In a landmark deal that underscores the escalating investment in artificial intelligence infrastructure, Oracle has committed to spending around $40 billion on Nvidia's cutting-edge GB200 Blackwell chips to power OpenAI's new data center in Abilene, Texas.
The massive facility, currently under construction, is the first site of the U.S. Stargate Project – an ambitious joint venture between OpenAI, Oracle, SoftBank, and Abu Dhabi's MGX that aims to invest up to $500 billion in AI infrastructure across the United States by 2029. The project was formally announced in January 2025 by President Donald Trump, with SoftBank CEO Masayoshi Son serving as the venture's chairman.
According to sources familiar with the matter, Oracle will purchase approximately 400,000 of Nvidia's most powerful GB200 chips and lease the computing power to OpenAI. The Abilene data center, spanning 875 acres with eight buildings, will require a staggering 1.2 gigawatts of electricity when fully operational. Oracle has signed a 15-year lease for the site, which is expected to be completed by mid-2026.
The deal represents a strategic shift for OpenAI, which has primarily relied on Microsoft's data centers for its computing needs. Earlier this year, the two companies terminated their exclusivity agreement after OpenAI's demand for computing power exceeded Microsoft's supply capabilities. For Oracle, the partnership presents an opportunity to strengthen its position in the cloud computing market against competitors like Microsoft, Amazon, and Google.
The Abilene facility is being developed by Crusoe Energy Systems in partnership with Blue Owl Capital, which together have raised approximately $15 billion in debt and equity financing. JPMorgan has provided a significant portion of the debt financing, with two loans totaling $9.6 billion.
Beyond the U.S., OpenAI, Oracle, and Nvidia are also collaborating on a Stargate project in the United Arab Emirates, where a new AI data center will be constructed in Abu Dhabi. The UAE facility is expected to utilize over 100,000 Nvidia chips, with its first phase scheduled to come online in 2026.