OpenAI and Oracle have entered into a binding agreement to significantly expand the Stargate network by adding 4.5 gigawatts of new, power-dense data center capacity. This move propels the overall capacity of the network past the 5 GW milestone, cementing Stargate´s trajectory to accommodate more than two million Artificial Intelligence chips in the near future. Oracle will serve as the primary financier and operator for this particular phase, with SoftBank not contributing funds to this expansion despite being featured in earlier announcements. The expanded capacity will be connected with the existing Stargate I site in Abilene, Texas, where the first racks of NVIDIA GB200 hardware were recently installed. Training for OpenAI´s most advanced frontier models is already underway at this flagship site.
Construction for the new data center footprint is a nationwide endeavor, with labor sourced from more than twenty states and a projected jobs boost exceeding 100,000 positions over four years. These roles will span a wide array of specializations, including electricians, mechanical technicians, and indirect manufacturing positions. The expansion arrives at a pivotal moment, following recent media reports characterizing Stargate as delayed or at risk. OpenAI CEO Sam Altman has publicly refuted such claims, underscoring that overall investments into the project will substantially surpass previous commitments, and that the target of deploying over two million high-end compute chips remains within reach for this year.
The exact location for this sizeable 4.5 GW addition remains undecided, with options including Texas, Michigan, Wisconsin, and Wyoming still on the table. Oracle´s commitment is further underscored by a substantial order for additional NVIDIA silicon, intended to populate the new data halls. A critical hurdle for the expansion is electric power availability, as the planned capacity demands an amount of energy comparable to what 3.5 million US homes consume. In response, negotiations are ongoing with utilities to secure dedicated transmission and on-site generation solutions. The Abilene campus, with its H-shaped halls each planned to accommodate 50,000 NVIDIA GB200 units, is at the forefront of this new generation of infrastructure, serving as a prototype for liquid cooling and modular power systems that will be replicated in future Stargate deployments.
