NVIDIA CEO Jensen Huang took the stage at GTC 2026 in San Jose this week, delivering a keynote that outlined the company's vision for the next era of AI infrastructure. The conference, running March 16-19, brought together thousands of developers and researchers both in-person and virtually to explore breakthroughs in agentic AI, physical AI, and what NVIDIA calls 'AI factories'.
The keynote emphasized a shift from traditional AI inference to reasoning systems that can plan, act, and adapt. Huang described AI factories as "scalable intelligence engines that turn data centers into production facilities for intelligence," capable of efficiently producing, managing, and deploying AI across enterprises.
From Inference to Reasoning
According to the official GTC 2026 agenda, a major focus this year centered on agentic AI—specialized AI agents that can reason, plan, and take actions to deliver business results. The conference featured sessions on building these agents using the latest open models, libraries, and NVIDIA's expanding platform ecosystem.
High-performance inference and training sessions demonstrated how organizations can scale model training alongside AI inference, ensuring demanding reasoning models, agents, and robotics workloads run faster and more cost-effectively in production environments.
Physical AI Takes Center Stage
Physical AI emerged as another cornerstone theme, with NVIDIA demonstrating how open models, simulation frameworks, and libraries enable developers to build next-generation factories, robots, and autonomous vehicles. Sessions covered everything from warehouse automation to self-driving car development.
Featured speakers included Pras Velagapudi (CTO of Agility Robotics), Mira Murati (Founder and CEO of Thinking Machines Lab), Jeff Dean (Chief Scientist at Google DeepMind), and Chelsea Finn (Assistant Professor at Stanford and Co-Founder of Physical Intelligence).
Open Models and Developer Tools
GTC 2026 highlighted NVIDIA's commitment to open models, providing developers with customizable AI systems they can deeply integrate into agent architectures. The conference offered hands-on training through NVIDIA's instructor-led courses, with participants able to earn technical certifications validating their expertise in AI, data science, and accelerated computing.
The company also showcased advances in CUDA, CUDA-X libraries, and developer tools for building, optimizing, and deploying GPU-accelerated applications. These tools aim to make it easier for organizations to adopt AI at scale while maintaining performance and cost efficiency.
What This Means for AI Development
The conference signals a maturation of AI from experimental models to production-ready systems that can reason and act autonomously. For enterprises, the AI factory concept offers a roadmap for transforming existing data center infrastructure into scalable AI production environments.
Attendees can now access all GTC 2026 sessions on demand, with recordings covering topics from quantum computing integration to AI for scientific discovery. The next NVIDIA GTC events will continue exploring how accelerated computing and AI transform industries from healthcare to climate science.
