Founders Podcast · EXTRACTED

How Jensen Works

The flat org, the top five emails, and the speed of light benchmark. How Jensen built a company that runs on unfiltered information and permanent urgency.

▶185K views on YouTube

"I look in the mirror every morning and say, 'You suck.'" — Jensen Huang
David Sunra, on the episode

This is a solo episode from the Founders podcast, where host David Sunra rereads Tae Kim's book The Nvidia Way and strips it down to the operational ideas behind how Jensen Huang runs his company. The pop framing of Nvidia is a lucky AI bet that paid off. The actual operating system underneath is three decades of deliberate organizational design: a flat structure built for speed, a direct information channel from the edge of the company to the CEO, and a benchmark for execution that uses the laws of physics as the ceiling, not last quarter's numbers. Sunra pulls roughly 19 ideas from the book and organizes them into a single portrait of Jensen's management philosophy. This protocol surfaces the three plays with the clearest action for a founder or operator to run immediately.

Tactic 01

Run the Top Five Email System

As Nvidia grew, Jensen faced the same problem every scaling founder does: information gets filtered, softened, and politicized as it climbs through layers of management. His solution was structural. He asked employees at every level of the organization to send him a short email detailing the top five things they were working on and what they had recently observed, covering customer pain points, competitor moves, technology developments, and potential project delays. The format was specific: five bullet points, each starting with an action verb such as finalize, build, or secure. To make them searchable, each department tagged their emails by topic in the subject line, so Jensen could pull every recent email on, say, a specific account category with a single keyword search. Every day he read around a hundred of these emails. He described the goal plainly: it is easy to pick up on the strong signals, but I want to intercept them when they are weak. The emails gave him a live feed from the edge of the company, bypassing the distortion that builds up inside management layers. One of the weak signals he caught years before it became obvious was a recurring mention of an unusual development in machine learning. That intercept became one of the inputs that led Nvidia to invest in CUDA and accelerated computing before almost anyone else saw it coming. The Wall Street Journal described the system this way: strategy is not what I say, it is what they do. Jensen needed to know what everyone was actually doing, not what had been approved for him to hear.

The play

This week, send a one-paragraph note to your team, however large or small, asking each person to email you their top five priorities and one thing they have observed in the market, a customer complaint, a competitor move, or a signal that something is ahead of or behind schedule. Require the first word of each bullet to be an action verb. Ask each function to tag the subject line with their department. Read every response yourself. Do this for four consecutive weeks before delegating any part of it. The pattern you are looking for is the weak signal that appears in two or three emails from unrelated people before it surfaces anywhere else.

Tactic 02

Set the Speed of Light, Not the Historical Average

Jensen has a specific name for how he measures execution speed, and the name is doing real work. He calls it the speed of light. The idea is that every project gets broken into its component tasks, and each task gets a target completion time that assumes no delays, no queues, and no downtime. That number is the theoretical maximum, the fastest it is physically possible to complete the work given the constraints of reality. The team then measures itself against that number, not against what similar projects took in the past, not against what competitors are doing, and not against what seems reasonable. Jensen was explicit about why the benchmark had to be physics and not history: he needed to prevent the internal rot he had observed at other companies, the slow drift where the reference point for acceptable speed keeps shifting down toward comfortable. If you benchmark against last year, you are building in tolerance for last year's waste. If you benchmark against the theoretical maximum, every gap between actual and possible becomes visible and named. The practical effect is that the team can no longer hide behind what we used to do or what is normal in the industry. The only question is: what is the fastest this could physically be done, and what is standing between us and that number.

The play

Pick one active project this week. Map every task required to finish it. For each task, estimate the completion time assuming zero waiting, zero handoff delay, and full focus from the right person. Add those up. That is your speed of light for this project. Compare it to your current projected timeline. The gap is not schedule buffer. It is a list of problems to solve. Name each one and assign it to someone.

Tactic 03

Criticize Publicly So the Whole Company Learns

Jensen does not follow the conventional advice to praise publicly and criticize privately. He inverts it deliberately. When someone makes a mistake or produces poor work, he calls it out in front of the whole organization, and his reason is straightforward: if one person made this mistake, others are at risk of making it too, and keeping the feedback private means only one person gets to learn from it. His framing was direct: for what reason are you the only person who should learn from this. We should all learn from that opportunity. I don't take people aside. We are not optimizing for not embarrassing somebody. We are optimizing for the company learning from our mistakes. One of the clearest examples in the book came after a chip shipped with serious problems. Jensen called out the engineers at an all-company meeting and asked them point blank whether this was the product they intended to build. He described the architecture as a shitty job. He then invited an executive from Best Buy to speak to the company and spend most of the session walking through customer complaints about the product's poor performance. Jensen agreed with the criticism publicly. His broader point was that he had tortured himself with the same standard for decades. He says he looks in the mirror every morning and tells himself he sucks. The public criticism is not cruelty. It is the same standard applied outward. Johnny Ive, after Steve Jobs pushed back on him for wanting to soften criticism, put it clearly: he said no, Johnny, you are just really vain. You want people to like you. I thought you held the work up as the most important thing, not how you were perceived. Jensen operates from the same conviction.

The play

At your next all-hands or team review, when a mistake or a missed standard comes up, resist the pull to handle it privately afterward. Name the specific failure in the room, explain clearly what the standard was, what happened instead, and what a correct decision would have looked like. Then ask whether anyone else is facing a similar situation right now. You are not trying to humiliate anyone. You are converting a single mistake into a lesson the whole team can use. Do this consistently and the standard becomes self-enforcing.

Tactic 04

The Lua Warning Signal

Jensen runs meetings where the standard is blunt, concise, and direct communication. Short like a haiku. When an employee starts rambling instead of answering, Jensen has a single word he deploys: Lua. It is a warning sign that his patience is growing thin. When Jensen says it, he wants the employee to stop and do exactly three things. One, listen to the question. Two, understand the question. Three, answer the question. Everyone who works for Jensen has heard it. The word functions less like a reprimand and more like a reset, a way to pull someone back to the point before the conversation loses all momentum. This connects directly to how Jensen thinks about communication across the whole company. When a person and their ideas are easy to understand, those ideas spread. When someone can get to the point, their thinking is testable. When they cannot, they are usually hiding something, either from the room or from themselves. Lua is a tool for enforcing the standard Jensen holds in every meeting: say what you mean, answer what was asked, and do it quickly enough that the next person can build on it.

The play

Pick a word or phrase your team will recognize as a signal that an answer has gone off track. When you use it, the speaker must stop, repeat the original question back to you, and answer it directly. Practice it in your next meeting the first time someone starts answering a question that was not asked. The discipline is not about being harsh. It is about making clear that the most valuable thing in a meeting is precision, and that everyone's time depends on it.

Tactic 05

Pilot In Command Accountability

Jensen organizes every project around a single named leader he calls a pilot in command. That person reports directly to Jensen. The rule is simple and hard: no project exists without a name attached to it. No one gets to hide behind "the team is working on that" or "such and such department is handling it." The moment a project starts, a person owns it. The idea connects directly to how Jensen thinks about accountability at scale. Nvidia organizes employees into functional groups, a general pool of talent that can be redirected toward whatever the current mission demands. That flexibility is powerful, but it creates a real danger: diffused ownership. When responsibility belongs to everyone, it belongs to no one. The pilot in command structure is Jensen's answer to that. One person is accountable. Everyone else knows who it is. Jensen shared exactly the same instinct as Elon Musk, who built a similar rule into his operating algorithm: every requirement must carry the name of the real person who made it. Never a department. Always a human being. This is not a bureaucratic title. It is a mechanism for keeping accountability visible and personal inside a flat organization. Jensen wants to know the name. Whenever he asks about any project, the answer has to include a person, not a function. That single habit prevents the diffusion of responsibility that grows silently inside large companies and turns hard decisions into committee outcomes that belong to no one.

The play

For every active project in your company, assign one pilot in command, a single named person who owns the outcome and can be reached directly. Write that person's name into every status update, every meeting agenda, and every email thread about the project. When someone reports progress on anything without naming the owner, ask immediately: who is the pilot in command on this? Run that check for two weeks and you will surface every initiative that is currently drifting without an owner.

Tactic 06

Ship The Whole Cow

Jensen read The Innovator's Dilemma and took one lesson seriously above the others: the threat usually comes from the low end of the market. A cheaper competitor gets a foothold at the bottom, then climbs. So Jensen looked at what Nvidia was already throwing away. Chips that failed quality tests were not suitable for the company's flagship-grade products, but they were otherwise functional at lower speeds. That was a product line sitting in the trash. Nvidia repackaged those rejected parts into a cheaper, less capable derivative of their mainline chips. No new research. No new development cycle. The raw material cost was already sunk. Jensen described it plainly: rejected parts were generating no revenue. This was making something out of nothing. The strategy was named Ship the Whole Cow, a reference to how a butcher finds uses for almost every part of the carcass, not just the prime cuts. The commercial logic went beyond the margin on those cheaper chips. Nvidia could price its low-end line so aggressively that any competitor whose entire business depended on the low-cost chip would be forced to sell at a loss to compete. Jensen was not entering the budget segment because he wanted to. He was entering it to make it uninhabitable for anyone trying to use it as a ladder.

The play

Map your production or delivery process for outputs that currently fail your premium standard but still function at a lower tier. Repackage them into a distinct, cheaper product line rather than discarding them. Price that line at a level a pure low-cost competitor cannot sustain. The goal is not margin on the cheap version. The goal is removing the foothold that would otherwise let a low-end challenger climb toward your core market.

Tactic 07

Educate The Market On New Technology

When Nvidia created CUDA, they faced a problem that had nothing to do with the technology itself. The technology worked. The problem was that the people who needed it most, scientists and engineers who could use GPU computing for non-graphical applications, did not yet know what to do with it. Jensen's response was not to wait for the market to figure it out. He went and taught it. Nvidia's chief scientist began offering schools CUDA-capable machines if they committed to teaching a class on the subject. He gave more than 100 talks over the course of a single year. He taught a class himself at the University of Illinois. There was no textbook, so he wrote one. That textbook sold tens of thousands of copies, was translated into several languages, and was used by hundreds of schools. As Jensen saw it, the goal was saturation: the more people who had CUDA in their hands, the faster the technology would establish itself as a standard. Jensen said it plainly: push this everywhere and make it a foundational technology. This is not a new idea, but Jensen executed it at a level few companies ever have. Intel ran the same play when they invented the microprocessor. As Jensen put it, the goal of investing in CUDA was to expand Nvidia's reach into every corner of the tech industry. You cannot do that if the people at the corners do not understand what you built. The lesson is not just that Jensen believed in CUDA. It is that he understood belief alone does not create a market. Education does.

The play

When you introduce a genuinely new technology or product category, budget for market education as aggressively as you budget for the product itself. Identify the upstream group that needs to exist before your customers exist, whether that is students, researchers, or practitioners. Get your technology into their hands directly: offer machines, sponsor courses, give talks, write the missing textbook if one does not exist. Jensen's team gave over 100 talks in one year and wrote a textbook from scratch. The goal is saturation, not awareness. The more people who have direct hands-on experience with your technology, the faster it becomes the standard and the harder it becomes for a competitor to displace it.

Tactic 08

Recruit By Telling The Truth About Your Ambitions

When Jensen wanted to hire the chief engineer from Silicon Graphics, he did not prepare a polished pitch deck or ease the person into a conversation about culture fit. He invited the engineer to lunch and told him directly: "John, you should really think about coming to our company because ultimately I'm going to put SGI out of business." Most founders soften their pitch when recruiting. They emphasize upside, downplay competitive tension, and try to make the candidate feel comfortable. Jensen went the other way. He told a target recruit exactly what he planned to do to the company that recruit currently worked for. It was not a threat dressed up as an opportunity. It was an honest statement of intention, and the honesty itself was the pitch. If you believe in the mission enough to say that out loud to someone whose current employer is in the crosshairs, you are signaling conviction, not arrogance. This directness runs through everything Jensen does. He describes his own approach as going for the direct path, which is probably the only path he knows. The same quality that leads him to criticize a failed chip at an all-company meeting leads him to tell a prospective hire the unvarnished truth at a lunch table. Consistency between those two moments is what makes the signal credible. When a founder says "I will put your current company out of business" and then actually tries to do it for three decades, the words carry weight.

The play

The next time you recruit someone you genuinely want, skip the polished pitch and say the true thing. Tell them what you are building, what you intend to displace, and why you believe you will win. Name the company or category you are going after. If your conviction is real, say it plainly. Candidates who are drawn in by that honesty are the ones who will stay when the work gets hard. Candidates who need a softer story will leave when reality arrives.

Tactic 09

Build A Flat Organization You Can Drive Directly

Jensen has 60 direct reports and does not do one-on-one meetings. That is not an accident or an oversight. It is a deliberate architectural choice he made after watching what bureaucracy and politics did to a small company that should have been nimble. He thought carefully about what an ideal organization would look like from scratch, and he kept arriving at the same answer: flatter, always flatter. The logic runs in two directions at once. A flat structure speeds up information. When there are fewer layers between Jensen and the edges of the company, decisions move faster and the people closest to the work feel more empowered to act. Jensen put it plainly: "It turns out by having a lot of direct reports, not having one-on-ones, we made the company flat and information travels quickly and therefore employees are empowered." The second effect is self-selection. A flat structure quietly removes people who are waiting to be told what to do. Those who cannot think independently, who need a manager to hand them their next task, find the environment uncomfortable and eventually leave. The structure does the filtering. Jensen has resisted every push to change this, including from new board members who recommended he hire a chief operating officer to reduce his administrative load. His answer was direct: no thanks. He sees his management structure the way a racing team sees its car. "The company's organization is like a race car. It has to be a machine that the CEO knows how to drive." The point is not to copy what other companies do. The point is to build a company you can manage directly, one that fits the way you actually think and work.

The play

Map your current reporting structure and count the layers between you and the people doing the work. Identify one layer that exists primarily to pass information up and down rather than to make decisions. Remove it or collapse it, and have the people beneath it report to you or your direct team instead. Hold one open group meeting with that expanded set rather than individual check-ins. The goal is not to overwhelm yourself. It is to make sure information from the edge reaches you without being filtered, softened, or delayed by a layer of management that is optimizing for its own position.

Tactic 10

Choke You With Gold

Jensen wants to work with the best people in the world, and he treats compensation as a direct expression of that belief. He looks at his stock like his blood. He pours over the stock allocation reports himself, not delegating that review to HR or a compensation committee. Managers can refer an employee for special consideration to senior executives, and Jensen will then review the list personally and issue one-off grants that vest over a four-year period. When the grant is approved, the employee receives an email with a subject line that reads, "Special grant authorizing the RSU grant in recognition of your extraordinary contributions," with a clear description of the rationale behind the award. What separates this from a standard bonus program is the timing. Jensen can reach down into the organization at any moment and award stock directly, without waiting for an annual compensation review. The reward lands when the work is done, not six months later during a review cycle the employee barely remembers. That immediacy is the point. It sends a signal that someone at the top of the company is paying close attention and that great work will be recognized in the moment it happens. The Medici family once wrote to a sculptor they were trying to recruit, concluding a letter with the line, "Come, I will choke you with gold." Jensen operates the same way. The goal is not to be fair in the aggregate. It is to make your best people feel that working here is the highest-value decision they can make.

The play

Identify your top two or three contributors from the past 90 days. Do not wait for a performance review cycle. Write a one-paragraph rationale for each, naming the specific work and its impact on the company. Issue a meaningful equity or cash award now, with that rationale attached in writing so the person knows exactly why they are receiving it. The recognition tied to a specific action reinforces the behavior you want repeated.

Tactic 11

Swarm Your Greatest Opportunity

The idea that turned Nvidia into a multi-trillion dollar company did not come from a boardroom strategy session. It came from researchers who had hacked their GPUs to run non-graphical computations and were reporting significant speed improvements over CPU-only machines. Jensen picked up on this early. He hired the person who discovered it. Then he made it a priority to make it easier for others to find more non-graphical applications for GPUs. The underlying principle he acted on is one he and I have seen repeated across the best company histories: anytime you make something easier for people to do, the market expands. To act on that principle, Jensen's team built a new programming model called CUDA, compute unified device architecture, which made it possible and easier for scientists and engineers to use the GPU's computing power. The investment was not cheap. Nvidia converted its entire GPU line for CUDA compatibility at a cost that dropped gross margins from 45 to 35 percent. This happened in 2007 and 2008, during the global financial crisis, while Nvidia's stock fell more than 80 percent. Investors demanded a strategic course correction. Jensen refused. "I believe in CUDA," he said. "We are convinced that accelerated computing would solve problems that normal computers couldn't. We had to make that sacrifice. I had a deep belief in its potential." When developers did not yet know what to do with CUDA, Nvidia educated the market. Jensen's chief scientist gave more than 100 talks over the course of a year, taught a class himself at the University of Illinois, and wrote a textbook because no textbook existed. That textbook sold tens of thousands of copies, was translated into several languages, and was used by hundreds of schools. When Jensen's own executive team pushed back on deep learning, arguing it was a passing fad, Jensen overruled them. "Deep learning is going to be really big. We should go all in on it." When he announced the shift at a company all hands, he was direct: "We need to consider this work our highest priority." The entire arc runs from a 2002 research paper on simulating cloud movement to a multi-trillion dollar company. It played out over two decades, through an 80 percent stock decline, against the advice of investors and lieutenants alike. As Jensen put it at a later all hands: "AI was going to be more important than anything else they could possibly be doing."

The play

Identify the single opportunity in your market that, if it works, dwarfs everything else you are doing. Commit to it with disproportionate resources even if it compresses margins, draws investor criticism, or conflicts with short-term revenue targets. Educate the market yourself if adoption is slow: run workshops, publish content, train developers, do whatever it takes to get the technology into more hands. Then hold the course. The window where conviction separates you from the field is exactly the window when the pressure to quit is highest.

Tactic 12

Complacency Kills Mantra

Jensen believes that Nvidia's worst enemy is not the competition, but itself. That belief is not a sentiment he keeps private. He repeats it constantly, in company meetings, in interviews, in every period of the company's history covered in the book. The message is the same every time: complacency is the thing that kills successful companies, and success is precisely when it takes root. All the way back in 1997, when Nvidia was a fraction of the size of its rivals, Jensen would stand up at each monthly company meeting and tell employees, 'We're 30 days from going out of business.' He wanted the urgency to feel real, not hypothetical. At the same time, he was telling his team that Intel, then 860 times larger than Nvidia by revenue, had told its own employees they were going to put Nvidia out of business. Jensen's response was direct: 'Our job is to go kill them before they put us out of business.' The pattern holds across decades. Jensen resists flattering accounts of Nvidia's own history. When asked about the early years, he said, 'When we were younger, we sucked at a lot of things. Nvidia wasn't a great company on day one. We made it great over 31 years.' He slips into the third person when assessing his own early tenure: 'If Jensen wasn't even involved in the first 15 years of our company, I would really like that.' He was not proud of how the company was managed then, or of his own naive and lack of strategic thinking. That kind of self-assessment, applied publicly and consistently, sets the tone for how the whole organization thinks about its own record. You do not celebrate past success at Jensen's company. You treat it as a data point and move forward. This is not a posture Jensen performs at town halls and then drops in private. He says the inner voice telling him that nothing he does is good enough is impossible to satisfy, and it reads that way throughout the book. The same executive who watched Jensen blow out a quarterly earnings result recalls Jensen standing in front of the company afterward and saying, 'I look in the mirror every morning and say, you suck.' The point is not self-flagellation for its own sake. Jensen sees complacency as the specific mechanism by which successful companies lose their edge, and he treats the fight against it as an ongoing operating requirement, never a problem that gets solved.

The play

Pick one meeting this month, a team standup, a quarterly review, an all-hands, and say out loud that the company's biggest threat is its own success. Name a competitor who is moving faster. Give a number, a timeline, a concrete consequence. Do not frame it as motivation. Frame it as a factual risk assessment. Then make it a fixture, not a one-time speech. Jensen repeated 'we're 30 days from going out of business' at every monthly company meeting. The repetition is the mechanism. Urgency that gets said once becomes wallpaper. Urgency that gets repeated at every cadence becomes culture.

Your action plan

the plays, one checklist.

01
Run the Top Five Email System
Ask every person on your team to email you their top five priorities plus one market observation. Action verb first on each bullet, department tag in the subject line. Read them yourself. Run this for four weeks straight. You are hunting for the weak signal that shows up in two or three places before anyone says it out loud.
02
Set the Speed of Light, Not the Historical Average
Take one active project. Map every task and estimate each one assuming no delays and full focus. Add them up. That is the speed of light. Compare it to your real timeline. Every hour of difference is a named problem, not schedule padding. Assign someone to each gap.
03
Criticize Publicly So the Whole Company Learns
Next time a mistake or missed standard comes up, name it in the room instead of handling it privately later. Say what the standard was, what happened instead, and what good would have looked like. Then ask whether anyone else is facing the same situation. One person's mistake becomes a lesson for everyone.
04
The Lua Warning Signal
Choose a reset word for your meetings. Use it the moment someone's answer drifts from the actual question. Require them to stop, restate the question, and answer it directly. Precision in communication is a standard, not a preference.
05
Pilot In Command Accountability
Assign a named pilot in command to every project. Require that person's name to appear in every update. When someone says 'the team is working on it,' stop them and ask who owns it.
06
Ship The Whole Cow
Identify what your process rejects but could still sell at a lower tier. Package it, price it to hurt low-cost competitors, and defend the bottom of your market before someone else uses it as a ladder.
07
Educate The Market On New Technology
Budget for market education as aggressively as the product itself. Get your technology into the hands of the upstream group that has to exist before your customers exist. Give talks, sponsor courses, write the missing textbook. Jensen's team gave 100+ talks in a single year and wrote a textbook from scratch. Saturation creates the standard. The standard creates the moat.
08
Recruit By Telling The Truth About Your Ambitions
Tell your next top recruit exactly what you intend to displace and why you will win. Name the competitor. If your conviction is real, say it plainly. The candidates who stay after hearing that are the ones worth having.
09
Build A Flat Organization You Can Drive Directly
Count the layers between you and the people doing the work. Collapse the one that exists only to pass information up and down. Replace individual check-ins with a single group session so information from the edge reaches you unfiltered.
10
Choke You With Gold
Identify your top two or three contributors from the past 90 days and issue a meaningful award now, with a written rationale naming the specific work. Do not wait for a review cycle. Recognition tied to a specific action reinforces the behavior you want repeated.
11
Swarm Your Greatest Opportunity
Find the opportunity that dwarfs everything else and allocate maximum resources to it, even at the cost of margins and against investor pressure. Educate the market yourself if adoption is slow. Hold the course through the period when quitting feels rational, because that window is exactly what creates the moat.
12
Complacency Kills Mantra
Say out loud at a recurring meeting that the company's biggest threat is its own success. Name a competitor, give a number, and treat the statement as a factual risk assessment, not a motivational speech. Repeat it every meeting. Urgency said once becomes wallpaper. Urgency repeated every cadence becomes culture.