenAI's. Jensen Huang's oft-repeated $3 trillion to $4 trillion figure is a forecast of global AI infrastructure spending by 2030, not any single company's outlay. PwC's $31.6 trillion is a cumulative estimate running all the way to 2050, roughly an order of magnitude larger mainly because the horizon is twenty years longer. Gartner's 49.5% is a single-year growth rate for worldwide AI spending in 2026, applied to a total AI spending base of about $2.7 trillion rather than to infrastructure alone. Stacking these as though they describe the same thing is how a company-level number and an industry-level number get read as corroborating each other.

For public-market investors the OpenAI figure matters mostly as a demand signal for the supply chain. Cash burn at that scale is, in large part, capital flowing to compute: GPUs from NVDA, networking and optical components, power and data-center capacity. That is the bull case and the concentration risk in a single sentence, because it also means a meaningful share of the sector's forward revenue depends on one privately funded buyer continuing to raise capital on favorable terms.

What to watch: whether OpenAI or its investors respond to the FT's reporting, the terms and size of its next funding round, and whether hyperscaler capital-expenditure guidance in the next reporting season moves in the same direction. Guided capex is a disclosed number, which makes it a firmer read than any leaked projection.