OpenAI Trims Its Compute Spending Plan to $600 Billion

Six hundred billion dollars is the sum OpenAI now considers enough to build the compute infrastructure of its dreams. The paradox is that this is a scaling back: not long ago, the capital-spending ambitions of Sam Altman's company for compute looked even bigger. But even the revised figure remains astronomical and exceeds the annual GDP of most countries in the world.

According to the Chinese outlet 36Kr, citing sources close to the talks, OpenAI is updating its long-term capital expenditure plans for investors (算力支出, suànlì zhīchū, literally "computing power spending"). The target through 2030 is now around $600 billion. In parallel, the company is pushing a new fundraising round that could exceed $100 billion, which on its own would be the largest private financing deal in the history of the technology industry.

To understand the context of this decision, it helps to recall how quickly OpenAI's financial trajectory has changed over the past two years. The company, which generated relatively modest revenue as recently as 2023, reached annual revenue of several billion dollars by early 2025. Now OpenAI's forecast assumes that cumulative revenue will exceed $280 billion by 2030. That means the company expects to join the world's ten largest technology corporations by sales, just a few years after its commercial debut.

The structure of expected revenue is particularly notable. According to the sources, OpenAI's consumer and enterprise segments should contribute roughly equally to total revenue. That is an important signal: the company is no longer betting exclusively on ChatGPT subscriptions for a mass audience. The enterprise side (APIs for developers, custom solutions for business, integrations into corporate systems) is treated as an equal growth driver. In effect, OpenAI is building a twin-engine model in which the B2C and B2B segments balance each other, reducing dependence on a single source of income.

The key change in the updated strategy is that spending is tied to projected revenue. Sources familiar with the talks stress that investment plans are now "more directly linked to expected revenue growth." That may sound like an obvious principle for any company, but for OpenAI it is a significant shift. Until recently, the logic of Altman and his team followed the principle of "infrastructure first, monetization later": large-scale investment in GPU clusters and data centers was seen as a prerequisite for building ever more powerful models that would then find commercial use. Now investors are being offered a more traditional financial logic, in which spending is justified by concrete revenue projections.

This turn is no accident. Despite continued enthusiasm for AI, the venture and strategic capital market is becoming more demanding about justifying investments. When a round of more than $100 billion is on the table, even the most loyal investors, whether SoftBank, Microsoft or Middle Eastern sovereign funds, want to see a clear link between money in and money back. Lowering the spending target from higher earlier benchmarks to $600 billion is essentially a gesture toward financial discipline, a signal that OpenAI is ready to think not only in terms of technological breakthroughs but also in terms of a sustainable business.

Still, the word "reduction" is very relative here. Six hundred billion dollars on compute infrastructure over five years is comparable to the combined capital expenditure of all the world's largest technology companies over the same period. For comparison, Microsoft, OpenAI's largest investor, plans to spend about $80 billion on data centers in fiscal 2025 alone. Google, Amazon and Meta* are also ramping up infrastructure investment, but even their combined plans fall short of the scale OpenAI is signaling.

A natural question arises: where will the money come from? Even if the $100 billion round goes through, it would cover only a small part of the stated needs. OpenAI is evidently counting on a combination of its own growing revenue, partner investment from Microsoft and other strategic allies, and probably debt financing and infrastructure partnerships with data center operators. It is possible that part of the spending will be spread across joint ventures, such as the Stargate project announced in early 2025.

For the AI industry as a whole, this revision of OpenAI's plans carries a double signal. On one hand, the company confirms that the race for compute continues and that investment will remain unprecedented in scale. On the other, the very fact of a downward adjustment suggests that even the market leader is starting to match its ambitions to reality. The era of unbridled optimism, when any request for capital in AI was granted without awkward questions, is gradually giving way to an era in which investors want to see a path to payback. And that is, perhaps, the healthiest signal the market could have received.

*Meta is designated an extremist organization and banned in Russia.