OpenAI Stargate Compute Infrastructure Update

OpenAI has surpassed its goal of securing 10GW of AI infrastructure in the United States by 2029, adding more than 3GW of capacity in the last 90 days alone. This expansion is part of "Stargate," a long-term effort to build the compute foundation necessary to deliver AGI broadly and reliably.

Compute as the Foundation for AI Scaling

OpenAI views compute as the critical input for training better models, improving reliability, lowering costs, and increasing accessibility. The company describes a "AI flywheel" where increased compute enables superior models, which in turn drive higher usage, leading to increased revenue and further reinvestment in infrastructure.

The Stargate Partnership Ecosystem

Because of the scale required for the Intelligence Age, OpenAI is employing a partner-centric approach. Stargate coordinates across a broad ecosystem including:

• Infrastructure and Hardware: Chipmakers, cloud providers, neoclouds, and NVIDIA GB200 systems.
• Energy and Construction: Utilities, energy providers, construction firms, and skilled trades.
• Finance and Operations: Investors and public sector partners.

This collaborative model is designed to reduce execution risk and maintain flexibility as technology and demand evolve.

Community Engagement and Sustainable Development

OpenAI aims for AI infrastructure to provide local benefits, including job creation, educator resources, and workforce readiness. A recent donation to the Port Washington-Saukville Education Foundation in Wisconsin, made alongside Vantage Data Centers and Oracle, exemplifies this community investment strategy.

To ensure high standards of construction and safety, OpenAI is working with North America’s Building Trades Unions (NABTU) and affiliated skilled trades unions to create job pathways for workers and apprentices.

Case Study: The Abilene, Texas Site

The Stargate site in Abilene, Texas, serves as a model for responsible infrastructure growth. A key technical detail is the use of closed-loop cooling rather than traditional evaporative cooling towers. In this system, water is recirculated through sealed pipes rather than consumed.

• Initial Fill: The one-time initial fill for each building requires roughly two Olympic-sized swimming pools of water.
• Ongoing Use: Annual water use at full buildout is expected to be comparable to a medium-sized office building or approximately four average households.

Impact on Model Capabilities: GPT-5.5

Increased compute capacity is directly translating into more capable AI systems. GPT-5.5, OpenAI's latest and smartest model, was trained at the Abilene, Texas site using Oracle Cloud Infrastructure and NVIDIA GB200 systems.

OpenAI states that GPT-5.5 is designed to close the "capability overhang" — the productivity gap between AI power users and non-power users — to ensure economic opportunities remain broad-based.