NVIDIA's GTC 2026 conference in San Jose delivered a cascade of announcements aimed at every layer of the AI stack, from silicon to agentic applications.
The headline hardware reveal was Vera Rubin, NVIDIA's next-generation GPU architecture delivering 10x performance per watt over the current Blackwell generation. Jensen Huang projected at least $1 trillion in combined Blackwell and Vera Rubin revenue through 2027, a figure that reflects the staggering scale of enterprise AI infrastructure investment.
For developers, two software announcements stand out. Dynamo 1.0 is an open-source platform designed as an operating system for AI inference infrastructure, helping companies run AI applications more efficiently at scale. Think of it as the runtime layer that sits between models and the hardware they run on. NemoClaw, built on top of the open-source OpenClaw framework, provides a reference stack for building enterprise-grade agentic AI workers — autonomous agents that can operate 24/7 on defined tasks.
The conference also showcased NVIDIA's push into physical AI. Cosmos world models for robotics and autonomous vehicle development are now available on GitHub, and partnerships with Nissan, BYD, Hyundai, and Uber for Level 4 autonomous driving on NVIDIA's Drive Hyperion platform signal that agentic AI is expanding well beyond software development.
