NVIDIA unveils 35 European AI supercomputers and ramps Vera Rubin to full production

At ISC 2026, NVIDIA announced 35 new AI and HPC supercomputers being deployed across Europe to serve more than 3 million researchers, a major sovereign-compute expansion on the continent. It also confirmed its next-generation Vera Rubin platform has reached full production, with the new Vera CPU powering scientific systems at Los Alamos National Laboratory.

The company paired the hardware news with software for science, introducing the BioNeMo Agent Toolkit to support agentic workflows in life-sciences research, and reiterated its headline market thesis: roughly $1 trillion in AI infrastructure demand through 2027. Jensen Huang's recurring 'five-layer cake' framing positions AI as a sustained, compounding driver of computing demand rather than a one-off buildout.

The European push matters competitively because sovereign AI — nations and regions wanting domestically controlled compute — has become a key growth vector as the US tightens model-export controls. By seeding 35 systems across Europe, NVIDIA entrenches its accelerators as the default substrate for the region's research base.

The shadow over all this is custom silicon: OpenAI's Jalapeño, plus in-house chips from Google, Amazon, and Meta, are all attempts to erode NVIDIA's pricing power. NVIDIA's co-packaged optics roadmap, positioning TSMC's COUPE technology for the next infrastructure wave, is part of its answer — pushing further up the systems stack so customers stay locked into full NVIDIA platforms. Watch whether the $1T projection survives contact with rising memory-chip costs that are already forcing cloud price hikes.