Key Moments
Jensen Huang: The Mindset That Built NVIDIA
Y CombinatorWant to know something specific about what's covered?
We've already dissected every moment. Ask and we will deliver (with timestamps).
Key Moments
NVIDIA began with fundamentally flawed technology but a willingness to learn, leading to breakthroughs in computing; resilience and a 'how hard can it be?' mindset are crucial for founders.
Key Insights
NVIDIA's initial technology for reinventing 3D graphics was fundamentally wrong, leading to a realization in 1995 that required buying three textbooks on OpenGL to learn the correct approach.
The core idea that propelled NVIDIA was the ability to augment CPUs with accelerators to solve problems too difficult for general-purpose computers, a principle applicable to domains like molecular dynamics, image processing, and deep learning.
A pivotal moment for NVIDIA's survival was a $5 million contract with Sega for the Dreamcast project; Jensen Huang was honest about their flawed technology, enabling them to retain the funds and continue operating.
The realization that deep learning, as exemplified by AlexNet, is a universal function approximator was a significant breakthrough, enabling NVIDIA to envision applications in computer vision, robotics, and self-driving cars.
NVIDIA's success is attributed to a unique perspective on accelerated computing, focusing on algorithm acceleration rather than just chip design, and a belief in deeply held visions that are hard to pursue.
The future emphasis is on systems thinking, as most low-level tasks will be automated agentically; understanding system design, constraints, and information flow will be critical.
AI and automation are seen as job creators, not destroyers, by automating tasks within jobs, thereby increasing productivity and enabling new roles and industries with higher backlogs.
Physical AI, driven by generative AI's ability to articulate motion, is a key frontier, with self-driving cars being the first economically significant application, with a projected market of $10 billion and potential for a future $100 billion business.
Embracing failure and learning from textbooks
Jensen Huang recounts NVIDIA's challenging start in 1993, where their initial strategy to revolutionize 3D graphics for PCs was based on a fundamentally flawed technological approach. This realization, occurring around 1995, was a critical juncture, especially with numerous competitors in the market. Huang emphasized the importance of confronting this reality head-on, stating that the company would not survive if they didn't pivot. Ironically, the team lacked the knowledge to implement the correct approach. This led Huang to purchase three crucial textbooks on OpenGL and pipeline design from Fry's Electronics. This act of acquiring knowledge, rather than already possessing it, laid the foundation for NVIDIA to become a leader in modern computer graphics, highlighting a core philosophy: technology changes constantly, and the ability to confront reality and learn is more important than the initial technology itself. The lesson learned is that if a task is important, NVIDIA will learn the necessary technology, even if it proves much harder than initially anticipated.
The philosophy of augmenting general-purpose computing
The foundational idea that drove NVIDIA's inception was the conviction that general-purpose computers (CPUs) could be significantly enhanced by accelerators to tackle problems otherwise deemed too complex. While 3D graphics was their initial target, this principle of augmentation proved applicable to a much broader range of algorithmic domains. These include molecular dynamics, image processing, inverse physics, and most significantly, deep learning. The company's core realization was not merely about building a superior chip, but about accelerating specific algorithm domains. This perspective is central to NVIDIA's enduring success, allowing them to adapt and innovate across diverse fields by focusing on the computational challenges presented by emerging algorithms and problem sets.
A pivotal moment: honesty and survival with Sega
A critical moment that nearly ended NVIDIA's existence involved a contract with Sega to develop the graphics for their Dreamcast console. The project, originally valued at $12 million, was jeopardized because NVIDIA's core technology was, as Huang admitted, fundamentally flawed. Huang made the difficult decision to travel to Japan and inform Sega's CEO, Madsaki Tanaka, that they could not fulfill the contract due to their technology's limitations. Despite admitting their inability to deliver, Huang still requested the payment, framing it as essential for their survival and to provide them time to find a solution. Tanaka, recognizing Huang's honesty and belief in his company, agreed to provide the funds. This $5 million kept NVIDIA afloat, demonstrating the investor principle of backing people and trust, and was crucial in giving them the opportunity to pivot and eventually succeed, leading to NVIDIA's public offering with a valuation of $300 million in 1999.
The universal function approximator: deep learning's potential
The advent of AlexNet, a significant breakthrough in deep learning, provided NVIDIA with a new lens through which to view their core competency. Huang realized that deep learning was not just about specific algorithms like AlexNet, but represented a more fundamental capability: the ability to learn any function. This understanding led him to declare, about 15 years prior to the talk, that humanity had discovered the universal function approximator. This meant that with sufficient data, a system could learn virtually any function, even those that are imprecise or inherently difficult to define precisely. This insight was transformative, prompting NVIDIA to explore its implications across the entire computing stack – from processors and middleware to applications. It directly fueled their immediate focus on areas like computer vision, robotics, and self-driving cars, recognizing the foundational nature of this new approach to software and computation.
The CEO as a surfer: navigating rapidly changing technology
Huang describes the CEO's role as akin to surfing, requiring an understanding of the waves, reading the wind, and having good timing. This is achieved through continuous engagement and practice, by being 'in the weeds' to maintain a tactile sense of rapidly evolving technology. This hands-on approach serves multiple purposes: informing the CEO, breaking down complex problems for the organization, and inspiring others. Huang refutes the idea of needing conventional management techniques, likening the company to an F1 racer that the founder must build and adapt to their own driving style. The focus is on personal effectiveness and enabling the company to win, with the understanding that future leaders can adapt the 'car' to their needs. This approach is driven by curiosity and a desire to empower the company by sharing insights derived from deep technological understanding.
Systems thinking for an agent-driven future
Looking ahead, Huang stresses the paramount importance of systems thinking, awareness, design, and organization. This is because the future of computing, especially in software, will increasingly be handled by agents, automating most low-level tasks. Therefore, the ability to think abstractly about problems, understand system constraints, input/output flows, and information rates will be critical. This includes understanding bottlenecks related to processors, memory, and networking. This fundamental knowledge, Huang argues, will become more and not less useful. He also touches upon the recursive self-improvement of agents, where their long-term memory is constantly updated and processed. A key challenge and opportunity lies in developing fine-grained control over these agents, allowing for precise modifications to their outputs – whether it's a single pixel, a component in a CAD file, or a line of code. Controllability, he suggests, is the single biggest breakthrough needed for agents.
Physical AI and the rise of robotics
The generation of video through AI, initially demonstrated internally by NVIDIA, opened the door to physical AI and robotics. Huang realized that if AI could generate video of actions like finger movements or picking up a glass, it could be harnessed to control robots performing similar tasks. The critical next step is enabling robots to understand and obey the laws of physics, including causality, friction, and tension. NVIDIA's work on the 'world foundation model' aims to create an AI that comprehends these physical principles. While the 'ChatGPT moment' for robots occurred a few years ago, unlocking imagination, the current focus is on creating robust training environments through 'real-to-sim' and 'sim-to-real' processes, using simulators like Isaac Sim. Economically, self-driving cars are seen as the first significant application of physical AI, with NVIDIA chips and software integrated into vehicles and data centers. The company's investment in this area, including open-sourcing autonomous navigation software, has already yielded a business valued at approximately $10 billion, with potential to become a $100 billion industry.
AI as a job creator and the future of work
Huang challenges the narrative that AI and automation destroy jobs. He argues that AI automates specific tasks within a job, rather than eliminating the job entirely, because jobs have a broader purpose composed of many tasks, some of which cannot be automated. Evidence suggests that productivity gains from AI lead to increased demand and new job creation. For instance, even as AI automates coding tasks, the number of software engineering jobs has increased. Similarly, while AI can read radiology scans, the demand for radiologists has grown due to increased patient throughput. This pattern indicates that by automating existing tasks, AI frees up human capital to pursue more ambitious projects, tackle larger backlogs, and drive economic growth, ultimately leading to more employment opportunities across industries.
The 'how hard can it be?' mindset and daily resilience
For young entrepreneurs, Huang advises embracing the 'how hard can it be?' mindset. This attitude, coupled with a belief in one's ability to learn, is crucial for navigating the complexities of starting a company and encountering new technologies and markets. He acknowledges the inherent fear and uncertainty faced by founders, recalling his own apprehension about fundraising. However, he emphasizes that focusing on daily resilience—getting through one day at a time—is more important than overcoming everything at once. This approach prevents anxiety from paralyzing action. Learning is identified as the single greatest superpower, and approaching challenges with the belief that they can be overcome, especially with AI agents assisting, empowers founders to persevere. Ultimately, it's resilience and a continuous drive to learn and act that lead to significant achievements like NVIDIA's.
Mentioned in This Episode
●Products
●Software & Apps
●Companies
●Organizations
●Concepts
●People Referenced
Jensen Huang's Founder & Leadership Principles
Practical takeaways from this episode
Do This
Avoid This
Common Questions
NVIDIA initially focused on reinventing 3D graphics with the idea of turning every PC into a game console. However, the algorithms and technology they developed for this purpose were fundamentally wrong.
Topics
Mentioned in this video
The company founded by Jensen Huang, which started with a flawed vision for 3D graphics but became a leader in accelerated computing and AI.
A gaming company that partnered with NVIDIA for a game console project, providing a crucial financial lifeline during a difficult period.
Advanced Micro Devices, mentioned in the context of 'AMD doll's laws', likely referring to Amdahl's Law, a principle in computer architecture.
Cited as an example of organizations that teach how to start a company, contrasting with Jensen Huang's early experience of learning from books.
An automotive company for which NVIDIA provided chips for their self-driving car technology.
An automotive company where NVIDIA's chips and software stack are utilized in both the vehicle and the data center.
An API specification for rendering 2D and 3D vector graphics, which Jensen Huang used to learn and reinvent NVIDIA's approach to computer graphics.
An early, influential deep learning model considered a breakthrough, which NVIDIA recognized as an approach to learning any function.
An AI agent platform that enables companies to build their own domain-specific AIs, discussed in the context of enabling everyone to build AI.
A widely known AI chatbot developed by OpenAI, used as a benchmark for the 'ChatGPT moment' in robotics and for general AI use.
The world's first 'thinking' self-driving car, developed by NVIDIA through its open-sourced driving stack, demonstrating the power of reasoning in AI.
An open-source operating system that played a crucial role in the development of the mobile cloud industry and modern AI.
An open-source system for automating deployment, scaling, and management of containerized applications, crucial for modern AI infrastructure.
An open-source machine learning framework developed by Google, essential for the advancement of modern AI.
An open-source machine learning framework widely used in AI development, considered pivotal for modern AI.
An AI assistant from Anthropic, recommended for general use alongside ChatGPT.
NVIDIA's robotics simulation platform used for training and evaluating robots, part of the 'real to sim' process.
More from Y Combinator
View all 611 summaries
30 minWhat Big Tech Missed And How Startups Can Still Win
21 minWhy Physical AI Is the Next Platform Shift
48 minWhat Actually Makes A Startup Durable
45 minOpencode CEO: Blocked, 20X Growth in 6 Months, Building the Coding Agent for the World
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free