Nvidia GTC 2026: Vera Rubin, Physical AI, and the Next Wave of Chips

What happened

Nvidia's GTC 2026 conference ran 16-19 March in San Jose, drawing one of its largest crowds ever. CEO Jensen Huang delivered the keynote on Monday, outlining the company's vision for the next 12 months of AI development.

Key announcements included:

• Vera Rubin GPU and CPU architecture -- the next generation of Nvidia hardware, focused on AI inference and high-throughput workloads
• NVIDIA NemoClaw -- a new platform for building and deploying AI agents in enterprise environments
• Physical AI push -- Nvidia's Cosmos world models and Alpamayo models for autonomous vehicle development are now available open-source on GitHub and Foundry
• Groq LPU integration -- a partnership with Groq on language processing units for faster inference
• DLSS 5 -- the latest AI upscaling technology for gaming and creative applications

The conference featured 1,000+ sessions and 450 sponsors. Huang's central theme: the "token" -- the basic unit of modern AI -- is now powering systems from scientific discovery to physical robots operating in the real world.

What this means for your business

Most UK businesses will not buy Vera Rubin GPUs. But Nvidia's direction matters because it drives what AI services become cheaper and faster over the next 12-18 months.

The key signal from GTC 2026: inference is the priority now. Nvidia is building hardware specifically to run AI models faster and cheaper at scale, not just to train them. That means the AI tools you use -- ChatGPT, Claude, Gemini, Copilot -- are going to get faster and more affordable as cloud providers upgrade their infrastructure.

Physical AI (robots, autonomous vehicles, smart warehouses) is the next frontier Nvidia is betting on. If your business operates in logistics, manufacturing, or retail, pay attention -- the automation wave is moving from software into hardware.