At the GTC Berlin Conference, Jensen Huang refrained from announcing any new chips. Instead, he confirmed that the Vera Rubin platform has officially commenced full-scale mass production. His focus was centered on three pivotal areas: the swift establishment of computing power hubs, the governance and oversight of intelligent agents, and the practical deployment of physical AI. NVIDIA introduced its modular AI factory solution, DSX POD, which dramatically cuts down on-site deployment time from roughly six months to a swift 14 days. This solution is complemented by two distinct power and cooling systems, steering the benchmark for AI factories towards the volume of Tokens generated per kilowatt-hour of electricity. Concurrently, NVIDIA unveiled the OpenShell enterprise reference architecture, built upon a full-stack hardware foundation. This architecture facilitates hardware-level dynamic management and control of intelligent agents via an autonomous security framework on the DPU, thereby mitigating the risks associated with software dependency.

Furthermore, NVIDIA is harnessing Europe's robust industrial infrastructure to propel the implementation of physical AI. It has launched technical solutions and collaborative initiatives tailored for scenarios such as autonomous driving, humanoid robotics, and industrial simulation testing. This move is driving AI integration from foundational computing power to tangible, real-world applications.