Europe Launches Largest AI Supercomputing Build in History with 35 NVIDIA Systems, Serving Over 3 Million Researchers

NVIDIA has announced the launch of a record-breaking 35 AI HPC supercomputers across Europe, marking the largest single-year supercomputing expansion in the continent’s history. Once operational, the new infrastructure will provide over three million researchers with next-generation computing power to accelerate AI development, scientific discovery, and industrial innovation across the region.

The 35 systems span national supercomputing centers, AI factories, and university research institutions, all built on NVIDIA’s full-stack AI infrastructure. They will support a wide range of research disciplines including climate science, healthcare, clean energy decarbonization, quantum computing, and fundamental science.

Since last year, Europe has deployed or announced plans for 800 EFLOPS of AI computing power. NVIDIA is backing these efforts with its Quantum InfiniBand networking, CUDA-X libraries, NIM microservices, and AI Enterprise software — delivering a complete platform that covers model training, simulation, inference, and agentic AI workflows.

“AI has become the new instrument of science, and Europe is building the infrastructure to make it accessible to millions of researchers,” said NVIDIA founder and CEO Jensen Huang. “With NVIDIA accelerated computing, researchers can simulate more complex systems, train scientific AI models, and build agentic AI workflows that turn Europe’s data and expertise into breakthroughs that benefit the entire world.”

Among the flagship projects in this latest expansion:

MareNostrum 5 — the first EuroHPC facility purpose-built for AI — will add NVIDIA GB300 NVL72 and GB200 NVL4 systems interconnected via the Quantum-X800 InfiniBand platform. The upgrade delivers approximately 20 EFLOPS of AI training performance and 33 EFLOPS of AI inference, accelerating generative AI, climate modeling, medical and biotech research, sustainable agriculture, energy systems, and government AI services.

Blue Swan will deploy 1,000 GPUs at Friedrich-Alexander University Erlangen-Nürnberg and the Leibniz Supercomputing Centre using NVIDIA GB200 NVL4 systems and Quantum-2 InfiniBand networking. With up to 11 EFLOPS of training and 22 EFLOPS of inference performance, it will support Bavaria’s foundation model initiative, driving open multimodal models for science, public administration, medical research, robotics, and perception technologies.

IT4LIA will be equipped with over 8,000 GPUs based on NVIDIA GB200 NVL4 systems, Quantum-X800 InfiniBand networking, and AI Enterprise software, delivering 82 EFLOPS of AI training and 164 EFLOPS of inference.

HammerHAI, Germany’s first AI factory at the Stuttgart High-Performance Computing Center, will deploy more than 850 GPUs via NVIDIA GB200 NVL4 and Quantum-X800 InfiniBand. It offers approximately 8 EFLOPS of training and 15 EFLOPS of inference for secure computing in engineering simulation, large language model inference, and scientific discovery.

Mimer AI Factory, owned by the EuroHPC Joint Undertaking and hosted at Linköping University in Sweden, will deploy 100 NVIDIA GB200 NVL4 systems totaling 400 GPUs with ConnectX-8 networking. With up to 4 EFLOPS of training and 7 EFLOPS of inference performance, it will serve Sweden’s life sciences, materials research, autonomous systems, trustworthy AI, and data-driven innovation initiatives.

The coordinated rollout reflects Europe’s determination to build sovereign AI capabilities at scale, ensuring that the continent’s researchers have competitive access to the compute infrastructure necessary to tackle the defining scientific and industrial challenges of the coming decade.