#376: Jensen Huang: Founder of Nvidia
What this covers
Founders host David Senra works through the history of Jensen Huang and NVIDIA, setting it beside lessons from other founders and investors. The story begins with Sequoia backing the company on Huang's reputation, vouched for by LSI Logic's CEO, rather than on a pitch Huang later called horrendous. It continues through NV1, the first product, which failed because it was over-designed and its positioning was too complicated, a point Senra measures against David Ogilvy's definition. It then reaches the 2007-2008 collapse, when the stock fell more than 80%.
The middle of the episode is the technology and strategy. Senra traces how CUDA grew from a 2002 paper in which researchers hacked GPUs to simulate cloud movement. NVIDIA seeded universities with machines, sent chief scientist David Kirk on over a hundred talks, and published a textbook. The push took four years and cut gross margin from 45% to 35%. In 2013 Huang overruled lieutenants who saw deep learning as a fad. Other threads include a six-month chip cadence run by three design teams against rivals' eighteen months, and Ship the Whole Cow, which turned chips that failed top-grade tests into cheaper products. Huang's reading of Clayton Christensen on the Innovator's Dilemma is to attack his own company from the low end first.
The contested material concerns management. NVIDIA has no five-year plan and practices Continuous Planning, which Senra sets against Henry Singleton's steer-daily approach. Huang runs a flat structure of about 60 direct reports with no one-on-ones, criticizes mistakes publicly, and reads over 100 T5T emails a day. Senra also brings in Munger's Lollapalooza effect and Buffett on reputation, and notes that Huang prefers to describe the CUDA network rather than call it a moat.
Jensen Huang's management of NVIDIA, built on pain tolerance, public criticism, a flat organization, and fast decisions, was deliberately designed around his own temperament, and his early bet on general-purpose GPU computing (CUDA) is what turned that operating style into an AI-era advantage.
- A company built in the founder's own image, with direct reports, whiteboards, and T5T emails, is what allowed fast, unfiltered decisions
- Early failures (NV1, layoffs, cash crunch) forged the resilience and self-criticism that shaped the operating system
- Going all in on CUDA and then deep learning, against internal and investor pushback, created the moat
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
NVIDIA educated the market for CUDA by giving universities CUDA-capable machines in exchange for teaching classes, having chief scientist David Kirk give over a hundred talks, and writing a textbook that sold tens of thousands of copies.
“David Kirk gave over a hundred talks over the course of a year on this.”
Jensen recognized that general-purpose GPU computing could expand NVIDIA's market beyond graphics, so he built CUDA, a programming model that lets scientists and engineers use GPU computing power, to generate demand by making GPUs easier to program.
“they create this new programming model for chips that they call the Compute Unified Device Architecture. It's called CUDA.”
In 2013, Jensen overruled several key lieutenants who believed deep learning was a passing fad, declaring 'Deep learning is going to be really big' and that NVIDIA was going all in on it.
“'Deep learning is going to be really big,' he said this in 2013. 'We are going all in on it.'”
Jensen's approach to the Innovator's Dilemma is to ask what he would do to attack his own company and then do it before someone else does, which means taking the threat from the low end seriously.
“So what if I wanted to attack myself? What would I do? And then I'll just go do that before somebody else does it to me.”
Charlie Munger's Lollapalooza effect holds that multiple psychological and social factors within an organization interact so that their combination makes the traits far more powerful than any one alone.
“the combination of them actually makes the traits more powerful.”
NVIDIA split its design team into three groups so it could release a new chip every six months, while competitors took 18 months per chip, making rivals' slightly better products irrelevant because a faster part would arrive within six months.
“This was called three teams, two seasons.”
Sequoia invested in NVIDIA based on Jensen's reputation, introduced by LSI Logic CEO Wilfred Corrigan, not on the business plan, and Jensen later admitted his pitch was horrendous.
“They succeeded on the strength of your reputation, not their business plan, not their pitch.”
NVIDIA's push to make CUDA a foundational technology took four years and cut gross margin from 45% to 35%, and during the 2007-2008 period NVIDIA's stock fell more than 80% as the financial crisis destroyed consumer demand.
“its gross margin fell from 45% to 35% as NVIDIA increased spending on CUDA.”
NVIDIA began repackaging chips that failed Ferrari-grade quality tests into cheaper derivative products, turning previously discarded parts into revenue and a low-end defense against competitors, a strategy called ship the whole cow.
“Rejected parts were generating no revenue for NVIDIA—they were thrown out.”
Decisions should protect long-term reputation because a good reputation opens future deals and funding while a bad one can abort future opportunities, as Buffett told Congress during the Salomon scandal.
“If you lose a single shred of reputation, I will be ruthless.”
Huang's takeaway from Clayton Christensen's Innovator's Dilemma is that competitive threats come from the low end of the market, so NVIDIA built high-end Ferrari chips while creating its own low-cost line rather than letting price leaders climb from the bottom.
“I don't want to let someone come in and be the price leaders and lock me out of the bottom and climb their way to the top.”
Huang concluded that it is better to do fewer things well than to do many things, because customers buy specialized products, not all-purpose ones, as with the Swiss Army knife analogy that James Dyson also reached.
“We learned it was better to do fewer things well than to do too many things. Nobody goes to the store to buy a Swiss Army knife.”
The first use of NVIDIA GPUs for non-graphics work came in a 2002 research paper on simulating cloud movement, where researchers hacked GPUs to reframe computations as graphics functions and reported large speedups over CPUs.
“the first reference to using NVIDIA GPUs for non-graphical purposes was in a 2002 research paper on using computers to simulate the movement of clouds.”
NVIDIA does not do periodic planning because the world is a living, breathing thing, so it plans continuously; Jensen says strategy is not words but action, and there is no five-year plan.
“Strategy is not words. Strategy is action.”
NVIDIA's first product, the NV1, failed because it was over-designed with features no one cared about, when the market simply wanted the fastest graphics performance for top games at a decent price.
“We learned it was better to do fewer things well than to do too many things. Nobody goes to the store to buy a Swiss Army knife.”
NVIDIA's competitive advantage is a self-reinforcing network built around the programmable CUDA layer, which expands as more people learn CUDA and increases GPU demand, a network that Jensen prefers to describe instead of calling a moat.
“he prefers what he's really trying to build, and what he has built is a strong self-reinforcing network.”
David Ogilvy's 1965 definition of product positioning, that it is what the product does and who it is for, remains the best simple definition, and NV1's positioning failed because it was too complicated.
“David Ogilvy's definition of product positioning is simple: it's what the product does and who it's for.”
Jensen built a flat structure with about 60 direct reports and no one-on-ones because flat organizations fight slow decision-making, which drives great people away.
“You can drive great people away by making the speed of decision-making really slow.”
Henry Singleton believed in staying flexible and steering the boat daily rather than detailed long-term planning, which he thought constrains freedom of action because most outside influences cannot be predicted.
“My only plan is to keep coming to work every day. I like to steer the boat each day rather than plan way ahead into the future.”
Jensen believed AI would create the largest total addressable market expansion of software and hardware seen in decades, and the host argues that a decade later this has proven accurate.
“Jensen was sure that AI would create the largest total addressable market expansion of software and hardware that we've seen in decades.”
Jensen criticizes mistakes publicly, in all-company meetings, so the whole organization learns from a single person's error, as an antidote to turf protection, backstabbing, metric gaming, and political infighting.
“Feedback is learning. For what reason are you the only person who should learn from this? We should all learn from that opportunity.”
The great founders saw themselves primarily as teachers, and Costco's Jim Sinegal argued that a company leader who isn't spending 90% of time teaching is not doing the job.
“If you're not, as a leader of the company, if you're not spending 90% of your time teaching, you're not doing your job.”
Successful people like Jensen and Mitch Rales take extensive notes at conferences even after they are the most successful person in the room, because champions behave like champions before they become champions.
“champions behave like champions before they're champions.”
NVIDIA judges its speed against the physical limit of what is possible, breaking projects into tasks with target times assuming no delays, rather than against its past performance or competitors' actions.
“We're not going to judge ourselves on what we used to do. We're not going to judge ourselves on what other companies do.”
Jensen requires every employee to send a weekly-style 'Top 5 Things' email (T5T) listing the top five things they are working on or observing, and reads over 100 per day to catch weak signals before they pass through management layers.
“It's easy to pick up on the strong signals, but I want to intercept them when they are weak.”
Jensen organized employees into centralized functional pools such as sales, engineering, and operations, rather than business units, so that talent can be redirected to a new mission.
“They're treated as a general pool of talent and not divided by business units or divisions.”
A wonky machine-learning development that kept appearing in T5T emails led Huang to invest in tools for accelerating workloads on GPUs, a decision the host says helped make NVIDIA a $3 trillion company.
“One of the weak signals that he intercepted years ago was a wonky but exciting development in machine learning that kept popping up in T5T emails.”
NVIDIA's first five years were marked by talented people working very hard but being bad at their jobs, because building a company is a new skill that must be learned.
“We were just bad at our jobs. The first five years of our company, we had really talented people working super, super hard. But building a company is a new skill.”
Whiteboard meetings force people to demonstrate their thought process in real time before an audience, leaving no place to hide, so anyone who hasn't thought something through is immediately exposed.
“With the whiteboard, there is no hiding.”
NVIDIA does not chase market share in existing markets and instead creates markets through cutting-edge, revolutionary work that no one else can do, to avoid commodity pricing pressure.
“We do not—this is really the punchline—we don't have a culture of going after market share. We would rather create the market.”
Doing each opportunity in front of you as well as possible unlocks further unexpected opportunities, because reputation and results compound across a career.
“all-time opportunity handled well leads to more opportunity.”
Jensen believes NVIDIA's worst enemy is not competitors but its own complacency, and that the company survived its early failures because it was its own worst enemy.
“We survived ourselves. We were our own worst enemy.”
Huang believes NVIDIA's worst enemy is not its competitors but its own complacency, which grips any successful company, so he refuses to rest on past achievements and treats innovation as a necessity rather than an option.
“We survived ourselves. We were our own worst enemy.”
People with very high expectations have very low resilience, and resilience is what matters for success.
“People with very high expectations have very low resilience. Resilience matters in success.”
Sol Price's view is that the best leaders create alter egos through teaching, so employees eventually perform their jobs as well as or better than the owner.
“the best leaders create alter egos through teaching.”
Jensen wants information from the edge, meaning raw signals from employees closest to customers and markets, not information already filtered through layers of management, because strategy is what people do.
“What I want is information from the edge.”
NVIDIA treats the mission as the boss, assigning a named 'pilot in command' accountable for each project who reports directly to Jensen, so the company can redirect pooled functional resources when the mission changes.
“We have a PIC—a pilot in command—for every project. Nobody can hide behind 'such and such team is working that.'”
To apply the speed of light, Jensen reversed the usual process and had NVIDIA develop Riva 128 driver software before the prototype chip was finished, saving nearly a year on the production timeline.
“NVIDIA would have to develop their driver software for the Riva 128 before the prototype chip was completed. This was a reversal of the customary process.”
Excellence is the capacity to take pain, and Jensen Huang's career (reform school, AMD job plus Stanford master's on nights and weekends, early NVIDIA failures) demonstrates unusually high pain tolerance as the source of his success.
“excellence is the capacity to take pain. Jensen has unfathomable levels.”
Jensen holds that greatness is not intelligence but character, and character is earned only by overcoming problems, setbacks, adversity, and refusing to quit.
“Jensen said, 'Greatness is not intelligence. Greatness comes from character.'”
Jensen demands blunt, concise communication, using the shorthand LUA (Listen to the question, Understand the question, Answer the question) to stop rambling, and his emails are as short as haiku.
“Jensen's emails are short and sweet like a haiku.”
Huang uses the whiteboard as his primary communication tool because it forces people to demonstrate their thinking in real time, requires them to start from scratch, and makes it obvious when someone has not thought something through.
“With the whiteboard, there is no hiding.”
Teams should seek rather than avoid conflict because the friction of explosive debate about what to build and how to run the company produces sharper, more refined ideas, as a sword is sharpened by repeated honing.
“If I have to choose between agreement and conflict, I'll take conflict every... every time. Why? Because it always yields a better result.”
Overpaying top talent is nearly impossible to do wrong, so Jensen grants special one-off stock awards that vest over four years, reviewing a list of top contributors and giving stock in real time rather than waiting for annual reviews.
“You really should overpay for top talent because it's nearly impossible to overpay for top talent.”
Steve Jobs's rock tumbler story argues that a group of talented people rubbing against each other, creating friction and noise, polishes each other and their ideas, the way a tumbler turns ugly rocks into polished ones.
“It is through a group of incredibly talented people bumping up against each other, working together. They polish each other, they polish their ideas.”
Senra argues that NVIDIA is effectively Jensen Huang extended across its roughly 29,000 employees, and that a CEO must build an organization that he can personally drive, like a race car, which he borrows from Michael Moritz on Steve Jobs.
“The company's organization is like a race car. It has to be a machine that the CEO knows how to drive.”
Jensen looks in the mirror each morning and says 'you suck,' an internal self-criticism that persists even after NVIDIA smashed quarterly results, reflecting his drive to torture himself into greatness.
“I wake up every morning, I look at myself in the mirror and I say, 'You suck.'”
Jensen bought a roughly $1 million Ecos-style emulator that cut testing time, which cost runway from nine months to six but saved the company because the Riva 128 succeeded.
“But an Ecos machine cost a million dollars and would cut NVIDIA's runway from nine months down to six months. But Jensen realized it would speed up the testing process by significantly more time.”
Huang's extreme work hours, which he compares to Olympians training early, are not just personal intensity but a deliberate advantage, since NVIDIA's culture of speed and obsession let it outpace competitors whose employees moved in slow motion.
“People who trained for the Olympics grumble about training early in the morning, too.”
When Intel invented the microprocessor, it sent founders and executives to spend most of their time educating the market and running classes, which the host sees as a precedent for NVIDIA's approach to CUDA.
“Intel did this when they invented the microprocessor. In fact, they swarmed their greatest opportunity.”
Rockefeller turned waste materials from refining kerosene into profitable product lines, the same logic as NVIDIA repackaging discarded chips.
“Think about how many different products, byproducts that Rockefeller made from materials that were waste materials from the refining of kerosene”
James Dyson reached the same conclusion, that people want high-tech specificity rather than all-purpose products, after 5,127 failed prototypes.
“People do not want all-purpose. They want high-tech specificity.”
Jeff Bezos argued that slow decision-making can drive great people away because they will leave when they cannot get things done despite loving the mission.
“Why would great people stay in an organization where they can't get things done?”
Napoleon issued short orders that clearly expressed intent, and Churchill limited memo length, saying it is slothful not to compress your thoughts, since long orders are the enemies of speed.
“Long orders which require much time to prepare, to read, and to understand are the enemies of speed.”
Andy Grove's mantra 'Success breeds complacency, complacency breeds failure, only the paranoid survive' expresses the same principle that complacency is the main risk after success.
“Success breeds complacency. Complacency breeds failure. Only the paranoid survive.”
Founders need to stay in the game long enough to get lucky, and NVIDIA's AI dominance came from weak signals, customer-led uses, and decades of persistence rather than a single bet.
“this idea that you have to stay in the game long enough to get lucky.”
Michael Bloomberg held that planning must never get in the way of doing, and that a five-year plan is a form of central planning that fails for entrepreneurs as it did for Stalin and Mao.
“Central planning didn't work for Stalin or Mao and it won't work for an entrepreneur either.”
Marketing and positioning must be simple, because seduction requires a simple message, and the NV1 failed partly because its messaging was too complicated.
“Seduction requires a simple message.”
James Dyson's first autobiography, 'Against the Odds,' describes 14 years of struggle and 5,127 failed prototypes, and the host considers it his top recommended business book.
“It's him persevering and enduring through 5,127 failed prototypes.”
Steve Jobs told Jony Ive that he was too vain and cared more about being liked than about the work, showing that a harsh critique can reflect putting the work above perceived feelings.
“You just want people to like you. I'm surprised at you. I thought you held the work up as the most important.”
Huang and Steve Jobs both avoid dwelling on past accomplishments, arguing that looking back is poor use of time and that after a success one should figure out what's next.
“There's nothing worse than looking back at your own accomplishments—it's just not good use of your time.”
The host argues that Brad Jacobs, with a strong track record, could give a poor pitch and still attract capital because his reputation precedes him.
“Brad Jacobs could do a horrible job with a pitch too.”
Jensen has a long-term horizon, being impatient about certain things but infinitely patient about others, which allowed him to finish a Stanford master's degree over eight years while working and raising a family.
“I have a very long-term horizon. I can be impatient about certain things but infinitely patient about others. I plug away.”
Bob Kierland of Fastenal argued that the leader's core job is keeping everyone inside the organization aligned to the company's core objective, because human nature drifts away from it, and Huang grants project independence only when it stays aligned.
“the leader's job is to keep everyone inside of the organization working for the company's core objective. That is very natural in human nature to drift.”
Jensen steadfastly refused to hire a chief operating officer, even when new board members recommended one to reduce his administrative burden, because he wanted to manage the company directly.
“Says he steadfastly refused to change his management philosophy even when new board members joined and recommended that he hire a chief operating officer”
Jensen publicly berated engineers at an all-company meeting over the poor planning and execution of the NV30 chip, asking whether it was the piece of shit they intended to build.
“Is this the piece of shit you intended to build?”
Jensen believes in torturing people into greatness, and he applies this to himself as well, combining extreme self-confidence and charisma with an inner voice that is impossible to satisfy.
“I don't like giving up on people. I'd rather torture them into greatness.”
Jensen treats second place as failure, insisting 'second place is the first loser,' and believes competitive advantage comes from outworking rivals in time and effort, which he likens to Olympians grumbling about early training.
“People who trained for the Olympics grumble about training early in the morning, too.”
NVIDIA runs an intense email culture in which Jensen responds quickly and expects staff to act on his emails immediately, so nothing festers.
“If you get an email from him, you need to act on it right away.”
Huang gathered the whole company, read out the small amount of money left in the bank, then revealed a purchase order for 30,000 units from STB Systems, which prompted the entire company to cheer and shows his use of theatrical framing to motivate staff.
“One purchase order from STB Systems for 30,000 units of Riva 128. And the entire company erupted in cheers.”
Jensen's mother had her sons memorize ten random dictionary words every day, with definitions and spelling, to prepare them for America.
“she would select 10 random words from the dictionary every day. She'd make her sons memorize the definition and learn how to spell the words.”
Decisions made by parents before a child has memory can shape the entire trajectory of that child's life, as with Jensen's parents sending him to America and the host's grandfather leaving Cuba.
“The decisions that were made before Jensen even had a memory affected the entire trajectory of his life.”
Huang recognized the AI trend very early and invested heavily when others told him not to, positioning NVIDIA to capture the opportunity better than any other company.
“Jensen is going to recognize the AI trend very early. He's going to invest heavily when other people are telling him not to do this.”
Jensen attended the NeurIPS machine learning conference and took his own notes in the front row, saying he was there to learn, and he recruits people he meets at such conferences.
“And Jensen told him, 'I'm here to learn.'”
Reading biographies is valuable because human nature leads readers to put themselves in the subject's shoes and relate the subject's story and ideas to their own life.
“it's human nature to put yourself in their shoes and think about how their story or their ideas relate to your own.”
Huang says he will not be outworked, that working is relaxing for him, and that Olympic athletes do not complain about early practices, so long hours are a necessary prerequisite for excellence.
“People who trained for the Olympics grumble about training early in the morning, too.”
Jensen was smaller and younger than classmates at the Kentucky reform school and was beaten until he gained respect, and he later credited this experience with making him fearless and able to tolerate a lot of discomfort.
“As a result of this experience, I don't get scared often. I don't worry about going places I haven't gone before. I can tolerate a lot of discomfort.”
The ideal T5T email is five bullets, each starting with an action word such as finalize or build, and departments tag subject lines by topic so Jensen can search them.
“The ideal T5T email is five bullet points where the first word is an action word”
Special one-off stock grants vesting over four years are approved for top contributors, with Jensen copied on an email with the subject 'special grant' that explains the rationale.
“the subject line will say 'special grant' which authorizes the RSU grant in recognition of your extraordinary contributions”
Michael Moritz described the very best founders as unstoppable, irrepressible forces of nature, and the host argues Huang and Jobs both remade their companies in their own image.
“Founders, at least the very best of them, are unstoppable, irrepressible forces of nature.”
Relationships and personal networks often determine access to capital and opportunity more than the quality of the pitch itself.
“relationships run the world.”
Charlie Munger taught that concepts are best taught when tied to the lives and personalities of the people who developed them, which the host applies in this episode.
“I think it helps to tie them into the lives and personalities of the people who develop them.”
Huang stays involved in product decisions, sales negotiations, and investor relations, following Walt Disney's principle that if we lose the detail, we lose it all, and Senra argues that accusations of micromanagement miss the point of organizational control.
“If we lose the detail, we lose it all.”
Jensen recalled the career history of new hires at a party, including Peter Young's background from Sony PlayStation and 3dfx, showing his attention to employee details.
“You're Peter Young. You've been here for a year from Sony PlayStation, and then you worked at 3dfx.”
Jensen stays deeply involved in details such as product decisions, sales negotiations, and investor relations, following Walt Disney's principle that losing the detail means losing it all, and does not work from a private office.
“If we lose the detail, we lose it all.”
The host believes Jensen Huang is probably one of the most important founders alive right now, and that his ideas are useful for other founders.
“he's probably one of the most important founders alive right now.”