Funds from pharmaceutical drug maker Novo Nordisk are being utilized to establish among the world's most sophisticated AI supercomputers in Denmark in a cooperation with another current huge winner, chipmaker Nvidia.
Called Gefion, after the Norse goddess of tilling and abundance, and constructed by Atos Group subsidiary Eviden, it will be based upon the Nvidia DGX SuperPod architecture. Gefion will include over 1,500 of Nvidia's H100 Tensor Core GPUs and provide 6 exaflops of FP8 AI efficiency. It will link to the Nvidia Cuda Quantum open-source software application platform, which enables for simulations on hybrid quantum-classical computer systems.
The supercomputer will come from the “Danish Center for AI Innovation,” and be hosted by a Digital Realty information centre, which utilizes 100% renewable resource. The Novo Nordisk Foundation, managing owner of Novo Nordisk– maker of diabetes-turned-weight-loss drug Ozempic/Wegovy– is investing around EUR80mn in the task. The Export and Investment Fund of Denmark is likewise contributing around EUR8mn.
“Groundbreaking clinical discoveries are based upon information, and AI has actually now supplied us with an extraordinary chance to speed up research study within, for instance, human, and planetary health,” stated Mads Krogsgaard Thomsen, CEO of the Novo Nordisk Foundation in a declaration.
Clarifying the collaboration arrangement in a press instruction, Kimberly Powell, VP of health care at Nvidia, stated “In our cooperation contract, we'll be taking all of this generative AI and bring it over to their sovereign AI facilities so that [Denmark] can actually press into advancing medication, quantum computing, and social sciences.”
The geopolitics of AI
The
The most recent rumblings from the EU tech scene, a story from our sensible ol' creator Boris, and some doubtful AI art. It's totally free, each week, in your inbox. Register now!
AI and the supercomputers that power its training is no longer the domain of a couple of choose scientists. As the innovation ends up being ever more common, the calculate required to train bigger and bigger AI designs or run advanced simulations boosts.
Its significance varies in between whatever from efficiency and performance gains, to military and cybersecurity applications– something that is not passing undetected by nationwide security policymakers, as highlighted by United States export constraints on hardware utilized to train AI to China.
“In the present geopolitical environment, it is necessary that we reinforce our tactical positions,” Morten Bødskov, Danish Minister of Industry, Business and Financial Affairs, mentioned (albeit staying away from the expression of the day, “digital sovereignty”).
Gefion will link to quantum simulation through the Nvida Cuda-Q software application. Credit: Nvidia
Remaining on top of the high-performance computing (HPC) and AI training video game is a hard job. The truth that Gefion, a supercomputer revealed today and meant to be completely functional in 2025, will include the Nvidia H100 GPU is a testimony to simply how rapidly things relocate the world of HPC and expert system.
Simply the other day Nvidia,
