Europe is embarking on an unprecedented expansion of AI supercomputing capabilities, fueled by NVIDIA’s state-of-the-art hardware and software ecosystem. Thirty-five new all-NVIDIA supercomputers are rolling out across 23 countries, targeting a combined compute power of 800 exaflops. These installations aim to serve more than three million researchers working in academia, industry, and government sectors.

The major deployments include leading research centers and AI factories such as the Barcelona Supercomputing Center's EuroHPC AI Factory, BavariaAI’s Blue Swan, IT4LIA, HLRS’s HammerHAI, and NAISS’s Mimer EuroHPC AI Factory. These systems leverage NVIDIA’s latest accelerator chips—primarily Blackwell and Hopper architectures—as well as the upcoming Vera Rubin platform, designed to meet the demanding AI workloads of next-generation models.

Technical innovations underpinning these deployments include NVIDIA’s Quantum InfiniBand networking, CUDA-X and CUDA-Q libraries for GPU acceleration, NIM microservices, and NVIDIA AI Enterprise software. Together, these components create a full-stack AI platform tailored for the high-throughput, low-latency demands of modern AI research.

Among the flagship projects, Barcelona’s EuroHPC AI Factory plans to expand its MareNostrum 5 supercomputer by integrating NVIDIA GB300 NVL72 and GB200 NVL4 GPU systems interconnected by the Quantum-X800 InfiniBand network. This configuration is designed to deliver about 20 exaflops in AI training and 33 exaflops in AI inference, focusing on applications like generative AI, climate science, biotechnology, sustainable agriculture, clean energy, and governmental AI services.

BavariaAI’s Blue Swan provides 1,000 GPUs through NVIDIA GB200 NVL4 systems paired with Quantum-2 InfiniBand networking at the FAU Erlangen and LRZ supercomputing centers. It will offer 11 exaflops for training and 22 exaflops for inference to accelerate foundation models in scientific research, public administration, robotics, and healthcare.

IT4LIA stands out with a massive deployment exceeding 8,000 GPUs supported by Quantum-X800 InfiniBand and NVIDIA AI Enterprise software. This factory targets 82 exaflops of training and 164 exaflops of inference throughput, reflecting one of the largest European investments in AI compute infrastructure.

Meanwhile, HLRS’s HammerHAI becomes Germany’s first dedicated AI factory, integrating over 850 GPUs with Quantum-X800 InfiniBand to deliver around 8 exaflops for training and 15 exaflops for inference. This facility is set to provide secure AI resources for engineering simulations, large language models, and industrial applications.

These deployments mark a strategic effort by European nations to remain competitive in AI research, enabling breakthroughs across multiple disciplines. The adoption of NVIDIA’s comprehensive AI platform highlights the continent’s commitment to integrating cutting-edge technology into scientific and industrial ecosystems.