Denmark is standing up a sovereign AI stack on the Gefion supercomputer, giving public-sector workloads a domestic, EU-compliant alternative to hyperscale clouds.
Gefion, run by DCAI, now hosts Netcompany’s EASLEY AI platform, effectively turning a national HPC system into shared AI infrastructure for government, healthcare, education, and other regulated domains.
The core AI infra signal is data residency and control: all training and inference stay in European data centers under EU data protection and audit rules, which directly targets public-sector resistance to non-EU GPU clouds.
This is also an operational consolidation move, with Netcompany planning to migrate all its AI services onto Gefion, standardizing on a single European GPU backend for both public and private customers.
The economic claims (DKK 55 billion in Danish productivity and EUR 100 billion across the EU) are aggressive but show the political pressure to justify continued investment in sovereign compute and power-hungry data centers.
Gefion is positioned as a reference architecture for national AI supercomputers: shared GPUs, strong compliance boundaries, and a platform layer (EASLEY) that lets agencies build assistants without touching non-EU infra.
The piece is worth reading for how Denmark is turning a national supercomputer into multi-tenant, production-grade AI infrastructure and as a template for other EU states thinking about sovereign GPU footprints.
Source: Denmark launches sovereign AI on Gefion to boost public sector
