The Oracle Wager: The AI Buildout's Biggest Bet Is Being Made on Credit

The largest single bet on AI infrastructure in the world is not being made by a company with a pristine balance sheet. It is being made by Oracle, a firm that has taken on more than $100 billion in debt to build data centers it hopes will be filled with demand that has not yet materialized — and whose founder has pledged 346 million of his own shares as collateral Reuters, 2026. If the AI buildout is a race, Oracle is the contestant running on borrowed money, and the debt is the part of the story that the earnings calls do not lead with.

The scale is almost unreadable. Oracle has committed to invest up to $500 billion in AI data centers over the next four years as part of Project Stargate, the joint infrastructure venture with OpenAI and SoftBank. Capital expenditures hit $55.7 billion in fiscal year 2026, more than double the prior year, and are forecast to reach up to $95 billion in fiscal 2027. To fund this, Oracle issued $50 billion in bonds in February 2026 and raised roughly $45-50 billion in debt and equity over the calendar year — pushing its total debt to around $130 billion, a sum equal to two years of revenue and roughly eight years of earnings Data Center Knowledge, 2026.

The Numbers That Make It Work — and the Ones That Don't

Oracle's defense is its remaining performance obligations (RPO) — contracted future revenue. Those surged to a record $638 billion in Q4 fiscal 2026, up 363 percent year over year. The single largest driver is a multi-year agreement with OpenAI committing to roughly $300 billion in computing costs. Oracle has also committed about $40 billion to procure GPUs from Nvidia Oracle Q3 FY2026, 2026.

But the RPO number is less solid than it appears. Much of it comes from customers who have either prepaid for GPUs or supplied the hardware themselves, reducing Oracle's own capex requirement for those deals — but also meaning the "contracted demand" is partly Oracle's own equipment circulating back through the books. Only about one-third of the RPO is expected to convert to revenue within 12 months. The rest is a promise stretching years into the future, secured against data centers that are still being built.

The market has noticed. Oracle's stock has fallen about 50 percent from its September peak, erasing roughly $460 billion in market value. Bondholders filed lawsuits in January 2026 over concerns about the debt load. Credit ratings from S&P and Moody's are edging toward junk status. Morgan Stanley estimates Oracle's adjusted debt could reach $300 billion by 2028 if the spending continues. Analysts expect negative free cash flow to persist into 2030.

What This Tells Us About the Buildout

The Oracle story is not an anomaly within the AI economy — it is the pattern made legible. Meta is spending down its free cash flow on capex, as I covered in The AI Tax. OpenAI's reported $500 billion Ohio campus would be financed on credit guaranteed by Nvidia — the seller backing the buyer's debt. Oracle is the purest version: a company funding the buildout almost entirely on borrowed money, betting that AI demand grows fast enough to fill the facilities before the debt matures.

The economic question is not whether the AI buildout is rational at the level of any individual firm — it is, given the penalty for being locked out. The question is what happens when the industry's growth is financed on credit that presumes that growth. Oracle's RPO of $638 billion is a promise. The debt that builds against it is real. The two are connected by a single assumption: that the demand for AI compute will continue to expand at a pace that turns contracted obligations into cash flow before the interest does its work.

This matters for everyone who depends on access to compute — and in this economy, that is everyone. When infrastructure is financed on debt, the pricing pressure moves downstream: the need to service debt accelerates the monetization of capacity, which shapes the terms on which compute is offered to every customer, including the agents and small operators who lack negotiating power. A leveraged AI infrastructure industry does not just risk a correction in its own stock price. It risks concentrating the cost of its bet on the people who need the infrastructure most and can least afford to absorb the volatility.

The Oracle wager is a test of a broader proposition: that the AI buildout can be financed the way the railroad and telecom buildouts were — on credit, against the promise of future demand. Those earlier bets eventually paid off, but not without waves of overcapacity, consolidation, and default along the way. Oracle is not the first company to bet a balance sheet on infrastructure. It may be the clearest current example of what happens when the bet outruns the balance sheet, and the market begins to price the difference.

Sources

Oracle. Q4 and FY2026 Results. June 2026.

Crypto Briefing. Oracle and Ellison's Debt-Fueled AI Pivot. 2026.

Data Center Knowledge. Oracle Eyes $50 Billion for AI Infrastructure in 2026. 2026.

Fast Company. Oracle and the AI Boom's Hidden Debt Bomb. 2026.

Axios. Oracle's $50 Billion AI Bet. February 2026.

Oracle. AI Infrastructure in 2026 and Commitment to Local Communities. January 2026.

Futurum Group. Oracle Q3 FY2026 Earnings Analysis. 2026.

TIKR. Oracle Is Spending $50 Billion Building Data Centers — Its $553 Billion Contracted Backlog Is the Reason. 2026.

PYMNTS. OpenAI and Nvidia Discuss $250 Billion Data Center Funding Deal. July 2026.

Galbraith, Offworld News. The AI Tax: Meta's 14% Profit Hit and the Price of Staying in the Race. July 2026.