NVIDIA to Hold GTC 2026 March 16-19, With AI Infrastructure and Robotics on Agenda

NVIDIA announced that it will hold GTC 2026 in San Jose, California, March 16-19. More than 30,000 people from more than 190 countries are expected to attend, and the event will feature more than 1,000 sessions. CEO Jensen Huang's keynote will cover the company's work in areas including AI computing infrastructure, AI factories, open models, agent systems and physical AI.

NVIDIA says GTC will address AI across five layers—energy, chips, infrastructure, models and applications—and include robotics and physical AI in its program. For robot developers, the event offers a chance to follow technology trends spanning computing infrastructure and models as well as hardware, but the announcement does not provide performance figures or deployment records for individual products.

In addition to the expected attendance of more than 30,000 and more than 1,000 sessions, the announcement centers on presenting AI across five layers, from energy to applications. Rather than spotlighting a single AI model or robot product, the conference framework lays out a broader supply-side picture connecting computing resources, infrastructure, models and applications. Huang also described AI as essential infrastructure, but that is NVIDIA's positioning, not evidence of the performance of any particular technology or of its deployment record. Robotics and physical AI are among the subjects covered in more than 1,000 sessions, alongside AI factories, large-scale inference and digital twins. In the announcement, robots are therefore treated not as standalone machines but as a topic spanning computing, models and use in real-world settings. However, the company has not specified robot performance, fleet sizes or the conditions for real-world testing. A topic being featured at a conference should be distinguished from proof that a product or system is effective in the field. Our coverage of the product announcement in Humanoid.Guide introduces Geek+'s wheeled humanoid Gino 1 illustrates the need to distinguish descriptions of individual robots from performance validation in real-world settings. By contrast, the GTC announcement focuses less on robot specifications and more on bringing multiple technology areas, including robotics, together at one event. More than 240 startup exhibitors and more than 150 research posters also indicate the breadth of the venue for presentations and exhibits, but are not evidence of commercial deployment or performance. The industry implication is that NVIDIA is bringing computing infrastructure, model developers, application companies and robotics-related firms together, organizing a venue of its own that connects the layers of AI. The focus ahead will be on how computing resources connect with robot control and operations in sessions and exhibits on physical AI, and under what conditions performance and safety are demonstrated in real-world settings. The announcement does not establish specific deployment cases, operating rates, how training data is handled or evaluation methods.

The keynote is scheduled for 11 a.m. PT on March 16 at SAP Center, with an online stream also planned. NVIDIA says it will present AI across five layers—energy, chips, infrastructure, models and applications—and address how the layers connect through sessions on robotics, large-scale inference and digital twins. More than 240 NVIDIA Inception startups will exhibit their technology. The event will also feature nine all-day workshops, more than 60 hands-on labs and more than 150 poster presentations. It will span 10 venues across San Jose.