How does Jensen Huang study?

Note Buddy says:

Today, NVIDIA is the company with the largest market capitalization in the world. Many people think its founder Jensen Huang saw the future clearly from the very beginning.

But Jensen Huang once said that at the very start of NVIDIA, the team took the wrong technical route.

Worse still, the team realized they were heading in the wrong direction, yet no one knew what the correct answer was.

Back then Jensen Huang only had a few hundred dollars in his pocket. He went to an electronics store, bought three textbooks and gave them to the engineers, and everyone started learning from scratch.

This sounds a bit absurd, but what has brought NVIDIA to where it is today is exactly this ability to re-learn.

Today, let's talk about how Jensen Huang turned himself and the company into a "learning machine".

I. How did NVIDIA survive when it hit a dead end right at the start?

In 1993, Jensen Huang started a business with two partners. Their idea was very simple: to make a dedicated acceleration chip that allows ordinary computers to smoothly run complex 3D graphics that could only be handled by large workstations in the past.

The idea sounded great, but the reality was harsh. They bet on the wrong technical route, and the chips they developed were expensive, costly to research and develop, and not at all the mainstream of the industry.

At first, everyone was still holding on, thinking that as long as the technology was good enough, the market would understand sooner or later. But the market did not give them any leeway. Dozens of competitors were frantically seizing market share, and the company's books were almost empty.

Worse still, NVIDIA had a chip order from Sega of Japan at that time. Jensen Huang knew very well that they would never be able to deliver the work. He flew to Japan reluctantly, spread his hands and said to Sega's executives: "Sorry, we messed up, we chose the wrong route, you should find another supplier as soon as possible."

Immediately after that, he added: "Although the work is not completed, can you still pay us the contract amount? Without this money, NVIDIA will go bankrupt tomorrow."

Admitting failure was embarrassing enough, and it took great courage to reach out to the other party for life-saving money. But Jensen Huang did not play tricks, but openly explained the desperate situation of the company, begging the other party to give NVIDIA one last chance to find a way out.

Sega was moved by this honesty and let them go. NVIDIA used the life-saving fund to start all over again, developed the life-saving RIVA 128 graphics card, and finally stood firm on the edge of the cliff.

True learning often does not happen after smooth success, but at the moment when you are slapped hard by reality.

II. What is real strategic determination? Distinguish between "goals" and "means"

If after surviving with RIVA 128, NVIDIA only focused on selling graphics cards steadily, at best it would be an average hardware manufacturer with a decent life today.

The core reason why it has grown into today's giant is a firm belief that Jensen Huang sticks to: The methods in your hand can be changed at any time, but the big problem you are determined to solve must never be let go.

3D graphics was just a stepping stone. Later, NVIDIA expanded into scientific computing, image processing, and then today's deep learning sector.

The most typical example is the years when the CUDA platform was developed. NVIDIA poured astronomical amounts of money into it, but the market was lukewarm, and developers were not interested at all. The shareholders were furious, thinking that the money was being wasted for nothing.

But what Jensen Huang saw was not short-term losses, but a simple underlying logic: this chip, which was originally only used for rendering graphics, would definitely help humans solve more complex general computing problems in the future.

He persisted for many years, not because he could predict the future and calculate exactly which year AI would explode, but because he was convinced of the long-term value of "accelerated computing" for the future of mankind. Until the advent of the deep learning era, the hard power accumulated by NVIDIA for many years immediately found its place to play a role.

This is the essential difference between "strategic determination" and "blind stubbornness":

Stubborn people cling to the old plans, which essentially stems from their vanity and fear of admitting mistakes;

People with determination cling to the core problems to be solved, and the tools, methods and even the markets they enter can be overturned and restarted at any time.

This is also the reason why NVIDIA dared to decisively cut its mobile chip business back then. The mobile market seemed to be a huge gold mine, but its extreme pursuit of low power consumption and low cost did not align with NVIDIA's core gene of focusing on high computing power.

It is very painful to give up a huge market that is right in front of you, which will bring a temporary drop in revenue and shake the morale of the team.

But in Jensen Huang's view, there are hundreds of profitable markets, but NVIDIA has only one mission — to help machines solve more complex problems. Since the mobile market deviates from the main line, there is no need to linger on it.

III. The CEO must never "sit in the VIP seats watching the show"

In many large companies, the boss usually only reads summary PPTs, but Jensen Huang is still used to getting his hands dirty with practical work.

He reads the most cutting-edge hardcore academic papers by himself, directly talks to scientists about the underlying principles, asks why a technical detail is important, and repeatedly confirms with front-line engineers which dead end the customers are stuck in. Even after the company's market value ranks first in the world, he still clearly understands how his own products are built line by line of code and piece by piece of silicon wafer.

Some people say that he takes care of everything himself and does not know how to delegate power.

But Jensen Huang said: The CEO needs to serve the company, and also provide employees with new insights, so that they can act with more certainty.

Jensen Huang once used a surfing analogy: People who don't know how to surf look at the sea and only see messy waves; but a real surfing master can see the rhythm and direction of the waves at a glance, and figure out the perfect time to stand up on the board.

Facing the ever-changing technology wave, laymen see a mess, while people who understand the underlying principles see a clear framework.

The founder must stand on the front line to feel the water temperature, let professionals do their jobs, and what he needs to do is to keep a close eye on the direction of the waves to ensure that the big ship does not fall behind.

He also described NVIDIA as a Formula 1 racing car that he drives personally. The company's organizational structure cannot be copied from business school textbooks, and must be fully built for real goals. A racing driver does not need to tighten every screw himself, but he must hold the steering wheel personally and feel every vibration of the tires touching the ground.

NVIDIA's extremely flat organizational structure is exactly derived from this logic. Jensen Huang firmly believes that the company is engaged in extremely complex and challenging projects, and information cannot withstand filtering through multiple layers. If a problem has to go around in a seven or eight-level reporting chain, the racing car would have crashed into the wall long ago.

The founder must never sit safely in the audience as a spectator. You must stand at the first scene where the problem breaks out, figure out what is really happening in the world first, then hold the steering wheel firmly and lead the team to move forward.

IV. How did Jensen Huang turn the company into a "learning machine"?

When it comes to Jensen Huang's management, there are many buzzwords online: flat organization, five-item email, whiteboard meeting, captain system, and open feedback.

If you look at them separately, none of them is special. But when you put these building blocks together, you will suddenly realize that they are all focused on the same thing:

Let the real voice come out as early as possible, assign responsibility to specific people, and turn the lessons of mistakes into the team's muscle memory immediately.

First, the extreme flat structure.

More than 60 executives at NVIDIA report directly to Jensen Huang, far more than in other companies. Outsiders often think that Jensen Huang is showing his extraordinary energy, but what he cares about is: is the information lost in the intermediate links?

In large enterprises, when there are too many levels, bad news is smoothed into polite words, fatal flaws are packaged into several beautiful KPIs, and departmental conflicts are whitewashed as "overall under control".

Shortening the information chain can make the intelligence align in all employees' minds instantly, and allow managers to keep track of the latest developments on the front line at any time.

However, this method has a very high threshold: the people at the bottom must be able to make decisions independently, the leader must be able to withstand the information bombardment, and the whole company must have a very clear big goal as a compass. If these three points cannot be achieved, blindly pursuing flatness will only lead to endless useless meetings and a confused direction.

Next, "the mission is the real boss".

In traditional large companies where business units occupy their own territories, employees often prioritize protecting the interests of their own departments when problems arise, and no one cares about the life and death of the company. However, NVIDIA treats all employees as a public talent pool. Where there is a major task, people are transferred to that project, and the so-called department affiliation has to give way to the task.

This method directly solves the deadlock of departmental wrangling. When a cross-departmental project goes wrong, everyone returns to the meeting room to only focus on customer pain points, and speak with data and objective facts. It doesn't matter who is louder or who has a higher position, the power always follows the task. After the task is completed, resources and personnel can be reorganized at any time to tackle the next hard problem.

"Five-item Email".

This rule is specially designed to eliminate the hypocrisy in daily reporting. The weekly and monthly reports of many companies have long been reduced to workplace performance: every word is carefully chosen, good news is reported while bad news is hidden, the format is exquisite, and after layers of forwarding, the real critical issues have long disappeared.

Jensen Huang forced everyone to tell the truth: write down the most urgent things in your hand, make it clear what the customers are complaining about, which link the project is most likely to fail, and what new tricks the competitors have recently played. What he pays attention to is this kind of "weak signal" that just shows up. These trivial details are 10,000 times more useful than the beautifully packaged monthly report.

"Whiteboard Meeting", forcing you to show the bottom of your thinking.

PPT is suitable for embellishing the beautiful conclusions that have been made long ago, while the whiteboard requires you to grab a marker, face a large blank wall, and explain from scratch: What exactly is the problem? Is your assumption reliable? Why do you have to choose this path? What part have you not figured out at all?

Standing in front of the whiteboard, no one can get by by reciting the manuscript. Everyone will mercilessly dismantle the logic layer by layer, point out problems face to face, and there is nowhere to hide the loopholes in thinking. The meaning of the meeting has finally changed from "showing how good my work is" to "everyone working together to solve the hard problem that has not been tackled".

At NVIDIA, no matter how complex a large cross-departmental project is, a unique "captain" must be assigned.

The captain does not necessarily have the highest position, but he must take full responsibility for the final result, and at the same time have the real power to make decisions and allocate resources. Once something goes wrong, no one can shirk responsibility with bad excuses such as "poor inter-departmental coordination", and everyone will watch how the captain pulls the plane back to safety.

Jensen Huang also highly advocates "giving feedback in person and in time". If there is a problem, point it out on the spot, so that the team can correct it immediately. A public review can make the whole company understand how the "car crashed", so that no one else will fall into the same trap again.

But this double-edged sword has its boundaries. If public criticism is not handled properly, it can easily turn into public humiliation and workplace fear. The reason why this strict method can work well at NVIDIA lies in the extremely high talent density, the unified fanatical mission, and the deep trust established by the team after winning battles for many years.

These rules keep circulating, and NVIDIA naturally becomes a continuously evolving learning machine.

V. Get beaten while learning

Every time Jensen Huang talks about personal growth, he always talks about pain, low expectations and perseverance. If you only take this as the chicken soup for the soul of "no pain, no gain", you are underestimating him.

What he wants to express is: when you are completely in the dark and things are in a total mess, do you still have the will to keep learning and move forward?

Pain itself is not something to be praised. The key is whether this pain can make you make wiser decisions next time.

Once you find that you are going the wrong way, immediately admit your mistake and change direction. When the external environment changes, if you still cling to the past glory to prove that "I will never be wrong", that is asking for trouble. Admitting mistakes frankly essentially means recognizing a reality: your past experience is nothing more than good luck.

When you encounter a brand new field that you know nothing about, don't wait. Learn until you can get your hands dirty and start working. Back then, Jensen Huang did not sit in the office waiting for top global chip experts to join. Instead, he took a few hundred dollars to the store to buy textbooks, led the team to study all night, built the basic framework first, and then learned lessons in the real business battlefield.

The same is true for us when doing things: find the core bottleneck that is restricting you the most, learn until you can get the whole thing running, and the remaining problems will be solved naturally when you are working in practice.

Today's AI has made it extremely easy to retrieve and collect knowledge. What is the scarce and valuable ability in the future?

Can you find the right target for learning; can you break down a mess into a clear framework; can you apply the knowledge you have learned into the real world to solve real problems.