OpenAI NewsSep 16, 2025, 22:30 UTC+8

Introducing Stargate UK

OpenAI, NVIDIA, and Nscale launched Stargate UK, with OpenAI exploring offtake of up to 8,000 GPUs in Q1 2026 for UK sovereign compute. The project may scale to 31,000 GPUs over time for public services, finance, research, and national security use cases that require local jurisdiction. The key detail is local compute for regulated workloads; pricing, full site capacity, and launch timing are not disclosed.

Why it matters: OpenAI extending Stargate to UK sovereign compute with an explicit 8,000-GPU plan for Q1 2026 and a 31,000-GPU ceiling gives it HKR-H/K/R. It stays below 85 because this is an infrastructure partnership announcement, not a shipped model or product, and price, total site scale, or

Read the original ↗Export Markdown

Take

Stargate UK reads like a sales template for OpenAI for Countries, not proof that Britain has secured an AI supercomputer. OpenAI says it will explore offtake of up to 8,000 GPUs in Q1 2026, with a path to 31,000 over time. It names Nscale, Cobalt Park, and the North East AI Growth Zone, but gives no pricing, launch date, site split, or exact NVIDIA GPU SKU.

I don’t buy the full “sovereign compute” framing. The models, API surface, and operating stack still sit inside OpenAI’s system. Local jurisdiction solves procurement and compliance friction; it does not give the UK model sovereignty. Compared with the much louder US Stargate infrastructure story, this looks first like a regulated-market beachhead for public services, finance, research, and national security. 8,000 GPUs is a political signal; 31,000 decides whether it becomes infrastructure.

Source: OpenAI News · openai.com
Published: Sep 16, 2025, 22:30 UTC+8 · Sep 16, 2025