OpenAI is doubling down on its massive AI infrastructure ambitions under the $500 billion Stargate initiative, announcing five new data center sites in partnership with Oracle and SoftBank. The goal? To build a distributed network of AI supercomputing power across the U.S., helping meet surging global demand for generative AI capabilities.
As part of the latest announcement, OpenAI CEO Sam Altman unveiled a bold new vision: to eventually scale AI infrastructure development at the rate of one gigawatt per week — a level of computing capacity that, if realized, could transform industries ranging from healthcare to education.
Five New Sites, 7GW Capacity, and a Vision to Hit $500B
OpenAI’s latest expansion adds five new locations: three sites in Shackelford County (Texas), Doña Ana County (New Mexico), and an undisclosed Midwest location in collaboration with Oracle, and two more in Lordstown (Ohio) and Milam County (Texas) in partnership with SoftBank.
Together, these sites are expected to contribute nearly 7 gigawatts of AI computing capacity, moving the Stargate project closer to its 10-gigawatt target. OpenAI says the full $500 billion investment goal will be reached by the end of 2025, ahead of schedule.
This builds on an earlier $300 billion agreement between OpenAI and Oracle signed in July to build up to 4.5GW of capacity across multiple sites, with Oracle overseeing development and deployment on its cloud infrastructure.
AI at Scale: A “Compute Factory” for the Future
In a personal blog post released alongside the announcement, Sam Altman likened the project to creating a “compute factory” — a repeatable system capable of producing 1 GW of AI infrastructure per week.
“Maybe with 10 gigawatts of compute, AI can figure out how to cure cancer. Or provide customized tutoring to every student on Earth,” wrote Altman, signaling his belief in the transformative potential of compute-powered AGI.
The rollout timeline is ambitious: the first gigawatt of compute from this new buildout is scheduled to be deployed in late 2026 on NVIDIA’s new Vera Rubin platform — as part of a separate $100 billion hardware deal signed between OpenAI and NVIDIA.
More Than Just Power: Industry-Wide Implications
This infrastructure isn’t just about powering OpenAI’s own models. It’s also intended to support a broader ecosystem of AI developers and startups who rely on high-density compute environments.
The Stargate project is already being hailed as the largest private AI infrastructure initiative in the world. Analysts say that reaching this scale could consolidate OpenAI’s leadership in foundation model development, while also raising questions around vendor dependencies and equitable access to compute resources.
SoftBank and Oracle are expected to continue providing both financing and operational support, while Microsoft, MGX, and ARM remain strategic ecosystem partners.
