In a significant move that underscores the growing importance of artificial intelligence (AI) and high-performance computing (HPC) in research and innovation, Denmark is set to build one of the world’s most advanced AI supercomputers. Named Gefion, after the Norse goddess of ploughing and abundance, this state-of-the-art facility will be the result of a collaboration between two industry giants: chipmaker Nvidia and pharmaceutical company Novo Nordisk.
The Gefion supercomputer, which will come online later this year, is being built by Atos Group subsidiary Eviden and will be based on the Nvidia DGX SuperPod architecture. Boasting over 1,500 of Nvidia’s powerful H100 Tensor Core GPUs, Gefion is expected to deliver an impressive six exaflops of FP8 AI performance. Additionally, the supercomputer will connect to the Nvidia Cuda Quantum open-source software platform, enabling simulations on hybrid quantum-classical computers.
Hosted by a Digital Realty data center that runs on 100% renewable energy, Gefion will be part of the “Danish Center for AI Innovation.” Novo Nordisk, known for its groundbreaking diabetes and weight-loss drugs like Ozempic and Wegovy, is investing approximately €80 million in the project through the Novo Nordisk Foundation. The Export and Investment Fund of Denmark is also contributing around €8 million to support this ambitious endeavor.
Mads Krogsgaard Thomsen, CEO of the Novo Nordisk Foundation, emphasized the potential impact of this project, stating, “Groundbreaking scientific discoveries are based on data, and AI has now provided us with an unprecedented opportunity to accelerate research within, for example, human, and planetary health.”
During a press briefing, Kimberly Powell, VP of healthcare at Nvidia, further clarified the partnership agreement, saying, “In our collaboration agreement, we’ll be taking all of this generative AI and bring it over to their sovereign AI infrastructure so that [Denmark] can really push into advancing medicine, quantum computing, and social sciences.”
The establishment of Gefion comes at a time when AI and the supercomputers that power its training are becoming increasingly ubiquitous and essential across various domains, from productivity and efficiency gains to military and cybersecurity applications. As a result, national security policymakers are taking notice, as evidenced by recent US export restrictions on hardware used to train AI to China.
Morten Bødskov, Danish Minister of Industry, Business and Financial Affairs, emphasized the geopolitical significance of this project, stating, “In the current geopolitical climate, it is important that we strengthen our strategic positions.”
However, staying at the forefront of HPC and AI training is a challenging task, given the rapid pace of technological advancements. The fact that Gefion, a supercomputer announced today and intended to be fully operational in 2025, will feature the Nvidia H100 GPU, highlights the speed at which the industry is evolving. Just a day after the Gefion announcement, Nvidia, now the world’s third most valuable company, launched its latest AI chip, the Blackwell, which is said to be four times faster than the H100.
Denmark’s new supercomputer will join a growing list of European HPC and AI initiatives. The UK is building an exascale supercomputer in Edinburgh and an AI supercomputer in Bristol, while the joint undertaking EuroHPC is supporting two exascale supercomputers in the EU: Jupiter in Germany and Jules Verne in France, set to come online in 2025.
An exascale supercomputer can perform over one billion billion calculations, or one exaflop, per second. To put this into perspective, one second of work on an exascale computer is equivalent to one calculation every second for 31,688,765,000 years. Currently, the only known exascale supercomputer in the world is Frontier, located at the Oak Ridge Leadership Computing Facility in Tennessee, USA.
While Gefion is said to have a six exaflop AI performance, it is important to note that this refers to its capacity specifically in AI-related tasks. To be considered a true exascale computer, a supercomputer system must be capable of performing at exascale across a wide range of computational tasks, not limited to AI.
As nations continue to invest in HPC and AI capabilities, it is clear that these technologies will play an increasingly critical role in driving innovation, research, and strategic advantage. While Intel and AMD are making strides in closing the gap with Nvidia, the latter appears poised to maintain its dominance in the near future.
Denmark’s investment in the Gefion supercomputer, in collaboration with Nvidia and Novo Nordisk, represents a significant step forward in the country’s ambition to be at the forefront of AI innovation. With its potential to accelerate groundbreaking research in medicine, quantum computing, and social sciences, Gefion is set to make a lasting impact on the global scientific community and beyond.
