NVIDIA GTC Berlin Launches Sovereign AI Push Across European Industries

At GTC Berlin 2026, NVIDIA announced a wave of hardware, software, and partnership developments targeting European sovereign AI deployment. Jensen Huang introduced the DSX POD, a pre-integrated containerized AI factory that cuts on-site deployment from six months to 14 days, and the OpenShell Reference Architecture, an open-source policy-controlled runtime with kernel-level sandboxing for autonomous agents. European AI cloud providers Nebius, Nscale, and Verda are building infrastructure on NVIDIA's Vera CPU, while server manufacturers including Bull, Dell, HPE, Lenovo, and Supermicro are building Vera-based systems. Alongside the infrastructure announcements, NVIDIA reported a series of European industry deployments: Revolut built a foundation model called PRAGMA on Blackwell Ultra GPUs serving 80 million customers; Starling Bank trained an agentic model on UK sovereign cloud infrastructure; Stripe reported 42x faster fraud-detection pipelines and 92% cost savings using GPU-accelerated Ray Data in offline testing; and LSEG is working with NVIDIA to accelerate historical tick data analysis up to 10x. In retail, L'Oréal cut ingredient discovery from months to hours using NVIDIA ALCHEMI; Otto Group expanded an autonomous warehouse digital twin to Germany's largest fulfillment center. For weather and climate, NVIDIA launched StormScope EU, a regional European forecasting model, with AXA, Vaisala Xweather, TotalEnergies, and others deploying Earth-2 models. In the public sector, Poland, Germany, and France are each running separate sovereign AI initiatives, and NVIDIA hosted a European Open Models Codefest. NVIDIA also completed what it described as a first commercial phone call on T-Mobile's network with Nokia using CUDA after nearly eight years of work.