Alice Recoque, shaping (post) Exascale services in Europe toward the convergence of HP
Speakers: Eric Schnepf (ISC Group) · Gerhard Wellein (Friedrich-Alexander-University Erlangen-Nuremberg (FAU))
Session summary
In this session GENCI's Director of Technology presents the status and roadmap of Alice Recoque, the second EuroHPC exascale system, procured by the Jules Verne consortium of France, the Netherlands, and Greece with an investment of close to half a billion euros in total cost of ownership over five years. GENCI, the French HPC agency, supported over 2,400 research projects last year across HPC, AI, and quantum, with a rapidly growing share of AI and AI-for-science workloads, and operates three quantum computers of different architectures attached to its national systems. As one of the EuroHPC AI Factories, GENCI is addressing compute, data, and skills across twelve application verticals including health, materials, robotics, and dual-use cyber and defense. Alice Recoque is framed as a post-exascale system: it will be hosted by CEA at TGCC, whose site has been upgraded to 24 megawatts with expanded warm-water cooling, and built by Bull on a new platform generation. The unified partition combines AMD CPUs with next-generation AMD MI430 GPUs and supports multi-tenant partitioning of nodes for simultaneous scalar and accelerated workloads, alongside a standalone scalar partition with around 100,000 cores of European SiPearl processors and a European interconnect. The three French quantum computers will be re-attached to the machine, making it the first exascale system accelerated by quantum devices, and the design targets on-the-fly data ingestion from large-scale instruments to support digital twins. Installation is planned at the end of next year, with an application support team already preparing user communities.
Topics: eurohpc exascale systems · european processor technology · quantum accelerators · ai factories · digital twins · data center infrastructure
AI-generated summary of an auto-generated transcript (~2,043 words in full). Details may be imprecise — verify against the session recording.
Auto-generated captions from ISC 2026 session recordings · transcription errors likely, verify quotes against the video · timestamps are offsets into each recording · independent tool, not affiliated with ISC · a Radixia Labs experiment
