According to NVIDIA's live blog from GTC Berlin, CEO Jensen Huang opened the keynote by tying the company's European push to a broader industrial shift. He cited European firms including L'Oréal, Revolut, Spotify, Merck, Vodafone, Zara, Porsche, Nestlé and HSBC as AI collaborators, and said NVIDIA has committed nearly $10 billion to Europe's technology ecosystem through more than 12,000 partners. Huang described AI as a stack spanning energy, accelerated computing, AI factories, models and applications, arguing that infrastructure must exist before models emerge.
Huang announced DSX POD, a reference design for containerized Vera Rubin AI factories that partners integrate and validate before shipping. NVIDIA said onsite deployment can drop from six months to 14 days once site utilities are ready. The company also said its DSX AI Factory Platform can reduce cooling water use by up to 100% and cut the power needed for the same AI output by up to 30%. On safety, Huang introduced the OpenShell Reference Architecture for Safe AI Agents, which uses policy-controlled runtime, kernel-level sandboxing and dynamically adjusted privileges to secure autonomous agents.
For enterprise adoption, Huang framed agents as systems with models, orchestration harnesses, sandboxes, knowledge and tools, and introduced the Agent Knowledge Foundation for connecting them to unstructured documents and structured databases. NVIDIA also highlighted open models across language, robotics, autonomous vehicles and physics AI, arguing that companies can post-train them on proprietary data to create what Huang called sovereign intelligence. The source is a single NVIDIA live blog, so there are no independent or contrasting accounts to compare.