Oracle's $300 Billion OpenAI Bet Collides With a $20 Billion Borrowing Cap

Oracle's sprawling AI infrastructure deal with OpenAI carries both long-term revenue promise and substantial financing risk, and two developments this week have sharpened the question of which force will prevail. Bernstein published an assessment placing OpenAI's identified contract obligations through 2030 at $614 billion in total infrastructure costs, with roughly $300 billion of that tied to capacity agreements at Oracle. The arrangement runs five to six years, and OpenAI accounts for less than half of Oracle's remaining performance obligations — a detail that limits customer concentration provided the rest of the backlog converts to revenue on schedule.

That $614 billion figure rests on every identified contract proceeding as planned. Bernstein considers that unlikely given potential construction delays, cancellations and flexible hardware purchasing, so for Oracle the timing of when signed volume becomes actual business matters as much as the headline number. Analyst Mark Moerdler sees limited near-term risk for the largest cloud providers, a view that favors Oracle as long as the committed capacity is genuinely taken up and the buildout stays financeable.

Balance sheet meets backlog

The financing side is where the picture darkens. As of August 31, Oracle carried $125.34 billion in debt and other borrowings against roughly $36.37 billion in liquid assets, and heavy capital spending had pushed free cash flow negative. Management is trying to contain the funding requirement through supplier financing, customer-provided hardware and prepayments — tools that can lighten the load if suppliers and customers shoulder part of the expansion, making them a central link between infrastructure spending and later returns.

Guidance calls for free cash flow to turn positive in the second half of fiscal 2028, then exceed $30 billion in fiscal 2029, at which point Oracle expects to need no external financing. The company has also pledged to raise no more than $20 billion in additional debt through fiscal 2030, a self-imposed ceiling that turns capital discipline into a testable part of the story.

Capital markets have already been tapped. Early in the week Oracle priced an equity offering and senior notes, with net proceeds of about $9.9 billion and $10.0 billion respectively, earmarked for general corporate purposes. The broad use of proceeds puts the spotlight on which investments Oracle prioritizes and whether they fit inside the borrowing limit.

Contracts under pressure

Demand-side questions have piled up alongside the funding debate. Reuters reported Friday, citing Bloomberg, that Microsoft is letting its agreement to lease AI computing capacity from Oracle lapse, declining to renew. Oracle responded that demand for its OCI AI infrastructure remains strong, including for earlier-generation graphics processors. On Wednesday, the Financial Times reported that Oracle had scrapped a multibillion-dollar chip-leasing contract with Tencent after pressure from the Trump administration; Reuters could not independently verify that report, and Oracle, Tencent and the White House did not respond to requests for comment.

Bernstein flags rising risk for speculative projects and heavily indebted providers if the shortage of computing capacity eases — a caveat with direct relevance given Oracle's leverage. The long-term backlog supports the buildout above all when demand and cash receipts keep pace with the investments.

For investors, the yardstick is now the interplay of contract execution, capital needs and free cash flow. Oracle's order book gains weight as an argument for the stock if the company can carry its financing load from operating business as planned; the loss-making status of key customer OpenAI cuts the other way as its infrastructure commitments keep growing. Strong demand is worth most when it forms a dependable base for investment, and growth earns a valuation premium when Oracle demonstrates its financial sustainability at the same time.