How I Built This with Guy Raz

Jensen Huang: The Personal Cost of Building NVIDIA (April 2026)

· Summarized

Share this episode
Anyone with this link can open it.
Share Link
Start at 0:00

Key takeaways

  • Huang’s first existential lesson was to abandon NVIDIA’s flawed NV1 architecture even though doing so meant canceling a lucrative Sega contract and laying off two-thirds of the company.
  • NVIDIA survived by spending half its remaining cash on an emulator that let engineers test the Riva 128 without the normal chip-design iteration cycle.
  • CUDA became strategically valuable because NVIDIA carried it inside gaming-focused GeForce GPUs for years, distributing the platform before a profitable market existed.
  • Huang says AI’s breakthrough came when researchers used NVIDIA gaming GPUs and CUDA to train image-recognition systems far more effectively than CPU-based systems.
  • Huang believes AI warnings should be channeled into safety engineering rather than alarmism that discourages people from entering professions society will still need.
  • He argues that AI often automates tasks without eliminating a job’s broader purpose, citing increased demand for radiologists after AI improved scan analysis.
  • Despite NVIDIA’s extraordinary success, Huang says he would not repeat the journey if he had known its suffering and personal sacrifices in advance.

Summary

Jensen Huang’s childhood in Kentucky and Oregon

Jensen Huang says his parents moved him and a brother from Taiwan and Thailand to the United States at about age nine because of political unrest. They sent the boys to the affordable Oneida Baptist Institute in rural Kentucky, where every student worked; Huang cleaned bathrooms, shared an open dorm with a 16-year-old, and played swimming and soccer. He remembers recording monthly messages for his parents because international calls were too expensive, including an excited account of a coach taking the team to McDonald’s. After rejoining his family in Portland, he attended Oregon State, studied electrical engineering, and became lab partners with Lori, who became his wife.

From hand-drawn chips to NVIDIA’s founding

Huang describes entering chip design in California around 1984, at the transition from hand-drawn layouts to software-based design. At LSI Logic, he helped systems companies such as Sun Microsystems and Silicon Graphics build chips and saw startups succeed or fail through different combinations of technology and strategy. In 1992 or 1993, Sun engineers Chris Malachowsky and Curtis Priem asked him to lead a company making affordable 3D graphics for personal computers. Although he had a secure, well-paid job, two children, and only $30,000 in savings, Huang says the opportunity looked so clear that he never seriously imagined NVIDIA failing.

The NV1 disaster forced NVIDIA to change architecture

NVIDIA’s first product, the NV1, aimed to bring graphics comparable to the million-dollar systems used for films such as Jurassic Park into a roughly $300 consumer product. The company sold about 250,000 chips, but the underlying forward-texture-mapping architecture was incompatible with Microsoft’s DirectX, so the chips came back and Huang calls the result a technology disaster. NVIDIA had a $12 million Sega contract guaranteeing purchases of the next chips, but Huang went to Japan and advised Sega’s CEO to find another partner rather than continue with a flawed design. He persuaded Sega to convert the remaining $5 million into an investment, then laid off two-thirds of NVIDIA’s roughly 200-person workforce.

A $5 million gamble kept NVIDIA alive

With only six months of cash left, Huang says NVIDIA could not afford the normal one-and-a-half-to-two-year cycle of designing, testing, debugging, and remaking a chip. The company had shifted production to TSMC because European capacity was insufficient, but would run out of money before the new chip returned from fabrication. Huang bought an unwanted Icos emulator from the creditors of a failed company, spending half of NVIDIA’s remaining money on a machine that could imitate the chip and run its software before fabrication. The engineers did not object because, as Huang recalls it, the emulator was the only idea available. The resulting NV3, or Riva 128, worked well enough to save the company.

CUDA was carried through gaming GPUs for years

After Riva 128, NVIDIA focused on gaming and treated game developers as essential because applications, not chips alone, create industries. Inspired by Clayton Christensen’s The Innovator’s Dilemma, Huang says disruptive products often look like toys until they become “good enough.” NVIDIA’s parallel processors began handling physics, fluid dynamics, image processing, and other tasks beyond graphics. Around 2006, the company created CUDA, a software layer that let researchers use gaming GPUs for general-purpose parallel computing. Universities experimented with it, but those users did not pay the bills, so NVIDIA embedded CUDA in GeForce cards and gave it away, accepting poor margins while evangelizing it to universities and developers.

The AI breakthrough came from gaming GPUs

For years, CUDA had little commercial success: NVIDIA’s stock was flat or falling from roughly 2006 to 2012, and Huang says the company repeatedly had to keep itself alive for “just another few more days.” The architecture faced a chicken-and-egg problem because developers would not write software for a platform without users, while users needed software to justify buying it. In the early 2010s, researchers in Toronto used two NVIDIA gaming GPUs in a bedroom computer to train a neural network for image recognition. Groups preparing for the ImageNet computer-vision contest contacted NVIDIA for help accelerating deep learning with CUDA, and the dramatic improvement over earlier algorithms convinced Huang that AI could become the company’s future. In 2013, he set that direction after bringing colleagues along through extensive discussions.

Huang rejects both AI complacency and extinction alarmism

Huang says people should separate practical computer science from Hollywood versions of conscious AI, while still directing concern toward building safer systems. He criticizes warnings that AI will inevitably eliminate professions or cause extinction, comparing them to an auto industry repeatedly telling customers cars will kill them. As an example, he says AI has transformed radiology and reads scans faster and more accurately, but demand for radiologists increased because cheaper diagnosis enables hospitals to see more patients. He fears American alarmism will discourage young people from entering fields such as computer science, radiology, and engineering while other countries adopt the technology. AI assistants, he argues, will let future graduates build things that previously seemed impossible.

Taiwan, supply chains, and Huang’s leadership values

Asked about a possible conflict involving Taiwan, where advanced chips are produced, Huang says companies should seek resilient supply chains with diversity and redundancy, though those options are limited. He is less worried than most about a hot conflict and argues that leaders should avoid pushing adversaries into having no strategic alternative; he believes wisdom can still prevail. On leadership, he says employees care less about style than consistent values. He may be demanding or yell, but the confrontation ends when the issue ends, and employees should know he loves them, wants them to succeed, and is on their side. He calls himself a “battlefield CEO” who is most comfortable in crises.

The success came at a personal cost—and he would not repeat it

Huang says he would not start NVIDIA again if he had known the full cost: years of humiliation, bad financial news, layoffs, near-bankruptcy, and the need to forget yesterday and keep moving. He missed his children’s karate tournaments while working long days and completing a Stanford master’s degree, but credits Lori with managing the family without complaining. Their children, Spencer and Madison, later joined NVIDIA and had followed the company through shareholder letters, meetings, and conferences. Huang dislikes talking about himself because he built nothing alone and does not find his own decisions especially interesting; he prefers being brought into a crisis. He attributes NVIDIA’s path to hundreds of forks combining judgment, values, friends, critical thinking, and luck.

Sponsors

  • Anthropic · The team behind Claude describes itself as a public benefit corporation founded around the question of how to make AI go well for people. It says it asked more than 100,000 people about their hopes and fears and is publishing the results, including uncomfortable findings. The host also uses Claude to think about what separates people who build businesses when AI can help everyone write plans, make prototypes, analyze markets, or generate ideas.
  • Framer · Framer presents itself as an AI website builder for creators, teams, and businesses. Users can prompt, inspect, edit, and publish production-ready sites in one place. Listeners are offered help from a Framer specialist or a 30% discount on an annual Framer Pro plan.
  • Upwork · Upwork lets businesses hire highly skilled specialists for one-off or ongoing work, fill skill gaps, and scale support without full-time headcount. Business Plus offers access to the top 1% of talent, with AI-powered shortlisting producing a curated list in under six hours. The ad says more than 170,000 freelancers have five-star ratings and that posting a job is free.
  • Engine · Engine offers business travel booking with zero booking fees, no contracts, and no BS. Customers can book trips in as little as two and a half minutes, earn up to 10% back on hotels, and collectively saved more than $300 million in 2025. Businesses that sign up and start traveling can receive $500.
  • Canva · Canva says businesses can store colors, fonts, and logos in a shared brand kit so work does not depend on one designer. The ad jokingly promises that a vacationing brand manager can relax knowing nobody will use Comic Sans, and promotes Canva for brainstorming, collaboration, presenting, editing, resizing, and creating.
  • Indeed · Indeed Sponsored Jobs gives job postings premium status and is intended to help employers spend less time searching and more time interviewing candidates who meet their requirements. Listeners receive a $75 sponsored-job credit.
  • Philadelphia · Philadelphia promotes its original cream cheese, made with fresh milk and real cream, as the choice for holiday recipes including cheesecakes and cookies.
  • BambooHR · BambooHR is described as putting business data in one place and providing actionable insights so employees and businesses can perform at their best.
  • Odoo · Odoo offers all-in-one business management software connecting sales, accounting, inventory, marketing, and other functions. It emphasizes avoiding messy integrations, disconnected applications, and spreadsheets, and invites businesses to try it for free.
  • Delta · Delta, the official airline of the National Women's Soccer League, says travel creates opportunity, understanding, connection, and new horizons. It advertises up to 5,500 daily Delta and Delta Connection flights to more than 300 destinations on six continents.

Show notes

Given recent news about Jensen Huang, we're revisiting this unusually personal conversation with the Nvidia co-founder and former CEO. Recorded in April 2026, before the revelations that led to his resignation, the interview offers a fuller picture of Jensen: his family, the sacrifices he made during 33 years leading the company, his demanding leadership style, and his views on the U.S., Taiwan, and AI.

Nvidia is one of the most valuable companies in human history. Its chips run the AI systems transforming everything from entertainment to warfare. But for years, Jensen struggled to convince others of his vision. He spent nearly a decade pouring billions of dollars into a technology called CUDA, long before AI made it profitable.

Jensen also shares how Nvidia came within 30 days of going out of business, what it took to lead the company through decades of uncertainty, and why, despite its phenomenal success, he's not sure he'd choose to do it all again.

Watch this Episode on YouTube: 

Jensen Huang: The Man Behind NVIDIA (April 2026)

What You’ll Learn:

Why Jensen doesn't like talking about himself

What he missed out on while building Nvidia

Why past failures still haunt him

How a desperate trip to Japan helped save Nvidia

How AI researchers discovered the power of Nvidia's gaming chips

Why Jensen believes fears about AI and job loss are overblown

Production Credits

How I Built This is produced by Podcast Pointman and part of the Vox Media Podcast Network.

Follow How I Built This:

YouTube → @HowIBuiltThis

Instagram → @howibuiltthis

X → @HowIBuiltThis

Facebook → How I Built This

Follow Guy Raz:

Instagram → @guy.raz

X → @guyraz

Substack → guyraz.substack.com

Website → guyraz.com
Learn more about your ad choices. Visit podcastchoices.com/adchoices

Follow this show on AudioBrevity and new episodes get summarized for you. Create a free account

Couldn't load the audio. A blocker may be stopping the show's tracking links. Open the file