Here’s a focused three‑paragraph summary centered on autonomous vehicles from Jensen Huang’s March 2025 GTC keynote:
In his keynote at NVIDIA GTC San Jose (March 18, 2025), Jensen Huang revealed an important new partnership with General Motors, in which NVIDIA will supply both hardware and AI infrastructure to support GM’s forthcoming fleet of autonomous vehicles—spanning manufacturing, automotive design and in‑car systems (YouTube, Investing.com). A key pillar of this collaboration is the introduction of HALO, NVIDIA’s comprehensive automotive safety system. HALO encompasses tools from silicon design to system software, and Huang emphasized that NVIDIA has subjected its safety code (over 7 million lines) to third‑party audits to ensure transparency, explainability and robustness in autonomous systems (Video Highlight).
To power such safety‑critical AV systems, Huang showcased the convergence of NVIDIA’s Blackwell Ultra architecture and Omniverse/Cosmos synthetic simulation platforms. Blackwell Ultra—scheduled to ship in the second half of 2025—delivers massive inference performance for agentic AI workloads within vehicles (AP News). Complementing that compute is NVIDIA’s cosmos‑powered AV system training pipeline, where teacher‑student model distillation leverages synthetic data and simulation to train autonomy models more efficiently than ever, enabling iterative learning and real‑world validation using tools like their physics engine partnerships (Newton, with DeepMind and Disney Research) (Video Highlight).
In sum, Huang framed autonomous vehicles as a flagship domain where NVIDIA’s physical AI—the next stage after generative and agentic AI—is coming alive. By combining high‑performance compute (Blackwell/ Rubin roadmap), safety‑first software frameworks (HALO), and simulated training pipelines (Omniverse/Cosmos/Newton), NVIDIA aims to accelerate development of AVs that can safely reason and act in physical environments. Huang concluded that the late‑2026 “Vera Rubin” chip generation—and later Rubin Ultra—will push AV AI to the next horizon of performance and safety (AP News).
Let me know if you’d like details about other parts of the keynote—such as robotics, data‑center infrastructure, or enterprise AI.