Tomorrow is the day. Jensen Huang takes the stage at the SAP Center in San Jose at 11 AM Pacific — and if the pre-announcements are anything to go by, he’s got something big this time.
A Chip That Will ‘Surprise the World’
Those are Huang’s own words. And from a man known for leather-jacket keynotes that routinely move markets, that’s saying something. The strongest rumors point to a new inference chip — hardware specifically optimized for running AI models fast and cheap. Faster, cheaper inference is widely seen as one of the last major bottlenecks to scaling AI applications broadly.
More Than Just Chips
But GTC 2026 is far more than a chip unveiling. With over 30,000 attendees from more than 190 countries, it’s the biggest AI industry gathering of the year. Huang has promised to cover the full stack: chips, software, models, and applications.
Particularly interesting: Nvidia is reportedly investing up to $26 billion in open-source models. The previously announced NemoClaw platform for AI agents should take more concrete shape on Monday. And with the Rubin GPU and Vera CPU, two entirely new architectures are on the table.
Why This Matters
Nvidia is long past being just a chip maker. The company is building the infrastructure the entire AI industry runs on. When Huang says ‘AI is no longer a single breakthrough — it is essential infrastructure,’ he’s describing Nvidia’s own role pretty accurately.
The keynote will stream for free at nvidia.com. If you care about the future of AI infrastructure, tune in Monday at 11 AM PT.
Sources: