NVIDIA GTC: 110 robots on stage
NVIDIA’s GTC demo put 110 robots on stage and pushed agentic AI and confidential GPU workloads — a showcase of how AI, sensors and silicon are converging in real‑time industrial automation. The event underscored a shift toward agentic workflows and raised geopolitical alarms about manufacturing leadership and China’s response. ( )
NVIDIA positioned Vera Rubin as a five‑rack, seven‑chip platform anchored by the NVL72 GPU and dedicated Vera CPU racks to support agentic inference workflows at rack scale. (investor.nvidia.com) A single Vera Rubin POD was described in NVIDIA materials as scaling to as many as 40 racks and 1,152 GPUs with platform numbers cited in press coverage of roughly 60 exaflops for the largest configurations, with first cloud deployments expected in H2 2026. (hashrateindex.com) NVIDIA announced NemoClaw as an enterprise stack that installs Nemotron local models and the OpenShell runtime in one command, adding sandboxing, privacy controls and support for deployment on RTX PCs, DGX Station and DGX Spark. (nvidianews.nvidia.com) Automotive partners announced at GTC include BYD, Geely, Isuzu and Nissan adopting NVIDIA DRIVE Hyperion for Level 4 development, while NVIDIA and Uber committed to launching NVIDIA‑driven robotaxis across 28 cities by 2028, beginning in Los Angeles and San Francisco in the first half of 2027. (investor.nvidia.com) NVIDIA’s robotics slate named ecosystem partners such as ABB, FANUC, KUKA, Figure, Universal Robots and Yaskawa and introduced Isaac/Omniverse simulation updates plus new Isaac “GR00T” open models aimed at robot‑brain developers. (investor.nvidia.com) The company integrated Groq‑derived inference accelerators into the Vera Rubin lineup with the Groq 3 LPU after a reported $20 billion Groq asset/licensing deal, and that transaction has generated regulatory and Congressional scrutiny from senators including Elizabeth Warren and Richard Blumenthal. (tomshardware.com) At GTC Jensen Huang said NVIDIA has received purchase orders for H200 processors for customers in China and is “restarting our manufacturing,” and company commentary at the show also framed Rubin and related stacks as part of a plan to approach $1 trillion in AI‑related revenue by 2027. (bloomberg.com)
