> ## Content Index
> Fetch the complete content index at: https://askablackman.me/llms.txt
> Use this file to discover other available public pages before exploring further.

# OpenAI Expects to Burn $278 Billion by 2030
- URL: https://askablackman.me/openai-expects-to-burn-278-billion-by-2030-and-the-bill-is-already-landing-on-everyone-else/
- Published: 2026-09-20T14:30:05.000Z
- Updated: 2026-09-21T14:09:38.000Z
- Description: A leaked company presentation shows the largest planned cash burn in corporate history. The same week, the debt and the workers propping up the AI buildout started to show cracks.
- Author: Free Game News
- Tags: Finance, Business, Technology

On Friday, the Financial Times reported that OpenAI expects to burn through $278 billion in cash between 2026 and 2030\. The number comes from a private presentation the company prepared in July in connection with a computing deal, and it describes a cumulative free cash flow deficit with no real precedent in corporate history.

To appreciate what OpenAI is proposing, consider the shape of the plan. Revenue is projected to climb from $36 billion this year to $350 billion in 2030, a tenfold increase, generating a cumulative $840 billion over the period. And it still will not be enough. The company forecasts spending roughly $856 billion on computing power and infrastructure through 2030, its single largest expense. OpenAI raised $122 billion in March at an $852 billion valuation. According to the FT's account of the presentation, that money runs out in 2028.

So the most valuable private company in the world has handed investors a document that says, in effect: give us the largest revenue ramp ever achieved by any business, and we will still need hundreds of billions more in outside money to get through the decade.

Follow the money outward

OpenAI does not build most of its own data centers. It signs enormous commitments with partners, and those partners borrow against the commitments. That is where the risk stops being theoretical and starts showing up in places ordinary people can feel.

The same week the FT story landed, Reuters reported that roughly $18 billion in loans tied to Project Jupiter, a 1,400-acre Oracle-leased data center campus in Doña Ana County, New Mexico, were quoted at 89 to 91 cents on the dollar by syndicate banks including Santander and Jefferies. The project is part of Oracle's broader agreement to supply computing capacity to OpenAI. Efforts to sell the debt to a wider pool of investors have stalled, the FT reported, amid concerns about Oracle's rising borrowing and weakening creditworthiness. Local opposition over water supply and air quality has added to the pressure. Oracle's corporate credit rating sits one notch above junk.

Ninety cents on the dollar is the market's way of clearing its throat. Loans trading below par this early in the AI buildout suggest the people whose job is pricing risk are starting to price it.

Then there is the human financing layer. Oracle has cut about 21,000 roles over the past year, shrinking its workforce from roughly 162,000 to 141,000, and this month it expanded its 2026 restructuring plan by another $700 million, bringing the total to about $2.8 billion. In the quarter ended August 31, Oracle's capital expenditures hit $28.5 billion, up from $8.5 billion a year earlier, and the company posted negative free cash flow of $5.4 billion. Workers are, in a fairly literal sense, being converted into data centers.

The third leg of the stool is the neocloud IPO wave. Nscale, a London-based, Nvidia-backed cloud provider that rents GPU capacity to OpenAI, Anthropic, and Microsoft, filed this week to go public on the New York Stock Exchange under the ticker NSCL. Its prospectus is a perfect miniature of the whole economy. Revenue grew 1,252% in the first half of 2026, to $140.6 million. The net loss over the same period was $1.02 billion. The company carries more than $8 billion in debt, and one unnamed customer accounted for more than half of revenue in the first half. Nscale points to $56.4 billion in remaining performance obligations, which is another way of saying its valuation rests on contracts with the same small circle of AI labs whose own cash flow is deeply negative.

This is the architecture worth understanding. OpenAI commits to buy compute. Oracle and the neoclouds borrow to build it. Banks syndicate the loans to investors. Public market buyers are now being invited in through IPOs. Every layer of the stack is ultimately a claim on one thing: OpenAI's revenue actually reaching $350 billion by 2030.

The honest counterargument

The skeptics' case is easy to make, so it is worth stating the other side plainly. OpenAI's $278 billion burn projection is actually an improvement. A May version of the company's projections put the shortfall at $305 billion, according to the FT. The company says annualized revenue jumped about 20% in July following new model releases. And the people closest to the business have consistently argued that compute, not demand, is the binding constraint. If AI revenue keeps compounding anywhere near the projected rate, today's spending looks like railroads in the 1870s rather than fiber in 1999.

But note what the bull case requires. A company needs to grow from $36 billion to $350 billion in revenue in four years while cutting prices to compete with Anthropic and cheap open-weight models from China. It needs to keep raising private capital at ever-higher valuations, and Bloomberg reports it is already in early talks for a round that could value it above $1.2 trillion. It filed confidentially for an IPO in June, yet CEO Sam Altman said last week the company will not go public in 2026, citing concerns about AI safety. The plan works only if every one of those gears keeps turning in sequence.

Last November, Altman publicly put OpenAI's data center commitments at about $1.4 trillion over eight years. At the time, the company was doing around $20 billion in annualized revenue. The new projections do not contradict that ambition so much as put a price on it: $278 billion of somebody else's money, delivered on schedule, for five consecutive years.

The question the FT report forces is not whether AI is real. It plainly is. The question is what happens to the debt, the IPOs, the construction loans, and the 21,000 people Oracle already let go if the revenue arrives a few years late, or a few hundred billion short. A bet this size does not stay inside one company. It has already been distributed, quietly, into bond funds, bank balance sheets, stock indexes, and severance packages.

## Free Game Takeaway

You probably own a piece of this bet whether you chose to or not. Oracle is a standard holding in major stock indexes, its bonds sit in credit funds, and the AI trade has been carrying a large share of overall market returns. Here is what to watch concretely. First, Oracle's credit rating: it sits one notch above junk, and a downgrade would raise borrowing costs across the entire OpenAI supply chain. Second, the Project Jupiter loans: if $18 billion in New Mexico data center debt cannot find buyers above 90 cents on the dollar, that tells you the debt market's appetite for AI infrastructure is thinning, and every similar project gets harder to finance. Third, the Nscale IPO under ticker NSCL: a weak reception would signal public investors are done underwriting losses tied to a handful of AI customers, while a strong one keeps the funding machine running. Fourth, OpenAI's cash runway: the presentation says the $122 billion raised in March runs out in 2028, so watch whether the reported $1.2 trillion valuation round actually closes, and at what price. If you work in tech, note the Oracle pattern: companies are cutting payroll to fund capex, which means AI spending pressure can reach your job even at profitable firms. And if you live near a proposed data center, the Doña Ana County fight over water and air quality shows local opposition is now a real financial variable, priced directly into the loans.