NVIDIA's GTC 2026 developer conference opens March 16 in San Jose, with CEO Jensen Huang scheduled to deliver a keynote at 11 a.m. PT from the SAP Center. The event is expected to draw 30,000 attendees from 190 countries and will be streamed live for free online.

"Several New Chips the World Has Never Seen Before"

Huang has been teasing hardware that goes beyond NVIDIA's current Vera Rubin platform. In pre-conference remarks, he promised "several new chips the world has never seen before" — a hint analysts are interpreting as the first public preview of Feynman, the next-generation architecture designed specifically for agentic AI workloads. Feynman is expected to address the reasoning and long-term memory requirements that current inference chips struggle with.

The Vera Rubin platform, featuring HBM4 memory and custom Armv9 CPU cores, entered production earlier this year and is already shipping to hyperscalers. Feynman would be the follow-on generation.

Agentic AI Takes Center Stage

Beyond silicon, GTC 2026 is shaping up as NVIDIA's formal entrance into the AI agent software ecosystem. The company previously revealed plans for NemoClaw, an open-source enterprise agent platform being pitched to Salesforce, Cisco, Google, Adobe, and CrowdStrike.

GTC attendees can also experience Build-a-Claw — a hands-on event in GTC Park where participants deploy a personalized AI agent using OpenClaw, the fast-growing open-source agent framework that Jensen Huang recently called "the most important software release probably ever."

Open Models Panel

On March 18, Huang will moderate a panel on open models alongside LangChain CEO Harrison Chase and leaders from A16Z, AI2, and Thinking Machines Lab — a conversation expected to address whether open-weight models can hold ground against closed frontier labs.

GTC 2026 runs March 16–19. The keynote streams free at nvidia.com.