OpenAI Projects $278 Billion in Cash Burn Through 2030

OpenAI expects to run through $278 billion of cash between 2026 and the end of 2030. The Financial Times reported the figure on Friday, September 18, 2026, saying it had seen a recent company presentation that projected negative free cash flow of that size over the five-year stretch.

Related reading: Amazon's AWS Grew 39% While Free Cash Flow Fell to -$42.4 Billion · Microsoft Borrows $25 Billion as the AI Buildout Moves Onto the Bond Market.

The number is a forecast, not a result, and OpenAI has not published it. It still shows how the company behind ChatGPT frames its own finances as it raises money and prepares for a possible stock-market listing.

The Financial Times reported the projection first. Bloomberg corroborated it, citing a person familiar with the matter who said the materials had been shared in July as part of a major computing deal. Reuters, which carried the FT's account, said it could not immediately verify the report, and KSL published the same wire story. OpenAI declined to comment, Bloomberg said.

What the presentation reportedly shows

The burn figure is one part of a larger picture described in the reports. According to the FT account, the same materials project revenue rising from $36 billion in 2026 to $350 billion in 2030, a roughly tenfold increase over four years. Cumulative revenue through the end of the decade would reach about $840 billion.

The largest cost is computing power. The report said OpenAI expects to spend about $856 billion on computing and infrastructure by the end of 2030, which the FT described as the company's biggest expense category. That spending is what turns a fast-growing revenue line into a cash deficit: the company is building capacity faster than customers pay for it.

For scale, the five-year deficit is more than twenty times the roughly $13 billion in revenue OpenAI recorded in 2025, according to figures reported by CNBC and compiled by analyst trackers. The gap between what the company spends and what it collects is the whole story of the projection.

The financing math is tight. The FT reported that OpenAI raised $122 billion in March 2026 at an $852 billion valuation and is on track to exhaust that cash by 2028, two years before the projection window closes. The newspaper also reported earlier in the week that OpenAI had held talks with investors that could value the company at about $1.2 trillion ahead of a potential listing.

OpenAI filed confidentially for an initial public offering in June 2026. Sam Altman has said the company will not go public in 2026, citing AI-safety concerns, according to the Reuters account.

Why a cash-flow figure matters more than a headline loss

Free cash flow is the money left after operating costs and capital spending. Negative free cash flow means a business is consuming cash rather than generating it. A cumulative $278 billion shortfall over five years works out to an average of more than $50 billion a year, a scale few private companies have tried to finance.

The difference between a cash-flow deficit and an accounting loss is worth keeping straight. A loss reflects revenue minus expenses, including non-cash items such as depreciation. A cash deficit reflects money actually leaving the building. For a company buying servers, leasing data-center space and signing long-term power contracts, the cash number determines how much outside money is needed and when.

For comparison, the entire market value of most large technology companies is smaller than the cumulative outlay in the reported plan. Capital-intensive industries such as airlines and telecom carriers routinely carry heavy debt, but they do so against established revenue. OpenAI is asking investors to fund a deficit that runs for five years before the business turns cash-positive.

That timing is the real story. A burn that runs through 2030 is manageable only if investors keep supplying capital. The reported projection that cash runs out in 2028 gives OpenAI a window of roughly two years to raise more, whether through private rounds, debt or an IPO.

The funding question

A $278 billion deficit cannot be covered from operations. It has to come from investors, lenders or asset sales. OpenAI has already assembled large backers and closed funding rounds at rising valuations, and the reported $1.2 trillion figure attached to recent talks would place it among the most valuable private companies ever.

Each round comes with trade-offs. New capital dilutes earlier holders, and a higher valuation raises the return the business must eventually produce. Strategic investors who also supply chips or cloud capacity blur the line between customer and shareholder. The reported detail that the July materials traveled with a computing deal is a reminder that OpenAI's fundraising and its infrastructure commitments are entangled.

The IPO is the obvious next step. Altman's stated caution about a 2026 debut leaves 2027 and later in play. Anyone buying that offering would be buying a company that has told them, in its own materials, that it plans to spend more than it earns for years.

The 2030 revenue target has moved up

The new projection is also a revision. In February 2026, Bloomberg reported that OpenAI was telling investors its revenue would top $280 billion in 2030. The figure now tied to the presentation is $350 billion, about a quarter higher. The direction of travel is upward, which fits a company that has repeatedly raised its own expectations.

Higher revenue targets cut both ways. They make the spending plan look more defensible, because a larger future business can absorb more capital. They also set a higher bar that the company has to clear, and they depend on assumptions about consumer subscriptions, enterprise sales and advertising that have not been tested at this scale.

Can $350 billion of 2030 revenue justify it?

The arithmetic is demanding rather than impossible. A company with $350 billion in annual revenue would sit among the largest businesses in the world, bigger than most of today's technology giants. Whether it earns a profit depends on how much of each revenue dollar is consumed by compute.

OpenAI's reported cost structure shows why investors watch training and inference expenses closely. The product runs on expensive hardware that depreciates and draws power, so gross margins are thinner than those of a typical software company. Analyst estimates have placed OpenAI's gross margin in the low 30 percent range, well below the 70 to 80 percent common among mature software firms. A lower gross margin leaves less room to cover research, sales and administration before reaching break-even.

The comparison with a rival sharpens the picture. Analyst estimates compiled by Value Add VC put Anthropic's free-cash-flow break-even in 2027 to 2028, roughly two years ahead of the 2029 to 2030 target the same analysis attributes to OpenAI. The reported difference reflects business mix: Anthropic sells more through enterprise and developer API channels, while OpenAI leans on consumer ChatGPT subscriptions, where most users pay nothing. Value Add VC labels both timelines as analyst estimates, not company guidance.

Revenue estimates for OpenAI have moved through the year as well. Some analyst trackers put annualized revenue near $25 billion in the first half of 2026 after a plateau that began in February, with later estimates climbing toward $40 billion as enterprise sales grew. Those are outside estimates, not audited figures, and the company does not publish them on a fixed schedule.

The capex debate the number feeds

The report lands in the middle of an argument about whether the AI build-out has outrun demand. Data centers, graphics processors and electricity have absorbed hundreds of billions of dollars across the industry, and much of the return is still projected rather than realized. A company telling investors it will burn $278 billion before breaking even is a concrete data point for that debate.

The optimistic reading is that the projection shows confidence backed by real growth. A business expecting to go from $36 billion to $350 billion in four years is scaling a product that already sells, not chasing a hypothetical market. If the revenue arrives, the spending becomes capacity that rivals would struggle to match.

The cautious reading focuses on the gap. The same document that forecasts $350 billion of 2030 revenue also forecasts about $856 billion of compute and infrastructure spending by then, alongside the $278 billion cumulative cash deficit. For the plan to work, revenue must grow faster than costs for the rest of the decade. If growth slows, much of the spending is already committed, because data-center leases and chip orders are signed years in advance.

There is a market-level question too. The AI industry's largest customers and suppliers are connected through a web of deals in which the same handful of companies invest, build and buy. A funding slowdown at one large player would be felt by the others. That interconnection is why the OpenAI projection is read as a signal about the whole sector, not one company alone.

What to watch

Three developments would turn this from a reported forecast into a clearer picture. The first is whether OpenAI confirms the numbers or files a prospectus that discloses them, which would move them from leaks to audited figures. The second is whether the company closes the funding round implied by the reported valuation talks, and on what terms. The third is the revenue line itself.

If the run rate keeps climbing toward the reported 2030 target, the burn looks like the cost of winning a market. If it stalls, the same numbers read very differently. For now, the $278 billion figure is OpenAI's own projection, reported by two outlets from the same document and unverified by a third. It says more about the price of the company's ambitions than about whether they will pay off.

Sam Altman in November 2022. Photo: Village Global / Wikimedia Commons (CC BY 2.0).