Jensen Huang, the CEO of NVIDIA, sat down with a Stanford class and laid out how he thinks. Three of his ideas work for anyone, even if you never touch a line of code.
Jensen Huang runs NVIDIA, the company whose chips power most of today’s AI. In May 2026 he was a guest in Stanford’s CS153 Frontier Systems class, in a session called The Compute Behind Intelligence. Most of it is about hardware, but buried inside is a way of thinking that has nothing to do with chips.
Here’s the number that frames it. He says NVIDIA made computing about a million times faster over the last 10 years by designing the chips, software and systems together, which he calls co-design. Shrinking chips the old Moore’s Law way would have delivered only a small fraction of that. That jump is a big part of why AI suddenly feels like it’s everywhere.
You don’t need to understand chips to use the rest. Each idea below comes with a prompt so you can apply it to your own life today.
When NVIDIA designed its Hopper systems for training AI, the biggest supercomputer ever sold cost about $350 million. NVIDIA decided to build systems worth billions anyway.
In His Words
“You would have precisely zero customers.”
He called it building for "a marketplace of zero." There was no demand the day they started, but he reasoned from first principles that it was coming, and he calls Hopper "a great call."
Most of us do the opposite and chase what’s hot right now. By the time we show up, the crowd is already there. So ask where things are clearly headed, and start learning that skill now while it still feels lonely to be early.
A student asked about compute and pointed out that universities like Stanford don’t have a budget for a billion-dollar supercomputer. Jensen asked whose fault that was, and answered himself: "Stanford’s." Then he explained why he said it.
In His Words
“When somebody is at fault, you empower them to solve it.”
He went on to say that calling something "absolutely your fault" also empowers you to solve it. For Stanford, that meant changing how it budgets and building shared, campus-wide supercomputers, the same way it once built its linear accelerator.
The second you say "this isn’t my fault," you also give away the power to fix it. Blaming your boss or the algorithm feels good for a minute, but it hands the steering wheel to someone else. Owning it isn’t about guilt. It’s about keeping control.
Jensen pushed back hard on the loudest AI doom claims. He rejected the idea that nobody knows how these systems work. He also rejected the idea that AI will become infinitely powerful in a nanosecond and take over the world with no way to defend against it.
In His Words
“These things are all being made up.”
He added that this kind of fear even harms the students in the room, who are hoping people still care about computers when they graduate. You don’t have to agree with him on every risk to take the useful part. Fear that keeps you from even trying the tools has a real cost.
Here’s the test I use when a scary headline freezes me. Could any evidence prove this claim wrong? If nothing could, it’s a story, not a fact. Then I do the small brave thing and open the tool to try one real task.
One more number for scale. He said the energy computing will need is probably about a thousand times more than we have today. He isn’t planning for the world as it is. He’s planning for the one that’s coming.
Remember This
Build for where things are headed. Own your situation so you keep the power to fix it. And don’t let fear keep you on the sidelines.