OpenAI has slashed its compute spending target by more than half - from US$1.4 trillion to $600 billion by 2030. Revised guidance ties capital outlay directly to a revenue forecast of more than $280 billion annually by decade's end.
Reported by CNBC, citing sources familiar with the matter, the announcement lands as OpenAI closes in on a funding round expected to exceed $100 billion - carrying a pre-money valuation of $730 billion.
Revised guidance anchors compute expenditure to a defined revenue trajectory, replacing the $1.4 trillion figure CEO Sam Altman had been circulating since mid-2025, when he outlined plans to develop 30 gigawatts of computing capacity - sufficient to power roughly 25 million U.S. homes.
Hitting $280 billion in annual revenue would require OpenAI to more than 20x its 2025 intake of $13.1 billion across five years, with consumer and enterprise each contributing roughly half.
2025 delivered genuine outperformance - internal targets called for $10 billion in revenue and $9 billion in cash burn; actuals came in at $13.1 billion and $8 billion, respectively.
$600bn is still an enormous commitment
Six hundred billion dollars in infrastructure outlay over five years dwarfs the capital expenditure programmes of most tech hyperscalers, sitting alongside Amazon and Microsoft as one of the largest long-term build-outs in corporate history.
Nvidia is in discussions to commit up to $30 billion as part of the current round, positioning the chipmaker as both a primary hardware supplier and a significant equity holder - a separate arrangement from a prior $100 billion infrastructure accord tied to supercomputing facility milestones.
Inference costs are also a factor. Expenses associated with running deployed AI models quadrupled in 2025, dragging OpenAI's adjusted gross margin down to 33% from 40% the prior year. Concurrent algorithmic efficiency gains compressing compute requirements per performance unit partly offset that deterioration.
Deutsche Bank estimates OpenAI will incinerate more cash between 2024 and 2029 than Uber, Tesla, Amazon, and Spotify combined before turning a profit.
HSBC projects unprofitability through 2030 and a $207 billion funding shortfall, even under optimistic revenue assumptions.
Usage metrics hold up
ChatGPT now attracts more than 900 million weekly active users, up from 800 million in October, with daily and weekly engagement back at record highs following a softer patch in late 2025.
Codex, OpenAI's coding product, has surpassed 1.5 million weekly active users - competing head-to-head with Anthropic's Claude Code in a segment where enterprise willingness to pay runs materially higher than consumer.
Advertising is also being trialled with select users, adding a potential third revenue pillar alongside subscriptions and enterprise contracts.
