OpenAI Opens RFP for US AI Infrastructure and Domestic Supply-Chain Expansion

What OpenAI Announced

OpenAI said on January 15, 2026 that it is launching a public RFP process to expand US AI infrastructure and strengthen domestic manufacturing links across the stack. In the company’s framing, this is an execution step to convert long-range compute plans into concrete power, land, construction, and hardware capacity.

The announcement ties directly to Stargate, the initiative OpenAI says it announced with Oracle and SoftBank to invest up to $500 billion in US AI infrastructure over four years. The new RFP is positioned as the pipeline-building mechanism for that scale.

Priority Criteria in the RFP

• 100MW+ sites available in the near term
• Clear ability to expand toward 1GW+ over time
• Participation from data-center operators, utilities, EPC contractors, and architecture/engineering firms
• Preference for US states able to streamline permitting and support infrastructure investment

Manufacturing and Hardware Notes

OpenAI also described upstream supply-chain milestones. It said TSMC has started volume chip production at Arizona Fab 21, while Amkor is building advanced packaging and testing capabilities in Arizona with operations expected in 2027. OpenAI further stated that first AI racks are being assembled with UEC in Houston, and that the first 100,000 NVIDIA GB200 superchips produced and tested in the US are expected by the end of 2026.

Why This Matters

The significance is less about a single model release and more about infrastructure scale-out. The 100MW-to-1GW requirement signals a shift from short-cycle product launches to multi-year industrial planning across energy, permitting, hardware integration, and supply logistics. If executed, this type of procurement framework can shape where AI capacity is physically built and how quickly large clusters come online in the US.

Source: OpenAI announcement