OpenAI plants Stargate in Norway with 100,000 GPUs by 2026

OpenAI is taking its infrastructure ambitions to the Arctic Circle. The company announced Stargate Norway, its first European data center under the ‘OpenAI for Countries’ program, set to rise in Narvik with 230MW of initial capacity and a planned 100,000 NVIDIA GPUs operational by the end of 2026. An expansion of an additional 290MW is already on the drawing board, making this one of the most significant AI compute buildouts on the continent.

The project brings together an interesting pair of local partners. Nscale, a specialist in AI cloud infrastructure, will design and build the site, while Aker—a Norwegian industrial heavyweight with a century-long track record in energy—will co-own the facility through a 50/50 joint venture. OpenAI comes in as the initial offtaker, with room to scale its consumption over time. The location isn’t accidental: Narvik offers abundant hydropower, cheap energy, and a naturally cool climate that’s practically custom-made for power-hungry GPU clusters.

The facility will run entirely on renewables and use closed-loop, direct-to-chip liquid cooling. There’s a nice secondary benefit baked in too—excess heat from those GPUs will get piped to low-carbon enterprises in the region. That kind of heat reuse is becoming table stakes for responsible large-scale compute, but it’s still good to see it specified upfront rather than bolted on later.

OpenAI says Norwegian startups and researchers will get priority access to the compute, with surplus capacity serving the UK, Nordics, and Northern Europe. The move comes amid a flurry of European chess moves from the company: a UK government MOU, an Estonia deal putting ChatGPT in secondary schools, and interest in the EU’s AI Gigafactories initiative. Norway’s ChatGPT weekly active users have already quadrupled in the past year—most under 35. That’s the user base staring down this new pipeline of compute. The question is whether sovereign AI goals and commercial offtake can coexist without one crowding out the other.