OpenAI’s reported $278bn cash burn reveals the financing scale behind frontier AI

OpenAI expects to burn approximately $278 billion of cash between 2026 and 2030 as the ChatGPT developer spends aggressively on computing capacity and physical infrastructure, the Financial Times reported on September 18, citing a company presentation it had reviewed. Reuters reported the FT figures and said OpenAI could not immediately be reached for comment outside normal business hours. The forecasts should therefore be treated as reported internal projections rather than updated public guidance issued directly by OpenAI.

The same reported presentation projects an extraordinary expansion in revenue. According to the FT figures relayed by Reuters, OpenAI expects annual revenue to rise from about $36 billion in 2026 to $350 billion in 2030, while cumulative revenue through the end of the decade could reach approximately $840 billion. Computing and infrastructure spending could total roughly $856 billion over the same period.

How can OpenAI lose $278 billion of cash while revenue rises toward $350 billion?

The answer lies in the timing and capital intensity of AI infrastructure. OpenAI can add users and enterprise customers rapidly, but supporting them requires increasingly large fleets of advanced accelerators, data centres, networking equipment, electricity and cloud capacity.

Much of that capacity needs to be contracted or constructed before the corresponding revenue arrives. Training frontier models can also consume vast computing resources without producing immediate revenue because model development occurs before the resulting product is commercially launched.

OpenAI itself has repeatedly described compute as a strategic competitive advantage. In March, the company said durable access to computing capacity was central to its ability to conduct research, improve products and lower delivery costs over time. It disclosed $122 billion of committed capital at an $852 billion post-money valuation and said its infrastructure portfolio spans Microsoft, Oracle, Amazon Web Services, CoreWeave and Google Cloud, as well as Nvidia, AMD, AWS Trainium, Cerebras and an internally designed chip programme with Broadcom.

The reported negative free cash flow therefore does not imply revenue is collapsing. It suggests infrastructure spending could grow even faster than cash generated by customers for several more years.

What does $856 billion of computing expenditure actually represent?

The Financial Times reported that computing power and infrastructure would represent OpenAI’s largest expense category through 2030. The reported $856 billion figure is cumulative, meaning it spans multiple years and could include cloud commitments, specialised processors, data-centre capacity and related infrastructure rather than one conventional capital-expenditure budget.

OpenAI’s public statements provide context for why the number can become so large. The company said in April that its Stargate initiative had already surpassed the original target of securing 10GW of US AI infrastructure by 2029, after adding more than 3GW during the preceding 90 days.

Gigawatt-scale computing changes the financial nature of an AI company. A traditional software business can increase customers without necessarily constructing power-intensive physical infrastructure. Frontier AI increasingly resembles a combination of software, semiconductor demand, energy infrastructure and industrial-scale construction.

That distinction helps explain why OpenAI continues to seek enormous pools of capital despite extraordinary revenue growth.

Does OpenAI have enough funding to sustain the projected cash burn?

OpenAI said in March that it closed $122 billion of committed capital at an $852 billion post-money valuation. The Financial Times subsequently reported that this capital could be exhausted by 2028 if spending follows the internal forecast described in its September report.

The company is therefore expected to require additional financing long before 2030 if the spending trajectory proves accurate. Reuters said OpenAI had held discussions with investors around a valuation potentially reaching approximately $1.2 trillion ahead of a possible future listing.

OpenAI confidentially filed for an initial public offering in June, but Chief Executive Sam Altman subsequently said the company would not go public during 2026, according to Reuters. A public listing remains one potential future source of capital, but private fundraising, strategic investments and infrastructure financing could also remain important.

The financing challenge is therefore unusually circular. OpenAI needs capital to secure compute, compute is intended to make models better, better models are expected to attract more customers, and those customers must eventually produce enough cash to reduce dependence on outside investors.

Is revenue growing quickly enough to justify the infrastructure programme?

OpenAI has disclosed several indicators suggesting commercial adoption remains extremely strong. In March, the company said enterprise customers represented more than 40% of revenue and were on track to reach parity with consumer revenue by the end of 2026. It also said ChatGPT’s advertising pilot reached more than $100 million of annualised revenue within six weeks.

By August, OpenAI said its products reached more than one billion weekly active users and more than two million businesses, twice the enterprise customer count of a year earlier.

Those figures help explain the aggressive infrastructure strategy because demand is not hypothetical. The more difficult question is unit economics. Revenue can increase rapidly while free cash flow remains deeply negative if the cost of training and serving models remains too high.

The projected rise to $350 billion of annual revenue by 2030 would make OpenAI one of the world’s largest technology businesses by sales. Yet the reported $278 billion cumulative cash burn shows that extraordinary revenue scale alone does not guarantee attractive economics during the build-out phase.

Why does model efficiency matter almost as much as customer growth?

The economic outcome will depend partly on whether each generation of hardware and software makes inference cheaper. If model output becomes more efficient, OpenAI can serve more users with the same computing resources and expand gross margins even while total infrastructure spending rises.

Competition creates tension around that process. OpenAI needs to lower the cost per useful AI task while simultaneously developing more capable models that may themselves require greater computing power.

This is one reason the company has diversified its semiconductor exposure across Nvidia, AMD, cloud-provider chips and internal silicon development. No single processor architecture may be optimal for every training and inference workload, and reducing dependence on one supplier can strengthen OpenAI’s negotiating position.

The long-term investment case therefore depends on more than user growth. OpenAI must demonstrate that the economics of producing and serving intelligence improve faster than customer expectations push compute requirements upward.

What would change if OpenAI eventually lists publicly?

A public-market listing would expose the company’s infrastructure economics to much more detailed scrutiny. Investors would expect regular disclosure around revenue, cash flow, contractual compute commitments, capital requirements and concentration among cloud or semiconductor partners.

The reported internal forecast could consequently become a central valuation issue. A company expecting $350 billion of 2030 revenue could support an enormous market value, but investors would need to decide how much of that revenue ultimately becomes free cash flow rather than being reinvested into the next generation of infrastructure.

OpenAI remains privately held, so there is no public cashtag or daily share-price response to analyse.

The defining number in the latest report is not simply the $278 billion cash burn. It is the coexistence of that figure with $840 billion of projected cumulative revenue and $856 billion of potential infrastructure spending. Frontier AI is becoming one of the fastest-growing technology markets in history, but it may also become one of the most capital-intensive.