DSX, the CUDA Wheel, and Europe's AI Stack at GTC Berlin
Huang's factory half: DSX from a digital twin, Vera Rubin pods, and a slide that stacks applications, models, factories, NVIDIA compute and energy.
Workstation team9 نومبر، 20261 min read
Before agents, Huang's GTC Berlin keynote was the factory. NVIDIA DSX is the full-stack design: compute, network, storage, software, power and cooling, starting from a digital twin. DSX Pod is the cabled Vera Rubin building block. The slides around it were a wheel of markets on one CUDA architecture, and a stack called Building Europe's AI Economy.
What we could read off the slides
The Europe stack runs applications, models, AI factories, NVIDIA compute, and energy. The wheel's centre is a full-stack AI factory, with cloud providers, AI labs, neoclouds, OEMs, and enterprise industries around it. The DSX platform slide shows Vera Rubin NVL72, NVLink 6, BlueField-4 compute and storage, Spectrum-6, a rack labelled Groq 3 LPX, and Vera RTX compute.
Huang said that at 2 gigawatts, DSX can save 54 million litres of water a day, and that Vera Rubin is another 30 times Blackwell in two years. The newsroom's tighter figure is 30 times the throughput per megawatt on agent workloads. Those are NVIDIA's numbers.
A factory is not the only place he put OpenShell. RTX Spark, DGX Spark and DGX Station are the edge of the same story. The long read, including the company names we could actually read, is DSX and Europe's AI economy.