Jensen Huang does not do small keynotes. But CES 2026 was something else entirely.
Vera Rubin platform: 5x more powerful than Blackwell, shipping H2 2026. Nemotron 3: open models designed specifically for agentic AI. Cosmos: a world simulator for training robots and autonomous vehicles. DLSS 4.5: neural rendering that is getting uncomfortably close to making traditional GPUs irrelevant for gaming.
And the line that will define 2026: “The ChatGPT moment for physical AI is here.”
Is he right?
ChatGPT’s “moment” was not when the technology was ready. It was when ordinary people could suddenly USE it. One week: 1 million users. Two months: 100 million. The technology existed before. The moment was access.
For physical AI (robots, autonomous vehicles, embodied agents), the technology is advancing fast. Boston Dynamics Atlas is on factory floors. Waymo is expanding to 11 cities. Humanoid robots are playing tennis.
But can a regular person buy a robot that does useful things in their home? No. Can a small business deploy an autonomous agent that handles physical tasks? Not yet. Can a developer build a physical AI application as easily as they can build a chatbot? Not even close.
Jensen is right that the foundation is being laid. He might be premature on the “moment” part. ChatGPT’s moment was a consumer product. Physical AI does not have its consumer product yet.
The numbers that do matter
NVIDIA announced over $1 trillion in pipeline commitments at CES. One. Trillion. Dollars. That is not hype. That is orders from companies that have done the math and decided the bet is worth making.
Mercedes-Benz announced the CLA as the first car running NVIDIA’s full autonomous stack. Not a concept car. A production vehicle.
The open-source Nemotron 3 models and Alpamayo reasoning models for autonomous vehicles signal that NVIDIA wants an ecosystem, not just hardware sales. Give away the software, sell the chips. The razors-and-blades model at unprecedented scale.
What I took away
Jensen is a showman and his keynotes are designed to move stock prices. But behind the theatrics, the concrete announcements are real: production hardware, production vehicles, production robots, open models.
Physical AI is not having its ChatGPT moment yet. But it is having its “GPT-3 paper” moment. The capabilities are being demonstrated. The infrastructure is being built. The ecosystem is forming.
The consumer moment comes next. And when it does, NVIDIA will be selling the shovels.
[Draft: Awaiting Carlos’s twist]