FINLAND LEADS, REGION FOLLOWS The next 18 months are the build-out window for Nordic AI infrastructure. Finland’s 3 E Network has locked 2027-2028 as peak construction years. Norway’s Oslo AI Factory is already zoned, permitted, and powered on 10,500 m². Microsoft’s Stargate Norway broke ground in 2025 with a $6.2 billion commitment. Sweden and Denmark are finalising joint feasibility studies for shared compute clusters, expected to publish by Q1 2027. WHAT IS HAPPENING NOW Right now, 79% of Nordic organisations report efficiency gains from AI, yet only 18% see revenue growth. The gap is compute. Deloitte’s 2026 survey shows 75% of firms expect AI-driven revenue within two years, but current GPU supply is the bottleneck. NVIDIA’s Vera Rubin NVL72, launched this month, delivers 2.4x higher inference throughput than the previous generation. That throughput scales linearly with hardware, but only if the metal is in place. Aixia reports GPU server orders up 40% quarter-on-quarter across Scandinavia, with lead times stretching to 16 weeks. WHY IT MATTERS The Nordics are positioning as Europe’s AI backbone. Renewable energy surpluses, cold climates, and existing hyperscale footprints make the region the lowest-cost location for large-scale inference. The joint feasibility study, once published, will set the rules for cross-border data flows, power contracts, and GPU allocation. Builders who secure capacity now will lock in pricing before the 2027-2028 demand spike. Those who wait will face higher costs and longer queues. ACTION THIS WEEK Map your GPU pipeline. Identify the single inference workload that, if scaled today, would move the revenue needle. Benchmark it on Vera Rubin NVL72 or equivalent hardware. If the throughput gain justifies the cost, place the order before the next quarterly price adjustment. Use the lead time to refactor your deployment stack for linear scaling. The metal is here; the window is open.

researched · 6 sources

16 SepInfra & computereaches nearby