Sweden now has a shared AI supercomputer, and some of the country’s biggest corporate names have a seat at the table. Sferical AI launched Sferical One on October 15, 2026, a machine built on Nvidia hardware and designed for industrial users.
What’s under the hood
The system runs on two Nvidia DGX SuperPODs. Together, they hold 1,152 GPUs spread across 16 Nvidia DGX GB300 systems.
The official launch date is something of a formality. Sferical One has been running since late summer 2026, so the hardware has already had a few months of real workloads behind it.
The roster of backers reads like a guide to Swedish industry. Founding partners include AstraZeneca, Ericsson, Saab, SEB and Wallenberg Investments. SKF has joined as a strategic partner.
That spread covers pharmaceuticals, telecom, defense, finance and manufacturing. The target uses span drug discovery, telecommunications and predictive maintenance on factory equipment.
On the software side, users can build on Nvidia’s open-source Nemotron multimodal models.
From a university stage to a running machine
The project traces back to May 2025. At an event at Linköping University, Marcus Wallenberg of Wallenberg Investments and Nvidia CEO Jensen Huang announced the partnership that set it in motion.
Sferical AI was formally created on August 22, 2025.
Nvidia’s role goes beyond selling the chips. The company supplies the hardware and is also working with partners to set up an AI Technology Center in Sweden.
Huang framed the launch as a boost to Sweden’s history of innovation, saying the shared supercomputer would empower that legacy. Sferical AI CEO Jenny Nordlöw focused on the model itself, stressing the value of cross-industry collaboration and shared infrastructure.
Why sovereign compute is the real story
The phrase doing the heavy lifting here is sovereign infrastructure. Sferical One is pitched as a way to give Swedish companies secure, scalable AI capacity that stays under domestic control.
The initiative’s stated goal is to strengthen Sweden’s standing as a hub for sovereign AI compute.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.