Nvidia CEO Jensen Huang is expected to take the stage at a packed hockey arena in San Jose today, March 16, 2026, to kick off the company’s annual GTC conference. Often dubbed the “Woodstock of AI,” this year’s event is seen as a critical moment for the $4.3 trillion chip giant to prove it can maintain its lead in a market that is rapidly shifting from “training” AI to “using” it.

The Next Frontier: “Feynman” and Inference The centerpiece of the keynote is expected to be the reveal of the “Feynman” chip architecture, named after the legendary physicist Richard Feynman. While Nvidia’s current “Rubin” chips dominate the training of massive models, Feynman is designed to conquer the “inference” market—the process of actually running AI applications to answer user queries. This move comes as tech titans like Meta and OpenAI shift their massive spending away from model development and toward serving hundreds of millions of daily users.

Strategic Reinforcement via Groq Investors are also awaiting details on how Nvidia will integrate technology from Groq, the AI inference startup it licensed for $17 billion late last year. Analysts expect Huang to introduce a new line of servers that combine Groq’s ultra-fast “LPU” (Language Processing Unit) tech with Nvidia’s existing software and networking. This partnership is viewed as a defensive play against rivals like Cerebras and Intel, who are challenging Nvidia’s dominance in cost-effective AI operations.

Beyond Chips: “NemoClaw” and Physical AI The conference is also expected to mark a major expansion into software and robotics:

• Agentic AI: Reports suggest Nvidia will launch “NemoClaw,” an open-source platform for building autonomous AI agents. Unlike traditional chatbots, these “agents” can navigate software, manage tasks, and communicate with other AIs to complete complex work for humans.
• Physical AI: Huang is anticipated to showcase breakthroughs in robotics, featuring AI models that power the next generation of humanoid machines and autonomous factories.
• Optical Computing: Following $4 billion in recent investments in laser technology, Nvidia may demonstrate “co-packaged optics,” which use light instead of electricity to move data between chips, drastically reducing power consumption in massive data centers.

The Stakes for 2026 As the world’s most valuable company, Nvidia faces mounting pressure from both competitors and its own customers—many of whom are now building their own custom silicon. By pivoting toward inference-specific hardware and autonomous agent software, Huang aims to transition Nvidia from a “chip company” to the indispensable operating system for the entire AI-driven global economy.