Nscale inks massive AI infrastructure deal with Microsoft (techcrunch.com)

Nscale has struck a major infrastructure deal with Microsoft to deploy roughly 200,000 Nvidia GB300 GPUs across four data centers — three in Europe and one in the U.S. The rollout will be delivered via Nscale’s own operations and a joint venture with investor Aker. About 104,000 GPUs are headed to a Texas site leased by Ionic Digital over the next 12–18 months (Nscale aims to expand that campus to 1.2 GW), 12,600 will arrive at Start Campus in Sines, Portugal beginning Q1 2026, 23,000 are slated for Loughton, UK in 2027, and the remaining ~52,000 will go to Microsoft’s AI campus in Narvik, Norway. For the AI/ML community this is significant on multiple fronts: it dramatically expands high-performance GPU capacity available for training and inference at hyperscale, underscores Microsoft’s strategy of colocating cutting‑edge accelerators with cloud partners, and signals rapid growth in dedicated AI data‑center buildouts that require massive power and cooling investments. The deal also highlights how fast-growing infrastructure startups (Nscale was founded in 2024 and has raised >$1.7B) are becoming critical suppliers in the GPU supply chain alongside larger names — a trend mirrored by recent large procurements from OpenAI and others. Technically, the large deployment of GB300s will enable more parallel model training, larger context models, and lower-latency inference at scale, while forcing further innovation in networking, power, and facility-level efficiency.