OpenAI has agreed to lease approximately 4.5 gigawatts of computing capacity from Oracle in a deal valued at around $30 billion annually, making it one of the largest cloud contracts for artificial intelligence to date, according to reports from Bloomberg and the Financial Times.
The agreement represents a significant expansion of OpenAI’s “Stargate” data center initiative, launched earlier this year in partnership with SoftBank, aimed at securing the vast computing resources necessary to develop advanced AI models and support growing demand for applications like ChatGPT.
Sources familiar with the matter told Bloomberg that Oracle plans to build multiple data centers across the United States to fulfill the requirements of the Stargate project. The anticipated 4.5GW capacity would amount to roughly a quarter of the total operational data center capacity currently available in the U.S.
OpenAI and SoftBank have previously indicated that the Stargate project could involve investments of up to $500 billion for data center development both in the United States and worldwide.
The joint venture has so far secured roughly $50 billion in funding from its founding partners, which include Oracle and Abu Dhabi’s sovereign wealth fund MGX.
Possible sites under consideration for new data center facilities span several U.S. states, including Texas, Michigan, Wisconsin, Wyoming, New Mexico, Georgia, Ohio, and Pennsylvania, according to sources familiar with the plans. Oracle also intends to expand the 1.2-gigawatt Stargate facility in Abilene, Texas, a project currently being developed and financed by data center startup Crusoe.
Oracle has committed $7 billion to the Stargate venture and is preparing for $25 billion in capital expenditures next year, exceeding prior forecasts.
To support the Abilene data center, Oracle plans to purchase around 400,000 of Nvidia’s high-performance GB200 chips, representing an investment of approximately $40 billion, according to the Financial Times.
The developments also highlight how OpenAI is engaging new cloud providers to secure the computing resources needed for its growing portfolio of AI products.
