What Jensen Huang Said This Week Changes the Math
We told you to watch GTC. Here's what actually happened.
Jensen Huang took the stage Monday at the SAP Center in San Jose and delivered what may be the most consequential tech keynote in years. The headline number: NVIDIA now sees at least $1 trillion in confirmed orders for its current and next-generation chips through 2027. One year ago at GTC 2025, that number was $500 billion. It doubled in twelve months.
To put that in context: the entire global semiconductor industry generates roughly $600 billion in annual revenue. NVIDIA alone is looking at $1 trillion in demand for two product generations.
The Hardware
The centerpiece of Monday's keynote was Vera Rubin — NVIDIA's next-generation AI infrastructure platform. It integrates seven chips into five rack-scale systems that operate as a single unified supercomputer. Vera Rubin is already in full production and shipping to customers this year.
Alongside it, Huang unveiled the first chip from Groq — the company NVIDIA acquired for $20 billion last year. The Groq LPU (Language Processing Unit) is built specifically for inference, the process of running AI models in production rather than training them. It ships in Q3 2026. The architecture is different from a GPU: it's designed for ultra-low-latency token generation, which matters when AI agents are making decisions in real time. Huang's recommended deployment: use Vera Rubin for the heavy lifting, add Groq for the 25% of workloads that demand speed above all else.
This is NVIDIA completing its stack. Training. Inference. Now both covered by proprietary silicon.
The Software
NemoClaw was the software announcement that got the most attention. It's NVIDIA's enterprise-ready reference stack for OpenClaw — the AI agent platform that has gone viral in recent weeks. One command installs it, configures it, and deploys a working AI agent. The focus is security: agents that can work with proprietary data without exposing it.
Huang was direct about why this matters. "The OpenClaw event cannot be understated," he said. NVIDIA is betting that the next wave of AI demand doesn't come from training bigger models — it comes from deploying millions of agents that need to run securely inside enterprise environments. NemoClaw is the on-ramp.
The Physical World
The third major theme was physical AI — robots. Huang said he can't think of a company building robots that isn't working with NVIDIA. The roster on stage included an Olaf robot from Disney, autonomous vehicles from Nissan, BYD, Hyundai, and Geely, and an Uber partnership to deploy NVIDIA-powered robotaxis across 28 cities by 2028.
Huang called it "the ChatGPT moment for autonomous driving." Whether that framing holds up, the capital commitment behind it is real.
What This Means
Last week we closed the GDP thesis: the economy is producing more with less human labor, and the data confirmed it isn't temporary. GTC is the supply side of that same story.
The companies deploying AI aren't slowing down. They're doubling their orders. The infrastructure buildout is accelerating, not plateauing. For investors tracking the AI super-cycle, the question was always how long the spending continues. NVIDIA just gave the clearest answer yet: at least through 2027, and likely beyond.
The picks-and-shovels thesis we published for Pro subscribers holds. The companies controlling the chokepoints — silicon, networking, power, security — are the ones capturing this buildout. GTC just confirmed the buildout is larger than anyone projected a year ago.
MasicotAI — Tracking the intersection of artificial intelligence and economic reality.