All-In with Chamath, Jason, Sacks & Friedberg

Jensen Huang LIVE: Nvidia's Future, Physical AI, Rise of the Agent, Inference Explosion, AI PR Crisis

March 19, 2026

AI Summary

5 min read

Jensen Huang joined the All-In hosts live at Nvidia's GTC for a discussion on the company's shift from GPUs to full AI factories, the rise of agentic processing, inference economics, and strategic expansions into physical AI, robotics, and biology. He outlined Nvidia's operating principles amid explosive demand, emphasizing disaggregated computing, hard problems, and ecosystem leadership.

AI Factory and Disaggregated Inference

Huang introduced Dynamo as the operating system for AI factories, inspired by the dynamo that powered the last industrial revolution. Central to it is disaggregated inference, which breaks down the complex inference pipeline—now the hardest computing problem—across heterogeneous hardware like GPUs, CPUs, networking processors, BlueField storage processors, and new Grok LPUs. This evolved from acquiring Mellanox and supports agentic workloads involving working memory, long-term memory, tools, and multi-model interactions (e.g., large autoregressive, diffusion, smaller models).

Continue reading the full summary in the app — free to try.

Read Full Summary →

Free • No credit card required

Listen to Audio Summary Open in App

What you'll learn

1 (00:00) **Intro Banter and Grok Purchase** - Hosts joke about preempting show for Jensen; light tease on xAI acquisition
2 (01:31) **Dynamo OS and Disaggregated Inference** - Jensen unveils AI factory OS, evolving from GPU to heterogeneous computing with Mellanox and Grok LPU
3 (04:08) **Shift to Agentic Processing** - From LLMs to agents using memory, tools, storage; diverse models in data centers
4 (05:30) **Three Core AI Computers** - Training/creation, Omniverse evaluation gym, edge robotics (cars, teddy bears, telecom base stations)
5 (07:11) **Inference Explosion and Factory Economics** - Inference now 1M-1B x training; $50B factory yields lowest token cost at 10x throughput
6 (09:29) **CEO Strategy and Decision-Making** - CEO shapes vision from technologists; pursue insanely hard problems tapping NVIDIA superpowers
7 (11:21) **Physical AI and Digital Biology Bets** - Physical AI: $10B/year, inflecting after 10-year build targeting $50T industry