DCAI shifts Gefion supercomputer strategy toward AI inference

The Danish Center for AI Innovation (DCAI) is shifting the strategic direction of its supercomputer, Gefion, to adapt to a rapidly evolving artificial intelligence market. Originally designed for training massive AI models using 1,500 Nvidia H100 GPUs, the facility is now pivoting toward providing capacity for running smaller open-source models, a process known as inference.

Under the leadership of new director Martin Svensson, DCAI aims to make Gefion more accessible and establish an economically sustainable business model. This shift follows a challenging initial period characterized by high complexity, low utilization, and significant financial deficits. The project, funded by the Novo Nordisk Foundation and the Export and Investment Fund of Denmark, represents a major investment in national AI infrastructure.

Parallel to these infrastructure shifts, industry experts warn that the rapid integration of AI into Danish businesses and the public sector has outpaced the debate regarding security, accountability, and transparency. While 96 percent of Nordic organizations plan to increase AI investments, there are growing concerns that safety considerations are being deprioritized in the rush to implement business-critical AI technology.