Sovereign AI Infrastructure Expands
- National and cloud players are expanding sovereign AI compute projects, broadening hardware demand maps. - Microsoft added Norway datacenter capacity reportedly with access to about 30,000 NVIDIA Vera Rubin chips. - Canada launched a sovereign compute program with roughly $890 million, creating new buyers for accelerated compute systems and integration work. ( )
Governments and cloud companies are building AI capacity closer to home, with Microsoft adding Norway compute and Canada opening a national supercomputer program. (nscale.com) (ised-isde.canada.ca) In Norway, Nscale said on April 14 that it expanded its Microsoft agreement at its 230-megawatt Narvik campus and will deploy more than 30,000 NVIDIA Rubin graphics processors there in 2027. Bloomberg and CNBC reported the site had been intended for OpenAI’s Stargate project before Microsoft took the capacity. (nscale.com) (bloomberg.com) (cnbc.com) Microsoft’s Norway buildout extends a deal announced in September 2025, when Nscale and Aker said Microsoft had signed for about $6.2 billion of AI compute capacity, with staged deliveries beginning in 2026. Nscale said the latest expansion builds on its previously announced Vera Rubin rollout for Microsoft. (prnewswire.com) (nscale.com) Canada moved on a parallel track on April 16, when Innovation, Science and Economic Development Canada opened applications for the Artificial Intelligence Sovereign Compute Infrastructure Program. The guide says the program can provide up to about C$890 million over seven fiscal years starting in 2026-27 for the design, construction, and operation of a compute system. (ised-isde.canada.ca) (betakit.com) The Canadian program is aimed at a “large-scale, Canadian-owned AI supercomputer” for researchers and Canadian businesses, with applications due by 1 p.m. Eastern on June 1, 2026. That turns Ottawa into a direct buyer of advanced servers, networking gear, power systems, and integration work tied to AI clusters. (betakit.com) (ised-isde.canada.ca) “Sovereign compute” means keeping critical AI infrastructure under domestic ownership or control instead of relying entirely on foreign cloud regions. Canada’s guide requires a Canadian-owned entity to build and run the system, while Norway’s Narvik project adds capacity in Europe under a long-term Microsoft contract. (ised-isde.canada.ca) (nscale.com) The hardware involved is also moving up the stack. NVIDIA said on March 16 that its Vera Rubin platform had entered full production, framing it as the next rack-scale system for training and running large AI models. (investor.nvidia.com) Those projects widen the map of who is ordering accelerated compute. Instead of demand coming only from United States hyperscalers, vendors now have to plan for state-backed procurements, regional cloud campuses, and local operating rules in places like Canada and Norway. (ised-isde.canada.ca) (nscale.com) The next milestones are concrete: Nscale said the added Norway deployment is scheduled for 2027, and Canada’s first application deadline lands on June 1. By then, sovereign AI will be measured less by speeches than by signed power, chip, and construction contracts. (nscale.com) (ised-isde.canada.ca)
