NVIDIA's GPU Technology Conference opens March 16 in San Jose, bringing together an estimated 30,000 attendees from 190 countries for what has become the largest annual gathering in AI infrastructure. Jensen Huang will deliver the keynote at 11 a.m. PT at the SAP Center, home of the San Jose Sharks.
This year's conference spans topics from physical AI and AI factories to agentic systems and inference at scale. More than 700 technical sessions are scheduled across ten venues in downtown San Jose. The keynote streams free for virtual attendees.
The central hardware story heading into GTC is the Vera Rubin architecture. NVIDIA has confirmed Vera Rubin is in mass production. The architecture introduces NVIDIA's proprietary Vera CPU, replacing the current Grace CPU, paired with sixth-generation High Bandwidth Memory (HBM4). That combination represents a comprehensive upgrade across compute cores and memory bandwidth compared to the Blackwell generation.
A second architecture, internally codenamed Feynman, is expected to be previewed at the conference. Details remain limited ahead of the keynote, but Huang has told investors and analysts that GTC will feature chips "the world has never seen." Feynman is understood to be the roadmap successor to Vera Rubin, and any disclosure of its specifications would extend NVIDIA's public hardware timeline further than it has gone before.
The broader framing for this year's conference is what Huang has described as the "industrial AI era." In his characterization, AI is no longer a discrete application category but infrastructure: a layer of computing investment that every company and nation will need to build out. GTC 2026 is structured around that thesis, with sessions organized around energy supply, chip manufacturing, data center construction, model development, and industrial deployment.
The pregame session on Monday morning, starting at 8 a.m. PT, features the CEOs of Perplexity, LangChain, Mistral, Skild AI, and OpenEvidence. On Wednesday, March 18, Huang will moderate a panel on open model development.
The business context surrounding GTC is unusually complex. NVIDIA is managing the transition from Blackwell to Vera Rubin while navigating U.S. export controls that constrain sales of advanced GPUs to Chinese customers. Any guidance from Huang on international sales strategy or product differentiation for restricted markets will be closely watched by analysts and investors.
The competitive landscape is also shifting. Custom silicon from hyperscalers, AMD's MI-series, and new entrants are all pressing at the margins of NVIDIA's market position. GTC is the company's most important annual opportunity to reset the technology narrative, and 2026 arrives with more at stake than most prior editions.
Sources: NVIDIA Blog, TradingKey
–
By the Control Plane Editorial Team