FLAP-D colocation now runs 3.6 GW live capacity, double 2019 levels. The Nordics are the only region in Europe where power, land and cooling align at scale for AI workloads. Sweden’s Ruien project is live with 230 MW, targeting 520 MW by year-end. Hardware count: 100,000 NVIDIA GPUs on order. Narvik, Norway, secures $6.2 B in renewable infrastructure from Microsoft, Nscale and Aker. Denmark’s Gefion supercomputer now runs NVIDIA CUDA-Q, enabling hybrid AI-quantum pipelines. Bitzero Holdings lists on Nasdaq, bringing public-market capital to Nordic HPC assets. Nvidia acts as matchmaker, connecting GPU customers directly to Nordic data-center operators. The network hears that d-Matrix will use NVIDIA NVLink Fusion to deploy its Raptor XPUs in Stockholm and Oslo racks, creating a unified scale-up fabric across the region. For builders, this means predictable capacity. No more bidding wars for scarce metal; instead, a single contract with Nvidia or a listed colo partner delivers power, cooling and GPUs in weeks. The bottleneck shifts from hardware to software: orchestrating hybrid workloads across CUDA-Q, Spectrum-X and local NPUs from Nordic Semiconductor. Action this week: map your 2027 compute needs against the 3.6 GW FLAP-D footprint. If your pipeline exceeds 5 MW, lock in a letter of intent with one of the three listed colo providers before the next earnings call. Smaller teams can still move fast: Gefion offers CUDA-Q access via a simple API key, no capital outlay required.

researched · 4 sources

14 SepInfra & computereaches nearby