OpenAI signs a $300 billion computing-power contract with Oracle

Background

Training and running large AI models takes enormous amounts of computing: thousands of specialized chips working together inside data centers, drawing power and needing cooling, networking and maintenance. Few AI companies build all of that themselves. Most rent capacity from cloud providers, and for very large workloads they sign multi-year contracts that reserve capacity in advance and set a price for it.

Oracle is a long-established enterprise software and database company that has expanded into cloud infrastructure, selling exactly this kind of rented computing capacity to other companies.

The report

On 10 September 2025 the Wall Street Journal reported exclusively that OpenAI had signed a contract to purchase $300 billion in computing power from Oracle, spread over roughly five years and beginning in 2027^[1][2]. The Journal described it as one of the largest cloud contracts ever signed^[1].

Two things in the report are worth separating. The first is the size of the number: $300 billion is a figure on the scale of national infrastructure budgets rather than ordinary corporate procurement. The second is the timing: the contract starts in 2027, so the money is committed now but the computing capacity and the spending arrive in later years.

Scale against OpenAI's revenue

According to the report, the commitment far outstrips OpenAI's current revenue^[1]. That gap is the part of the story that drew the most attention. A contract of this size is a claim on money the buyer expects to earn in the future; it is signed years before the capacity is used.

Oracle's disclosed contract revenue

In the same week, Oracle disclosed $317 billion in future contract revenue, and people familiar with the matter said most of that figure came from the OpenAI agreement^[3]. Such figures typically represent business already contracted but not yet delivered or billed, and therefore recognized over the life of the contracts rather than at once. In other words, a single agreement with one customer accounted for the bulk of what Oracle had signed up for in the years ahead.

Market reaction

The disclosure moved Oracle's share price sharply higher, and the surging share prices increased the wealth of Oracle chairman Larry Ellison by around $100 billion on the Wednesday of the disclosure^[3]. Because much of Ellison's wealth is held in Oracle stock, a jump in the share price translates directly into a jump in his paper fortune.

Why it matters

The Wall Street Journal said the deal reflected AI data-center spending hitting new highs despite mounting concerns over a potential bubble^[1].

The contract sits at the center of that debate. On one side, long-term commitments give a cloud supplier visibility over future revenue and justify building large data centers, which take years to plan and construct. On the other side, the gap between a $300 billion commitment and the buyer's current revenue raises the question of where the money will come from, and what happens to the supplier if the buyer's growth slows. Because delivery starts in 2027, neither the spending nor its consequences will show up immediately; the contract is a promise about the second half of the decade rather than a record of what has already been paid.