Stargate: Oracle Bets $40B on NVIDIA Chips

Oracle is poised to spend about $40 billion to acquire Nvidia’s high-performance chips as part of the Stargate project, of which it serves as the technology partner, according to the Financial Times (FT).

Unveiled in January from the White House, this colossal $500 billion initiative aims to build an AI infrastructure entirely dedicated to the development of OpenAI. It began with the construction of a first data center in Abilene, Texas. When complete, the site will deliver 1.2 gigawatts of power, positioning it among the largest data centers in the world.

The deal envisions Oracle acquiring roughly 400,000 Nvidia GB200 chips—the latest generation of “superchips” designed for training and operating AI systems—before leasing this compute power to OpenAI. The complex, expected to be fully operational by mid-2026, will span eight buildings and is backed by $15 billion in funding raised by its owners, Crusoe and the American investment firm Blue Owl Capital. Oracle has committed to a 15-year lease.

OpenAI’s Diversification Strategy

For OpenAI, this project marks a pivotal step in its effort to reduce reliance on Microsoft. Until now, the creator of ChatGPT depended exclusively on Microsoft for its computing power. A substantial portion of its nearly $14 billion investment has taken the form of credits to cover these needs.

According to the FT, the two companies have agreed to end their exclusive arrangement later this year. Negotiations are ongoing to determine how long Microsoft will retain licensing rights to OpenAI’s models.

The scale of the Texas project rivals Elon Musk’s ambitions, who plans to expand his Colossus data center in Memphis to house around one million Nvidia chips. Musk this week announced that the next phase of “Colossus” would constitute “the first gigawatt AI training supercluster.” Amazon is meanwhile developing a data center in Northern Virginia with a capacity exceeding 1 GW.