Nscale raises $790 million

AI infrastructure is getting very physical. Not abstract. Not “just more GPUs.” Power, cooling, land, grid access, and bank debt are now the real gating items. That’s why Nscale’s new $790 million financing for its Narvik campus in northern Norway matters — it shows the AI buildout is moving from venture-story mode into heavy-infrastructure mode. ### What actually happened? Nscale said on May 11, 2026 that it secured an additional $790 million in financing for its AI data-center project in Narvik, Norway. The lenders are ABN AMRO, DNB, Eksfin, Nordea, and SEB. The company says the money supports continued buildout of the site and strengthens what it calls the largest AI infrastructure investment in Norway. (nscale.com) ### Why Narvik? Narvik is attractive for the boring reasons that end up mattering most. Northern Norway offers abundant renewable power, a cool climate that helps with heat management, and room to build at scale. For AI data centers, that combination is gold — dense GPU clusters draw huge amounts of electricity and dump out huge amounts of heat. If you can make power cheaper and cooling easier, the economics improve fast. (nscale.com) ### How big is this buildout? The key number is 115 MW. That’s the extra expansion capacity tied to this financing structure. Narvik was already planned as a 230 MW campus, and the package includes an additional uncommitted accordion feature of another $790 million that could fund that further 115 MW expansion. In plain English — this is not money for a boutique AI cluster. It is financing for utility-scale compute infrastructure. (nscale.com) ### Who is the capacity for? This is where the story gets more concrete. Nscale had previously pitched the Norway project as part of “Stargate Norway” with Aker and OpenAI. But more recent reporting says Microsoft has signed on for the Narvik capacity, including more than 30,000 Nvidia Rubin GPUs, after Nscale expanded its Microsoft agreement in April. So the campus looks less like speculative capacity and more like pre-sold industrial AI infrastructure. (datacenterdynamics.com) ### Why use bank financing here? Because this asset now looks like infrastructure, not a startup experiment. Banks lend when they think the project has durable collateral, credible customers, and predictable cash flows. A big AI campus with contracted demand from hyperscalers starts to resemble a power plant or telecom network more than a software company. That shift matters — it means AI expansion can increasingly tap debt markets instead of relying only on equity rounds. (nscale.com) ### Why does Europe care so much? Europe has worried for years about falling behind the U.S. on large-scale AI compute. Building in Norway offers a different pitch — sovereign or regional capacity, renewable-heavy power, and less dependence on a handful of U.S. campuses. Nscale has also been assembling a broader European footprint, including Portugal and Finland, which makes Narvik look like one piece of a continental strategy rather than a one-off bet. (nscale.com) ### What’s the real bottleneck now? Not just chips. The harder problem is stitching together power delivery, cooling systems, financing, permits, and customers on the same timeline. Think of GPUs as the engines and the campus as the airport — buying more engines does nothing if the runway, gates, and fuel lines are not ready. This financing matters because it funds the runway part. (nscale.com) ### Bottom line? Nscale’s raise is a sign that AI infrastructure is maturing into a capital-intensive utility business. The winners won’t just be the companies with access to chips. They’ll be the ones that can line up megawatts, lenders, and anchor customers before everyone else does. (nscale.com 1) (nscale.com 2)