Oracle's massive AI infrastructure push, capped by a $300 billion agreement with OpenAI in September 2025, has produced

Oracle’s rapid AI data center expansion and a September 2025 agreement to provide services tied to OpenAI have left the company carrying more than $160 billion in liabilities, a development that has rattled markets and investors. The $300 billion OpenAI deal briefly sent Oracle shares up roughly 36% in a single day and made Larry Ellison the world’s richest man for a short time; since then the stock has fallen more than 43%, wiping out those gains and sharpening scrutiny of the company’s financing.

Over the past year Oracle has built large — scale AI compute capacity, including the Stargate data center campus in Abilene, Texas, designed to supply computing power to model builders and enterprises. Company comments and regulatory filings characterize the effort as a broader push to sell AI infrastructure — data centers, power and interconnects — to customers such as OpenAI, positioning Oracle as a provider of the physical stack beneath large language models and other generative AI systems.

Investors and asset managers have framed Oracle’s strategy through a “picks and shovels” lens: firms including Blackstone and others see data centers, chips and electricity as lower‑risk ways to play the AI wave compared with the models themselves. NVIDIA and its CEO Jensen Huang are commonly cited as leading infrastructure plays; some market participants have begun to view Oracle in a similar category because of its emphasis on supplying compute and related services.

That strategic pivot carries a clear balance‑sheet cost. Public filings and third‑party studies indicate Oracle now reports more than $160 billion in outstanding liabilities. JPMorgan research, as reported by Barron’s, attributes roughly $133 billion of that total specifically to the AI buildout. Oracle disclosed under $40 billion in cash on hand in its most recent SEC filing and stated it is burning cash, intensifying concerns about the company’s near‑term liquidity needs amid the ongoing capital intensity of its data center program.

The company’s revenue and contract mix deepen the exposure: of roughly $553 billion in remaining performance obligations, more than $300 billion is attributable to OpenAI, and OpenAI itself has reportedly been losing billions. Because OpenAI remains private, some investors treat Oracle as a public proxy for OpenAI’s prospective success, a concentration that shows up in market signals — credit default swaps have surged and Oracle’s debt rating sits just above junk.

Oracle’s debt‑to‑equity ratio of about 415% is far higher than hyperscalers such as Meta and Amazon, none of which top 80%, underscoring that Oracle’s financing profile and reliance on very large contracted obligations are atypical among major data‑center operators. That concentrated exposure leaves lenders, rating agencies and shareholders particularly sensitive to OpenAI’s financial trajectory and Oracle’s cash burn, making near‑term liquidity and the performance of contracted AI services the key variables investors will be watching.