Sources: Oracle will spend ~$40B on ~400,000 Nvidia GB200 chips and lease them to OpenAI at its 1.2GW Texas data center, billed as the first US Stargate project
The 1.2 gigawatts infrastructure project will be one of the largest in the world — Roula Khalaf, Editor of the FT …
Context & Ripple Effects
This report puts a concrete hardware-and-site commitment behind the Stargate joint venture announced in January, which paired OpenAI, SoftBank, and Oracle around US AI infrastructure investment. It identifies Oracle not merely as a project participant but as the buyer and lessor of the compute for an initial Texas deployment.
The arrangement matters because it combines Nvidia accelerator procurement, data-center power, and a long-term compute customer in one transaction. That makes Oracle’s cloud expansion dependent on delivering a very large, specialized facility rather than simply adding conventional cloud capacity.
First-order effects
- Oracle takes on roughly $40 billion of accelerator procurement and the task of turning a 1.2GW Texas facility into capacity it can lease to OpenAI.
- OpenAI gains a dedicated path to a large GB200 deployment, while Nvidia receives demand for roughly 400,000 of its chips tied to a named infrastructure project.
Second-order effects
- Oracle’s cloud business becomes more closely tied to a small number of capital-intensive AI deployments; construction, power delivery, and chip availability become immediate execution dependencies.
- The scale of the lease model raises the importance of Stargate’s original infrastructure-financing structure for other cloud providers and AI customers seeking capacity without directly owning the underlying hardware.
Third-order effects
- If replicated, AI compute procurement will increasingly resemble utility development: specialized operators finance power-constrained sites and sell capacity through long-duration customer commitments.
- That structure can concentrate both demand and execution risk among a few chip suppliers, cloud operators, and frontier-model customers, making financing discipline and facility delivery central competitive capabilities.
The trend: This is part of AI infrastructure’s shift from incremental cloud expansion toward financed, power-intensive compute campuses anchored by major model developers.