The New CLARA Supercomputer for AI

New CLARA Supercomputer for AI: A Specialized GPU Cluster for Cutting-Edge Brain Research

21.7.2026

Researchers from the Czech Institute of Informatics, Robotics, and Cybernetics (CIIRC) at the Czech Technical University in Prague, in collaboration with the INDRC research center, are launching a next-generation computing center, which is a key component of the new joint CLARA center.

What’s this about?

The center brings together leading Czech and international partners and focuses on brain research using artificial intelligence and quantum computing. At its core is a high-performance GPU cluster optimized for training AI models and processing large-scale data, integrated into the CLARA testbed. The infrastructure supplied is based on Hewlett Packard Enterprise (HPE) technology and the solution is tailored to the most data-intensive projects in Europe.

Research teams at CIIRC CTU and INDRC are addressing the growing demands for processing large volumes of data, particularly biological, image, and clinical data. This data is crucial for the development of advanced biomarkers, the training of AI models, and the simulation of the behavior of biological structures at the molecular level. The existing computing infrastructure has been unable to meet these requirements in terms of performance, scalability, or operational reliability. Thanks to the CLARA project and funding from the Ministry of Education, Youth, and Sports (MŠMT) under the Jan Amos Komenský Operational Program (OP JAK), it is now possible to acquire new state-of-the-art technologies.

“We needed a stable, high-performance, and sustainable platform capable of supporting interdisciplinary research on an international scale. The CLARA cluster is the technological foundation upon which we will build the next phases of research in the fields of AI, biomolecular interaction modeling, and neuroscience,” said Josef Šivic, a leading Czech scientist in the fields of artificial intelligence and computer vision, who is responsible for the CLARA project at CIIRC, CTU.

CLARA Computing Cluster

Following a public tender, we partnered with Hewlett Packard Enterprise to deliver the CLARA computing cluster as a fully integrated turnkey solution. The goal was to design a computing infrastructure that combines high performance with efficient management and the potential for future expansion—while also meeting the demanding requirements of scientific research in the fields of artificial intelligence, quantum simulations, and neuroscience—the areas on which the new CLARA center focuses.

“For customers like CIIRC and INDRC, it was essential to provide not only key computing power but also a stable, easy-to-manage, and efficiently scalable platform. Our solution combines the best of GPU computing, storage, and support—with an emphasis on long-term operational sustainability,” said Vojtěch Šubrt, Sales Manager at our company.

The computing power of the entire system is based on the latest NVIDIA H200 and L40S GPU accelerators, which significantly speed up AI model training, quantum simulations, and the processing of large datasets. The total computing power of the GPU accelerators in the computing cluster reaches 260 PFLOPS*, or 260,000,000,000,000,000 floating-point operations per second. A total of 9 TB of GPU memory is available.

The solution is built on a combination of HPE enterprise products and an open-source software layer based on the Warewulf, SLURM, and Lustre storage platforms. The Warewulf + SLURM cluster is unique thanks to the efficient deployment and management of compute nodes (Warewulf) combined with a powerful and scalable GPU job scheduler (SLURM). The Lustre parallel file system provides throughput of hundreds of GB/s; all data is accessible from all nodes, and—in combination with Warewulf—it works perfectly even on diskless nodes, as the client boots from a central OS image. In terms of price-performance ratio, this is an ideal solution for scientific organizations for which system robustness and maximum performance are paramount. Thanks to the combination with an open-source software layer, experts can tailor the system’s performance to their specific types of tasks.

Scientific Applications and International Impact

The CLARA cluster in Prague’s Dejvice district will serve as the backbone computing infrastructure for the new CLARA center and, together with partner sites in Ostrava at VŠB-TUO and in Brno at ICRC, will create a unique testing environment (the so-called CLARA testbed) for processing large-scale data on patients with Alzheimer’s disease, simulating protein structures, and developing neural networks for predicting cognitive disorders. At the same time, it will also be used in the context of international collaboration, particularly with the Paris Brain Institute in Paris and the Leibniz Supercomputing Centre in Munich.

Thanks to this new infrastructure, both CTU and the Czech Republic are gaining a tool that will significantly accelerate the development of new scientific models, enable the efficient sharing of computing capacity, and support interdisciplinary projects across Europe. The platform is open to further development and ready to tackle challenges in areas such as personalized medicine, biotechnology, and cognitive research.

The CLARA infrastructure thus represents more than just a technical solution—it is a strategic investment in cutting-edge science with a direct impact on our understanding of the human brain, the diagnosis of neurodegenerative diseases, and the future of medicine as a whole. CIIRC CTU is a scientific and research institute at the Czech Technical University in Prague focused on cutting-edge research and development in the fields of artificial intelligence, robotics, automation, and cybernetics. INDRC is an international nonprofit institute based in Prague that coordinates the CLARA Center of Excellence. CLARA (Center for Artificial Intelligence and Quantum Computing in System Brain Research), with a budget of 44.4 million euros from EU and Ministry of Education, Youth and Sports (MŠMT) funds, contributes to the development of diagnostics and treatments for neurodegenerative diseases. It is the first interdisciplinary center of its kind in Central and Eastern Europe dedicated to research on neurodegeneration using AI, machine learning, and quantum computing. It brings together leading research institutions in the Czech Republic, France, and Germany.

Cluster Specifications

KEY SPECIFICATIONS
6 high-performance servers with the latest-generation AMD EPYC Turin CPUs
Each with 8 NVIDIA H200 GPUs in SXM form factor
GPUs interconnected via NVIDIA NVLink
6 general-purpose servers with the latest-generation Intel Xeon 6 CPUs
Each with 8 NVIDIA L40S GPUs in PCIe form factor
HPE Clusterstor E1000 storage, 500 TB on SSDs for fast data access and 3 PB on spinning disks for archiving purposes
High-speed 400 Gb/s InfiniBand NDR network
Power consumption up to 110 kW, air-cooled

A Partnership That Drives Performance

M Computers is an HPE Gold Partner for compute-accelerated systems. Thanks to this status, we have access to Hewlett Packard Enterprise’s cutting-edge technologies and direct technical support. It is precisely this collaboration that underpins the robust and scalable CLARA computing cluster solution.

COMPUTING POWER: 260 PFLOPS*

GPU MEMORY: 9 TB