OpenAI's European Stargate plans shrink as Microsoft and Google take over capacity

Editor's take

OpenAI's ambitious plans for a massive data center in Narvik, Norway, have been significantly scaled back, with Microsoft and Google now slated to absorb a substantial portion of the projected computing capacity. This pivot suggests a pragmatic adjustment rather than a complete abandonment of OpenAI's European expansion goals, likely driven by the immense infrastructure demands and escalating costs associated with training frontier models like GPT-5.

The significance lies in the shifting resource allocation dynamics within the AI industry. The sheer scale of compute required for leading LLMs is pushing even well-funded organizations like OpenAI to collaborate or leverage existing hyperscale infrastructure, a trend that will likely accelerate as model sizes continue to grow. This also implies a greater reliance on cloud providers, potentially concentrating power and influence.

Future developments will hinge on whether this consolidation leads to more efficient, sustainable AI development or creates new bottlenecks. The key question is whether OpenAI can still achieve its performance targets for future models with reduced direct control over its dedicated hardware, and how this impacts its competitive positioning against rivals like Anthropic, which is also investing heavily in its own infrastructure.