At Y Combinator's Startup School 2026, NVIDIA Founder and CEO Jensen Huang sat down with Garry Tan for a candid conversation about the early days of NVIDIA, near-death pivots, and what it takes to lead a generational company in the age of AI.
Whether you are scaling a startup, running an agency, or leading a team, Jensen's perspective offers a masterclass in transparency, first-principles thinking, and continuous adaptation.
Watch the full talk: Jensen Huang: The Mindset That Built NVIDIA
Here are the key takeaways and direct quotes every builder needs to hear.
1. Confronting Failure: The $5M Sega Lesson
When NVIDIA was founded in 1993, the goal was to turn every PC into a 3D gaming console. By 1995, Jensen realized their underlying algorithm was completely wrong—and none of the 35–40 competing companies knew how to fix it either.
NVIDIA had a $12 million contract with Sega to build the graphics processor for their next-generation console. Facing total failure, Jensen flew to Japan to confess the truth to Sega's CEO, recommending they find another partner.
"What you're telling me is what I contracted you to do you can't do, but you would like all the money on the contract? And I said: 'You got it, that's exactly right.'" — Jensen Huang (08:49)
Appreciating Jensen's honesty, Sega's CEO granted the remaining $5 million payout. That lifeline bought NVIDIA the runway to buy $60 OpenGL textbooks from Fry's Electronics, retrain their engineering team, and rebuild the graphics architecture from scratch (03:42).
Takeaway: Honesty builds trust faster than polished performance. Confronting reality immediately is often the only thing that buys you the runway to survive.
2. Recognizing Universal Function Approximators
When AlexNet emerged in 2012, most observers saw an impressive image-recognition model. Jensen saw a foundational breakthrough across all computing.
"Fifteen years ago I was telling everybody that, hey, guess what—we just learned the universal function approximator. We can give it the answer for almost any function and it could learn what the function is." — Jensen Huang (11:42)
That insight prompted NVIDIA to reorient its entire stack around deep learning and domain acceleration:
- Silicon and chips: Custom GPU architectures optimized for tensor operations.
- Accelerated algorithms: CUDA libraries across physics, molecular dynamics, and neural networks.
- Systems and data centers: High-density interconnects, supercomputing infrastructure, and physical simulation engines.
3. Founder Mode: Build a Car You Can Drive
Addressing how he leads a company with a market cap in the trillions, Jensen shared an analogy for high-growth founders: build an F1 race car that fits your driving style.
"You're building an F1 racer, but you're building it in a way that you can drive. Adapt the car to you... Whatever it takes to fit the car to you, whatever it takes to fit the organization to you, that's what you ought to do." — Jensen Huang (17:37)
Jensen explicitly rejected rigid corporate structures or conventional management playbooks in favor of:
- Deep curiosity: Reaching past middle management to engage directly with researchers and principal scientists.
- First-principles thinking: Stripping complex technical systems down to their basic inputs and constraints.
- Continuous reshaping: Constantly tweaking internal workflows so the CEO acts as a direct force multiplier for the team.
4. Agents and Physical AI: The Next Frontier
Looking ahead at software design, Jensen highlighted two primary shifts shaping the industry.
Controllability in Agentic AI
As AI transitions from static chatbots to autonomous workflows, the primary engineering bottleneck is fine-grained control.
"Controllability is probably the single biggest breakthrough that we need for agents at every single level... I change one word in a plan file and that one word makes a delta difference—not complete difference, but specific difference." — Jensen Huang (22:13)
For founders, this is a product-design warning. Agents will not win only because they are powerful; they will win when customers can steer them predictably, audit their behavior, and make small changes that create precise outcomes.
Physical AI as a $100B Market
Physical AI—software grounded in the laws of physics that interacts with the real world—represents NVIDIA's next massive growth driver (38:50). The deployment pipeline operates as a three-part system:
- Real-to-sim: Capturing real-world environments into digital twins.
- Grounded simulation: Physics-accurate environments, such as Isaac Sim and Cosmos, for safe, rapid agent training.
- Sim-to-real: Deploying post-trained policies via reinforcement learning back onto physical machines.
5. The Ultimate Superpower: "How Hard Can It Be?"
Closing his talk, Jensen addressed the fear and anxiety that often paralyze early-stage entrepreneurs facing rapid technological change:
"This is absolutely the single greatest time to start a company... I always had this feeling: 'How hard can it be?' Truth be told, it is way harder than you think. But you don't want your mind to be there—you want your mind to be 'How hard can it be?' and let the suffering come to you a little bit at a time." — Jensen Huang (46:08)
In an era where artificial intelligence is fundamentally resetting the global computing stack, resilience matters more than having all the answers on day one.
"You don't have to overcome life in one day. You just have to overcome that morning... Just get through today, work towards tomorrow, keep following your dreams, and if you stick with it long enough, NVIDIA happens." — Jensen Huang (48:06)
The mindset is not naive optimism. It is a practical operating system for founders: stay close to reality, reshape the company around how you actually lead, and keep moving long enough for compounding to work.
Related Reading
For more on Jensen Huang's long-term AI worldview, read our breakdown of Jensen Huang on the AI factory era. To explore how agentic systems are changing software, see What Are AI Agents?.
