NVIDIA’s GTC 2026 (often nicknamed the "Super Bowl of AI") kicked off yesterday, March 16, 2026, at the SAP Center in San Jose.
CEO Jensen Huang’s keynote lasted nearly three hours and pivoted the company from being a "chip maker" to an "AI infrastructure and factory operator." Here is a summary of the major announcements:
1. Hardware: The "Vera Rubin" Era
NVIDIA officially unveiled the successor to the Blackwell architecture, named Vera Rubin (after the astronomer who discovered dark matter).
* Vera Rubin GPUs: Designed to handle the massive shift toward AI Inference. The new NVL72 architecture reportedly offers 35x more performance per watt than previous systems.
* Vera CPU: A new, high-performance CPU designed specifically for "Agentic AI" (AI that can reason and act independently), boasting twice the efficiency of traditional CPUs.
* Vera Rubin Space-1: A bold project to bring NVIDIA data centers into outer space to extend accelerated computing beyond Earth.
2. Software & AI Agents
* NemoClaw: A new "agentic AI" platform that allows developers to build autonomous AI agents (or "claws") that can perform complex tasks, browse for info, and manage workflows with enterprise-grade security.
* Nemotron Coalition: NVIDIA is rallying partners like Perplexity, Mistral, and Google to scale out the Nemotron-4 family of open frontier models.
* DLSS 5: The next generation of AI-driven graphics for gaming, focusing on "neural rendering" to further bridge the gap between AI-generated and real-time visuals.
3. Robotics & Physical AI
* Disney Partnership: In the most viral moment of the keynote, Jensen was joined on stage by a robotic Olaf (from Frozen). The robot was trained entirely inside the NVIDIA Omniverse simulator.
* Autonomous Machines: New foundation models like Cosmos 3 were released to help robots and self-driving cars navigate the real world more naturally.
* Uber Collaboration: NVIDIA announced a partnership with Uber to integrate AI-powered robotaxis into their network starting as early as 2027.
4. Financial Vision
Jensen Huang projected that AI infrastructure demand will lead to a staggering $1 trillion in revenue through 2027. He emphasized that we have reached an "inference inflection point," where AI is moving from being "trained" to being "used" at a global scale.
