Enabling Europe’s Sovereign AI Ecosystem: From HPC Systems to Hybrid AI and Quantum Co
Speakers: Sarra Refai (Jülich supercomputing center) · Ana Marija Sokovic (University of Illinois at Chicago)
Session summary
Crispin Keable, senior solution architect at Bull, presents the company's strategy for delivering sovereign European computing across HPC, AI, and quantum. Now a standalone entity after its period under Atos, with the French sovereign wealth fund as sole shareholder, Bull operates in over thirty countries and positions itself as a full-stack provider spanning silicon design, systems, software, and services. Keable highlights the expansion of the Angers factory in France, expected to build four exascale systems simultaneously by 2027, and a recently signed agreement with Foxconn to manufacture undifferentiated AI systems such as NVL72-class machines in Europe, starting with a production line in the Czech Republic, while differentiated systems remain in Bull's own factory. The portfolio includes the Bull Sequana XH3500 direct-liquid-cooled flagship, supporting up to 264 kilowatts per rack and the fastest current GPUs; the BXI (Bull Exascale Interconnect) Ethernet-based high-performance network going live at the end of the year; the Smart Management Center software stack; and Jarvis for cloud-like system presentation. He notes Bull-designed systems held the top three Green500 positions that week and cites JUPITER's energy efficiency. In quantum, Bull provides the Qaptiva software environment for simulation and hybrid QPU-accelerated applications in partnership with hardware vendors rather than building QPUs itself. Around 300 data scientists develop ready-to-go AI use cases aligned with government priorities such as health and transport, reflecting the view that governments are the primary customers for large sovereign AI investments.
Topics: european digital sovereignty · exascale system manufacturing · high-performance interconnects · direct liquid cooling · quantum software ecosystems · ai use cases for government
AI-generated summary of an auto-generated transcript (~2,005 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
