Nscale Turns Power into Margin

Nscale is trying to win AI cloud by turning power cost and facility design into product margin. Most GPU clouds buy chips, rent space, and pass through high electricity and lease costs. Nscale instead controls more of the physical stack, from hydro powered sites in Norway to liquid cooled racks and prefabricated builds, so it can keep more gross profit while still pricing reserved clusters and token based inference aggressively.

• This matters because electricity and real estate are not side costs in AI infrastructure. Stargate Norway is planned around 100,000 NVIDIA GPUs, 230MW initially, and hydro power in Narvik with below average regional electricity prices. That makes cheap power part of the product, not just procurement.
  2 nscale 3 openai
• The closest comparable is Crusoe, which also built around structurally cheap energy instead of just cheaper software. The difference is that Crusoe grew from stranded gas and modular power systems, while Nscale is leaning into sovereign European capacity, hydro power, and compliance sensitive workloads.
  4 nscale 5 sacra
• Nscale is also using partnerships to fill the machines faster. Singtel gives it a resale and distribution path in Southeast Asia, while Open Innovation AI adds a roadmap targeting 30,000 GPUs for Middle East enterprise demand. Anchor contracts like Microsoft then help justify building larger sites upfront.
  1 sacra 6 singtel 7 globenewswire 8 prnewswire

The next phase is a race to turn infrastructure ownership into a repeatable playbook across regions. If Nscale keeps pairing cheap power, pre sold capacity, and API compatible services, it can look less like a commodity GPU landlord and more like a new hyperscaler built specifically for AI workloads.