No PriorsNo Priors Ep. 13 | With Jensen Huang, Founder & CEO of NVIDIA
Sarah Guo and Jensen Huang on jensen Huang Reveals NVIDIA’s Long Game Building the AI Era.
In this episode of No Priors, featuring Sarah Guo and Jensen Huang, No Priors Ep. 13 | With Jensen Huang, Founder & CEO of NVIDIA explores jensen Huang Reveals NVIDIA’s Long Game Building the AI Era Jensen Huang traces NVIDIA’s origins from early chip design and accelerated computing bets that most of Silicon Valley initially dismissed, to becoming the core platform for modern AI. He explains how CUDA and architectural consistency, carried on the back of gaming GPUs, enabled NVIDIA to quietly build a developer ecosystem long before deep learning took off. The conversation covers the technical and organizational principles behind NVIDIA’s success, from FP8 and transformer-optimized chips to a bespoke company structure that balances refinement with skunkworks-style exploration. Huang also looks ahead to domain-specific foundation models, robotics, drug discovery, and climate science as the next frontiers for accelerated computing and AI.
At a glance
WHAT IT’S REALLY ABOUT
Jensen Huang Reveals NVIDIA’s Long Game Building the AI Era
- Jensen Huang traces NVIDIA’s origins from early chip design and accelerated computing bets that most of Silicon Valley initially dismissed, to becoming the core platform for modern AI. He explains how CUDA and architectural consistency, carried on the back of gaming GPUs, enabled NVIDIA to quietly build a developer ecosystem long before deep learning took off. The conversation covers the technical and organizational principles behind NVIDIA’s success, from FP8 and transformer-optimized chips to a bespoke company structure that balances refinement with skunkworks-style exploration. Huang also looks ahead to domain-specific foundation models, robotics, drug discovery, and climate science as the next frontiers for accelerated computing and AI.
IDEAS WORTH REMEMBERING
7 ideasPick hard, barely-solvable problems as a wedge for new architectures.
NVIDIA focused on applications that general-purpose CPUs couldn’t handle—graphics, molecular dynamics, seismic processing, then AI—using those “barely possible” problems to justify an accelerated computing architecture and sustain R&D.
Build one stable architecture and stay maniacally compatible over time.
Huang stresses that every NVIDIA chip was made CUDA-compatible even when few used it, sacrificing margins to ensure developers could target a single, stable instruction set—mirroring x86 or ARM and enabling the later AI boom.
Balance generality and specialization to maintain real acceleration.
If accelerators become too general, they turn into CPUs and lose their edge; if too narrow, the market stays too small. NVIDIA’s key skill is walking this line—broadening use cases without giving up outsized speedups.
Go up the stack only as far as customers need—and no further.
NVIDIA builds not just chips but domain-specific libraries (cuDNN, RTX) and, selectively, foundation models, but only to the point where developers and industries can build on top; they avoid becoming a horizontal AI-model provider.
Design your company’s org structure as a bespoke architecture, not a template.
Huang rejects generic corporate org charts; he has ~40 direct reports and combines a highly refined execution engine for shipping complex products with agile, shape-shifting skunkworks to explore 10-year bets.
Cultivate simultaneous conviction and humility as a founder.
He argues entrepreneurs must deeply believe in their core thesis (e.g., accelerated computing) while also believing they might be wrong, staying agile enough to learn, pivot, and endure repeated “face-kicking” setbacks.
Future AI breakthroughs will come from domain-specific foundation models.
Huang expects specialized models for proteins, chemicals, climate, 3D worlds, and robotics to be built on top of NVIDIA’s platform, unlocking drug discovery, Earth-scale climate simulation, and general robotic control from language and video.
WORDS WORTH SAVING
5 quotesWe decided to start a company on accelerated computing… to solve problems that normal computers can’t.
— Jensen Huang
For the very first time in the history of computing, the language of programming a computer is human.
— Jensen Huang
If you are doing something that’s barely possible, you call us.
— Jensen Huang
We try to do as little as we can, as much as necessary.
— Jensen Huang
Ignorance is one of the superpowers of an entrepreneur, and you’ll never get it again.
— Jensen Huang
QUESTIONS ANSWERED IN THIS EPISODE
5 questionsHow far can accelerated computing push beyond CPUs before hitting its own limits, and what might replace GPUs in the long term?
Jensen Huang traces NVIDIA’s origins from early chip design and accelerated computing bets that most of Silicon Valley initially dismissed, to becoming the core platform for modern AI. He explains how CUDA and architectural consistency, carried on the back of gaming GPUs, enabled NVIDIA to quietly build a developer ecosystem long before deep learning took off. The conversation covers the technical and organizational principles behind NVIDIA’s success, from FP8 and transformer-optimized chips to a bespoke company structure that balances refinement with skunkworks-style exploration. Huang also looks ahead to domain-specific foundation models, robotics, drug discovery, and climate science as the next frontiers for accelerated computing and AI.
What tradeoffs did NVIDIA face in keeping every GPU CUDA-compatible despite poor margins, and were there moments they nearly abandoned that strategy?
How should a startup today decide where to draw the line between being a platform provider versus building full-stack applications or models?
What does a “robotic foundation model” realistically look like in five to ten years, and which industries will feel its impact first?
If you were starting NVIDIA in 2025 rather than the 1990s, what would you do differently given today’s AI and hardware landscape?
EVERY SPOKEN WORD
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome