OpenAI Plans to Burn Nearly $280 Billion by 2030

Threat assessment: OpenAI’s internal math says the road to AGI costs roughly $280 billion in cash before the company breaks even. That’s the headline number, and it should reset how everyone in this industry thinks about the price of staying at the frontier.

OpenAI now forecasts nearly $280 billion in cumulative cash burn through the end of 2030, according to The Information, which reported the figure based on the company’s internal financial projections. The report lands as OpenAI keeps signing compute deals measured in gigawatts and hundreds of billions of dollars.

Here’s the briefing.

1. What Happened

The Information reports that OpenAI’s internal forecasts now show close to $280 billion in cash burn between now and the end of 2030. That’s cash out the door, not accounting losses. It covers the gap between what OpenAI earns from ChatGPT subscriptions, API sales, and enterprise deals, and what it spends on training runs, inference, data centers, and talent.

The source article is short on line items. But the direction is clear: the number went up. Way up.

2. How This Compares to the Last Forecast

This is the part that stands out. Roughly a year ago, The Information reported that OpenAI expected to burn about $115 billion through 2029. That already looked staggering at the time.

The new figure more than doubles it, with only one extra year added to the horizon. Two things could explain the jump:

• Compute commitments ballooned. Since last fall, OpenAI has stacked up massive infrastructure agreements with Nvidia, AMD, Oracle, Broadcom, and the Stargate partners. Sam Altman himself has put the company’s total compute commitments in the trillion-dollar range over the coming years. Someone has to fund the early years of those contracts before the revenue catches up.
• The revenue ramp is real but not fast enough. OpenAI’s revenue has been growing at a pace few companies have ever seen. It just isn’t growing as fast as the cost of building frontier-scale infrastructure.

My read: this isn’t a company losing control of its costs. It’s a company deciding that the winner of this race is whoever controls the most compute, and pricing that bet honestly.

3. Why It Matters

A $280 billion burn forecast changes the strategic map in a few ways:

1. Fundraising becomes a permanent operation. OpenAI will need to raise, borrow, or pre-sell its way through the rest of the decade. Expect more mega-rounds, more debt financing tied to data centers, and more creative deal structures where chip vendors effectively fund their own customer.
2. The moat is capital, not just talent. If it costs this much to stay at the frontier, the number of labs that can compete shrinks. Anthropic, Google DeepMind, Meta, and xAI all face versions of the same bill. Everyone else is building on top of them.
3. The IPO clock is ticking louder. Burn at this scale is hard to sustain on private money alone. A public listing, or something close to it, becomes less a question of if and more of when.
4. Pricing pressure flows downstream. OpenAI needs revenue. That means pushing enterprise contracts, ads in ChatGPT, higher-tier subscriptions, and agentic products that can charge per outcome rather than per token. Developers building on the API should expect the pricing model to keep evolving.

4. What Practitioners Should Watch

• Model pricing and rate limits. A company under this much cash pressure will optimize hard on inference costs. Cheaper models and aggressive tiering are likely; free-tier generosity is not guaranteed.
• Contract terms. If you run production workloads on OpenAI, the push toward committed-spend agreements will intensify. Lock in terms while you have leverage.
• The competitive gap. Watch whether rivals match this spending or choose efficiency. The answer decides whether the frontier stays a two- or three-horse race.
• Signals of strain. Delayed data center timelines, renegotiated chip deals, or slower model cadence would be early warnings that the forecast is biting.

5. The Bottom-Up View

There’s a version of this story where $280 billion is a bargain. If OpenAI captures even a modest share of global knowledge work, the payoff dwarfs the spend. There’s another version where compute gets cheaper faster than expected, and the company overbuilt.

What’s certain is that OpenAI has told us exactly how much it thinks the next four years cost. That level of clarity, even when leaked, is rare. Plan accordingly.

The Information has the full details of the projections.