EU Council reaches agreement on supercomputer use for AI development

EU Council agrees on regulation to expand EuroHPC’s supercomputing role, adding AI factories to boost Europe’s AI ecosystem and support startups.

The Council of the European Union announced plans to amend the current regulation that governs the European High Performing Computer Joint Undertaking (EuroHPC), which dictates supercomputer use for artificial intelligence (AI) development. 

On May 24, the EU updated the regulation to add the objective of developing and operating local AI factories. According to the EU, such an undertaking would provide additional infrastructure for AI supercomputing services.

The Commission’s proposal, which initiated this amendment, builds on the foundational work of EuroHPC, an entity established in 2018 to lead European supercomputing initiatives.

Currently, EuroHPC manages nine supercomputers across Europe, which have been developed since the initiative’s onset. The new AI factories proposed by the amendment will further this mission.

Additionally, larger supercomputing capacity for the EU would then allow AI startups in the EU better resources to innovate and train their models.

Willy Borsus, the vice-president of Wallonia and Minister for Economy, Research and Innovation, highlighted the collective effort in achieving this milestone. 

“Unity is the strength of Europe, and together we have achieved great goals. We have seen this with the development of our world-class supercomputers, and now we want to boost the outcomes of these powerful machines through trustworthy AI.”

The Council calls for the EuroHPC Governing Board to establish special access conditions for these AI supercomputers, prioritizing startups and small to medium-sized enterprises.

Related: Political correctness in AI systems is the biggest concern: Elon Musk

Under the new regulation, hosting entities can receive up to 50% of the acquisition and operational costs of AI supercomputers from the Union. Ownership of these machines can be transferred to the hosting entities five years post-acceptance testing.

The AI supercomputers will primarily focus on developing, testing, evaluating, and validating large-scale AI training models and emerging AI applications, thus advancing AI solutions within the Union.

First, the proposed regulation must undergo legal review and then upon approval, it will be published in the Official Journal of the European Union, taking effect twenty days post-publication.

The Commission’s proposed regulation is a part of a broader AI initiative announced by the EU’s president, Ursula von der Leyen, and signifies a strategic move to enhance Europe’s AI capabilities.

The EU has been at the forefront of developing AI policies to be implemented across the region. In March 2024, the EU passed its landmark AI Act, the world’s first set of comprehensive AI regulations aimed at safe and ethical AI development in the region.