OpenAI, Oracle, SoftBank Unveil $500 Bln Stargate AI Data Centers Across Five U.S. Locations

(RTTNews) – OpenAI, Oracle, and SoftBank announced five new AI data center locations in the United States under Stargate, OpenAI’s large-scale infrastructure platform. These new sites, along with the main campus in Abilene, Texas, and ongoing projects with CoreWeave, will bring Stargate’s planned capacity to nearly 7 gigawatts and total investment to over $400 billion within three years. This progress puts the companies ahead of schedule to reach their $500 billion, 10-gigawatt goal by the end of 2025.

In July, OpenAI and Oracle agreed to build up to 4.5 gigawatts of additional Stargate capacity. This partnership is expected to exceed $300 billion in investment over five years. Three of the new sites are located in Shackelford County, Texas; Doña Ana County, New Mexico; and a Midwest location that will be announced soon. An additional 600 megawatts may be added near the Abilene site. These projects could generate over 25,000 onsite jobs and tens of thousands more across the country. More locations are still being evaluated.

The other two sites, developed by SoftBank and OpenAI, are expected to scale up to 1.5 gigawatts within 18 months. One is in Lordstown, Ohio, where construction has begun on an advanced data center expected to be operational next year. The second is in Milam County, Texas, and will be built with SB Energy, a SoftBank Group company, which is providing the infrastructure for a fast-deployment facility. These sites aim to speed up development, improve scalability, and reduce costs—making powerful computing more accessible.

The five sites were selected from over 300 proposals submitted by more than 30 states since January. This marks the first round of selections, with more U.S. sites expected as the $500 billion investment target is surpassed.

The new Oracle-developed Stargate sites will join the Abilene campus, which is already active on Oracle Cloud Infrastructure. Oracle began delivering NVIDIA GB200 racks in June, and early AI training and inference workloads have already started using this new capacity to support OpenAI’s next-generation research.

The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.