GTC Berlin 2026: Live Updates From NVIDIA CEO Jensen Huang’s Keynote
NVIDIA Blog Brian Caulfield ● Covered by 2 sources
Jensen Huang used GTC Berlin to pitch Europe as an AI factory. He said NVIDIA is pouring nearly $10 billion into the region and pushing agents, robotics and physical AI.
Based on reporting by NVIDIA Blog, Brian Caulfield — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
NVIDIA turned the Tempodrom’s old-event glamour into a stage for a very different sort of show: Jensen Huang’s pitch that Europe should stop thinking of AI as something it buys and start building it. GTC Berlin runs for three days, with more than 100 sessions on agentic AI, robotics, physical AI, open models and accelerated computing taking place at STATION-Berlin. The keynote drew journalists, analysts and partners, but Huang kept returning to a single point: Europe can be part of the AI economy if it builds the infrastructure first.
He backed that argument with a list of European customers and a bigger spending claim. L’Oréal, Revolut, Spotify, Merck, Vodafone, Zara, Porsche, Nestlé and HSBC were all name-checked as companies working with NVIDIA on AI. Huang also said the company has spent the week moving across Paris, London, Stockholm and Dublin, while committing nearly $10 billion to Europe’s technology ecosystem and working with more than 12,000 partners. His framing was simple: AI is not just models. It is energy, computers, an AI factory, tokens and applications stacked on top.
That “factory” idea ran through the rest of the talk. Huang described NVIDIA’s stack as hardware, system software, CUDA-X libraries and models, then pointed to Vera Rubin and the new DSX POD reference design for partner-built AI factories inside modular containers. The deployment pitch was aggressive: once utilities are in place, onsite rollout can shrink from six months to 14 days. NVIDIA also said its DSX AI Factory Platform can cut cooling water use by up to 100% and deliver the same AI output with up to 30% less power.
The keynote then shifted from infrastructure to the agents running on it. Huang said agents think, reason, iterate and use tools, which means they need tight security, minimal privileges and constant monitoring. NVIDIA’s answer is OpenShell Reference Architecture for Safe AI Agents, built on an open source, policy-controlled runtime with kernel-level sandboxing. He followed that with open models such as Nemotron, Cosmos, GR00T, Alpamayo, Ising and Earth-2, plus the Agent Knowledge Foundation for connecting agents to company data and CUDA-X tools for the work itself.
From there, the examples spread across industries: Lila Sciences, Siemens’ Erlangen factory, Dassault Systèmes, telecom networks, adaptive radiotherapy with Siemens Healthineers, autonomous vehicles and a new reference humanoid called Agile ONE SN. Huang closed with RTX Spark laptops, which he compared to the DGX-1 he delivered to OpenAI in 2016, and a comic Eurovision-style send-off featuring four Microducks trained in simulation. The message was heavy on spectacle, but not subtle: NVIDIA wants Europe building the plumbing, the models and the robots.
My take — AI-written commentary, not fact-checked reporting
NVIDIA’s favorite trick is to make infrastructure sound inevitable, then sell the picks and shovels. Europe should listen carefully here, because “sovereign intelligence” is just the latest polite phrase for not handing the keys to everyone else’s platform. The real test is whether local companies build on these tools without confusing dependency for strategy.
Read more about this at: NVIDIA Blog
