How Jensen Huang runs NVIDIA: 19 principles from three decades

Executive overview

Most companies die from complacency, not competition. Jensen Huang has spent over 30 years building NVIDIA by treating complacency as an existential threat — and designing every management practice to fight it.

The result is a system: flat structure, radical transparency, relentless speed, and long-horizon bets on markets that don't yet exist.

The core insight: character is built through pain and suffering, and great companies are built the same way.

The 19 ideas Jensen uses to run NVIDIA

1. Professor Jensen — Jensen sees teaching as his primary job. He communicates so consistently that two unrelated NVIDIA employees will describe the same strategy unprompted.
2. Whiteboard — Every meeting centers on drawing and talking at a whiteboard. It forces rigour, exposes unclear thinking instantly, and reflects Jensen's belief in constant reinvention — erase and start again.
3. Complacency kills — Jensen's deepest fear. He ran 1997 all-hands meetings telling employees "we're 30 days from going out of business." At that time, Intel was 860× NVIDIA's size in revenue. He told the team: "We need to kill Intel."
4. Flat organization — Jensen has 60 direct reports and no one-on-ones. A flat structure speeds decision-making, empowers employees to act independently, and filters out people who wait to be told what to do.
5. Public criticism — Jensen does not praise publicly and criticise privately. He calls out mistakes in all-hands meetings so the entire organisation learns. "We are not optimising for not embarrassing somebody. We're optimising for the company learning from our mistakes."
6. Tortured into greatness — Jensen applies the same standard to himself. After a quarter where NVIDIA "blew the doors off," he stood up and said: "I look in the mirror every morning and say, you suck." He calls resilience and the ability to endure pain his greatest superpowers — not intelligence.
7. Speed of light — Every project is broken into tasks with target times that assume zero delays, zero queues, zero downtime. That sets the theoretical maximum. NVIDIA measures itself against that ceiling, not against competitors or past performance.
8. Unapologetically extreme — Jensen works every hour. Any hour not working, he is thinking about work. He has no tolerance for complaints about long hours. "There may be people smarter than me, but no one is ever going to work harder than me." Second place is the first loser.
9. Top five emails — Every employee sends a weekly email: five bullet points, each starting with an action verb, covering what they're working on and what they're observing in the market. Jensen reads ~100 per day. It is his primary method of intercepting weak signals before they travel through management layers. "I want information from the edge."
10. Blunt, concise communication — Jensen's emails are "short and sweet, like a haiku." Long orders are the enemy of speed. Get to the point; make it memorable.
11. LUA — When an employee rambles, Jensen says "LUA": Listen, Understand, Answer. A warning that his patience is running out.
12. Mission is the boss / pilot in command — Every project has a named individual accountable for it. No hiding behind "such and such team." The mission is the authority — not hierarchy or org chart.
13. Strategy is action — No five-year plans. The world is a living, breathing thing. Plan continuously, adjust daily. "Strategy is not words, strategy is actions."
14. Ship the whole cow — Chips that fail top-tier quality tests are repackaged and sold cheaper. Every part of the carcass gets used. This defends the low end of the market against competitors who try to climb from below.
15. Go to school on everybody — Jensen showed up at an academic machine learning conference he wasn't speaking at. Reason: "I'm here to learn." He knows biographical details about employees across the company. "The good ones know more."
16. Create the market, don't fight for share — NVIDIA targets zero-billion-dollar markets: "Where there are no customers, there are also no competitors." Jensen announced the robotics push by saying they were building something "where we are sure there are no customers" — and the room laughed. He was serious.
17. Choke you with gold — Jensen treats equity like blood. He reads stock allocation reports personally, can reach down and award special grants at any time without waiting for annual reviews, and writes personalised rationale for each award. Exceptional contributors get recognised in the moment.
18. Work on your highest priority first — Jensen completes his most important task before the workday starts. "Before I even get to work, my day is already a success." Focus relentlessly on the handful of things that actually matter; let everything else go undone.
19. Swarm your greatest opportunity — When Jensen sees a once-in-a-generation bet, he goes all in, for as long as it takes, even under pressure to stop. The AI bet unfolded over two decades: first GPU paper for non-graphics in 2002, CUDA launched in 2007-8 (gross margins fell from 45% to 35%, stock fell 80% during the financial crisis), Jensen held the course. Then deep learning. Then AI. He overruled key lieutenants who called it a fad. "Deep learning is going to be really big. We should go all in on it."

On CUDA and educating the market

• CUDA (Compute Unified Device Architecture) made GPUs programmable for science, not just graphics.
• Jensen believed that making something easier to use always expands the market.
• When developers didn't know what to do with CUDA, NVIDIA taught them: offered universities CUDA machines in exchange for teaching courses, ran over 100 talks in a year, wrote the textbook (no textbook existed).
• Intel did the same when they invented the microprocessor — ran more seminars than local colleges.
• The result: as more people learned CUDA, demand for GPUs compounded. Competitors faced a network they could not break into.

On building for the long term

• "Start, scale, sell" is a cultural trap. Jensen's model: find, build, retain, own for decades.
• The CUDA investment that now underpins a multi-trillion-dollar company began paying off 15+ years after Jensen first committed to it.
• Jensen prefers "strong, self-reinforcing network" to "moat" — because believing you have a moat breeds complacency.
• The last line of the book: "There are no shortcuts. The best way to be successful is to take the more difficult route and the best teacher of all is adversity."