OpenAI and Oracle's Stargate data center in Abilene, Texas, will eventually house more than 450,000 Nvidia GB200 GPUs, Larry Ellison has revealed.

Oracle founder, chairman, and CTO Ellison delivered the news during a keynote at the Oracle AI World event in Las Vegas on Tuesday. The Stargate data center will use 1.2GW, "enough power for one million four-bedroom homes in the United States," the billionaire said.

Ellison added: "That's a pretty good-sized city."

This seemingly confirms that OpenAI and Oracle will take up all the capacity at the campus, which is being constructed by Crusoe and has a stated target capacity of 1.2GW. The first two buildings are already operational, having launched in September 2025. The remaining six buildings are expected to be completed by mid-2026.

Previously, in March 2025, it was said that the site would have 64,000 GB200s by the end of 2026. Since then, OpenAI has signed a non-binding letter of intent to purchase 10GW-worth of hardware from Nvidia, which in turn is promising to invest $100 billion in OpenAI as the new systems are deployed.

As part of the keynote, delegates were shown a pre-recorded video of the project, which noted that the power will be provided by a combination of grid energy and natural gas turbines. It further stated that the GPUs will be housed in eight separate buildings that will all be connected together to support a single cluster.

The increased scale of the GPU deployment was hinted at when Oracle unveiled its Zettascale10 AI supercomputer. The cloud-based supercomputer, expected to be available in the second half of 2026, will connect hundreds of thousands of Nvidia GPUs across multiple data centers. Oracle noted that Zettascale10 was the "fabric underpinning the flagship supercluster built in collaboration with OpenAI in Abilene, Texas, as part of Stargate."