OpenAI signs $300B cloud computing deal with Oracle, triggering 36% stock surge
Sep 11, 2025
Key Points
- OpenAI signs $300 billion cloud computing contract with Oracle over five years, the largest deal of its kind, sending Oracle stock up 36% and boosting the company's backlog to $317 billion.
- The deal requires 4.5 gigawatts of power capacity and signals deepening alliance between Sam Altman and Larry Ellison to build AI infrastructure as part of the broader $500 billion Stargate project.
- OpenAI must scale revenue to afford $300 billion in compute costs while Oracle faces execution risk delivering capacity on schedule amid supply constraints; Oracle also posted double earnings miss.
Summary
OpenAI signed a $300 billion contract with Oracle over roughly five years to buy cloud computing power. Oracle's stock surged 36% on the announcement, which revealed the company had added $317 billion in future contract revenue during its latest quarter. The deal requires 4.5 gigawatts of power capacity, equivalent to two Hoover dams or the electricity consumed by 4 million homes.
The contract deepens the alliance between Sam Altman and Larry Ellison to build next-generation AI infrastructure. It also clarifies how capital from the Stargate project, the $500 billion investment initiative announced earlier with Trump, Ellison, and Masayoshi Son, is flowing into concrete infrastructure channels.
Historical context
The $300 billion figure fits historical capex spending by dominant internet companies. Amazon excluding AWS spent roughly $250 billion on retail and fulfillment infrastructure to support a trillion-dollar business. Meta spent approximately $130 billion on core AI and generative infrastructure. Google spent an estimated $210 billion on servers and data centers for search, YouTube, Gmail, and related services, the closest comparable since Google lacks warehouse fulfillment costs.
AI inference is far more compute-intensive than serving search results or email. If OpenAI scales to rival Google, Amazon, or Facebook in user reach and engagement, several hundred billion dollars in infrastructure capex becomes rational over the contract period.
Execution challenges
The deal ties Oracle and OpenAI together so tightly that both must execute nearly perfectly. Oracle must develop and deliver the capacity on schedule in a supply-constrained environment where electrical transformers, semiconductors, and real estate are bottlenecks. OpenAI must scale revenue to a level where the company can actually afford to pay for $300 billion in compute over five years. No company has ever raised $300 billion in capital. OpenAI's largest fundraise to date is $40 billion, contingent on its for-profit conversion.
Oracle just posted a double miss on earnings, raising questions about whether the company can execute infrastructure delivery at the scale and speed required. The backlog is real, but converting it to revenue and profit on a compressed timeline is uncertain. Whether Oracle can draw down the backlog into actual revenue next quarter will test whether the $317 billion figure represents genuine near-term obligations or mostly future optionality that needs heavy discounting.