Sweden’s Sferical One Shared AI Supercomputer Live

Sferical AI announced on 15 October 2026 the launch of Sferical One, a shared AI supercomputer in Sweden built for AstraZeneca, Ericsson, Saab, SEB and Wallenberg Investments, with SKF as strategic partner.

What the system is

Sferical says the system is an NVIDIA DGX SuperPOD built on 16 NVIDIA DGX GB300 systems, and that it has been running since late summer 2026. Sferical AI will also build services based on NVIDIA Nemotron open multimodal AI models so members can inspect weights, fine-tune for their use cases and run them in environments that meet their security requirements.

Do not treat secondary “1,152 GPU” arithmetic (16 × 72) as a figure from the Sferical release — stick to 16 DGX GB300 systems as stated. No cost is disclosed.

Who uses it for what

Rather than each company building the same stack alone, Sferical pools investment, capacity and expertise. Current areas of exploration named in the release:

AstraZeneca — drug discovery
Ericsson — foundation models for telecommunications
Saab — defence applications
SEB — financial services
SKF — predictive maintenance, quality control and R&D

Jenny Nordlöw, CEO of Sferical AI, said what makes Sferical unique “isn’t just the technology, but that several of Sweden’s largest companies share the same infrastructure, develop solutions together and share knowledge across industries.”

Marcus Wallenberg, chair of Wallenberg Investments, called it a long-term investment in capabilities Swedish industry will need to compete. Jensen Huang, NVIDIA’s founder and CEO, said the shared supercomputer will “bolster even more, mobilizing the researchers, entrepreneurs and scientists to build the Swedish AI ecosystem.”

How Sferical describes the company

Sferical AI was founded by AstraZeneca, Ericsson, Saab, SEB and Wallenberg Investments, with SKF as a strategic partner. Beside the shared supercomputer, the company also runs Sferical Labs, where the partner companies and Sferical jointly develop AI solutions and expertise. The stated aim is to accelerate AI adoption in industry and strengthen the companies’ competitiveness.

Open multimodal models matter for that setup: because Nemotron weights can be inspected and fine-tuned, members can keep sensitive fine-tunes inside environments that meet their security requirements rather than shipping proprietary data to a black-box frontier API. That is the industrial bargain Sferical is selling — shared CapEx, retained control.

Why shared sovereign compute matters

Industrial AI demand is rising faster than many single-firm clusters can justify. A shared SuperPOD in-country, on open Nemotron models members can inspect and fine-tune, is Sweden’s consortium answer: scale without five duplicate CapEx programmes, and more control than renting opaque foreign cloud. It sits in the same European “sovereign / shared industrial AI” lane as other national and corporate stacks TSN has covered.

See Germany’s reported Nvidia industrial-AI cooperation plans, NVIDIA joining the EC’s AI in Science Enablers, and Snowflake–Bleu Orange’s France sovereign AI data cloud.

What’s still unknown

CapEx / OpEx of Sferical One
Exact site details beyond Sweden (secondary outlets have named Linköping; not in the Sferical release used here)
How capacity is scheduled among the five founders and SKF

What this means

Sferical One live for AstraZeneca, Ericsson, Saab, SEB, Wallenberg Investments; SKF strategic partner.
Hardware: NVIDIA DGX SuperPOD, 16 DGX GB300 systems; running since late summer 2026.
Services on Nemotron open multimodal models (Sferical).
No cost in the release.