Nvidia Has No Five-Year Plan: Jensen Huang Built the Company Into a GPU
Jensen Huang's eight counterintuitive management rules aren't scattered points but a distributed operating system: 60 executives reporting directly, pay identical to the cent, no one-on-ones, no planning — the organization's communication structure determines its product architecture.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
60 direct reports, because the word "manage" is already wrong
The normal management span is a dozen or so people; 60 reporting directly to the CEO looks impossible. Wang Yajun's breakdown: the problem is the word "manage" — managing means giving orders to execute. Jensen Huang's position is that these people don't need managing; each of them does their own thing exceptionally well, and all they need is for him to create opportunities for them. These are two different management logics. Wang Yajun also went through a phase at Longfor where about 50 people reported directly to the CEO with no layers in between, so he says he wasn't surprised at all. But he admits it does raise a problem: what about developing people?
— Wang Yajun60 people paid the same, down to the cent
Jensen Huang's analogy: in an Excel spreadsheet there's a number representing a person's pay; you click the bottom-right corner and drag it down to everyone — these 60 people's pay is the same, accurate to the dollar or the cent. Wang Yajun sees this as matching a sense of equality with output value: anyone who can be placed in a role reporting directly to the CEO is a star who can run their own show; roles differ but magnitude is the same, so they should share a common sense of equality. More critically, in interactions no one is above or below anyone else; when discussing a problem you address the problem, not "you're more senior so you decide."
— Wang YajunNo one-on-ones; feedback happens in the room
Many management scholars stress that bosses should learn to have one-on-one conversations, because feedback and emotional issues are best handled privately. Jensen Huang does the opposite: he gives feedback on the spot. When different views emerge in a meeting, or some way of thinking has a problem, he won't pull that person aside; he'll say it right there so everyone hears. His logic is that this isn't criticism, it's learning — besides learning from your own mistakes, the best thing is to learn from other people's mistakes. If another 10 or 20 people hear it at the same time, the speed and frequency of learning is much higher. Wang Yajun says the precondition is that people aren't obsessed with face all day long.
— Wang YajunThe organization is like a GPU: Conway's Law in hardware
A commenter online said Nvidia's organization "sounds kind of like a GPU," and the comment got a lot of likes. Wang Yajun says the metaphor is exactly right. He learned from an AMD executive that AMD's chips use a chiplet-style modular design rather than one monolithic design; a GPU goes further, with a large number of weaker cores plus enormous inter-core transmission channels and bandwidth. Abstracted to software, this is Conway's Law: an organization's communication structure determines its product architecture. So talking purely about "what organizational form is good" is meaningless; the organizational form, the product form and market demand have to reach a continuously evolving best match.
— Wang YajunThe most common mistake in learning from Nvidia is copying the motions
Wang Yajun gives the example of Youzan founder Bai Ya: Bai Ya visited Nvidia, saw a large venue, learned it was where Jensen Huang holds monthly all-hands meetings, and reflected that once a company gets big, processes, systems and institutions really aren't much use — the top person still has to personally pay attention to key things. The practice he drew from it was to go past his direct reports and see what the layer below is doing. Wang Yajun thinks this leans toward control, whereas what Jensen Huang does at all-hands meetings is provide the information needed for decisions, similar to what Zhang Yiming calls context, not control. The same all-hands format — talking about performance metrics versus talking about how to make decisions — is a completely different thing.
— Wang YajunNo documents; reading hundreds of employee emails a day
Jensen Huang doesn't encourage writing documents internally; instead he encourages employees to email him the five most important things, and he spends a stretch every morning reading hundreds of emails from employees at every level — anyone can message him. Wang Yajun guesses what he wants is a sense of the company's reality, the company's pulse — what is actually happening. But he stresses that not everything should be sent: telling him actual information is very valuable, but telling the boss your conclusions is unnecessary. He mentions Ren Zhengfei saying that a frontline employee who writes the chairman a ten-thousand-word article on company strategy is the kind of person the company cannot keep.
— Wang YajunNo layoffs only works if you believe people can grow
Without layoffs, how do you weed out employees who don't push forward? Wang Yajun says he was quite shocked when he heard it, but some great companies really don't proactively lay people off — Toyota, for example. He lays out several preconditions: first, you have to believe that both people and companies grow continuously, that people aren't rigid and fixed; second, you have to be more careful in hiring, with growth potential perhaps the first factor and professionalism second; third, you can't chase short-cycle opportunities to make quick money — hiring in a rush when demand surges and then having no choice but to lay off when demand falls. Jensen Huang's line is: I will torture you into greatness.
— Wang YajunIgnore market share; only do what others can't
Jensen Huang only does what others can't do; he doesn't look at market share, he goes after a Zero Billion Dollar Market — a market that doesn't exist yet, where he wants to create a new market. His exact words are: Market Share is for the losers. Wang Yajun connects this to strategic choice: if you copy others and rush to imitate, it's easy for everyone to pile on capacity, capacity comes in very fast, the point where capacity matches demand arrives quickly, and then you have to lay people off fast. He gives real estate as an example: many companies died because around the three-year mark industry-wide demand surged, everyone rushed scale in, and when it came down they simply couldn't play the game.
— Wang YajunNo planning, because you simply can't predict
Jensen Huang doesn't make five-year plans, three-year plans, or even one-year plans. Wang Yajun says he wasn't shocked at all, because that's how he operated before. He went to Longfor in 2012 and found that the real estate industry faces extremely high uncertainty — market, environment, regulatory policy, financial policy — making it impossible to infer how big the market will be in three years; plus the market is highly fragmented, so talking about market share is meaningless. Later the whole company reached a consensus: we can't predict the future, but we can respond quickly. So there was no five-year strategy, no three-year strategy, only a one-year operating plan. He says Longfor's strategy at the time was just one page, used for several years, with half a sentence or a third of a sentence tweaked each year.
— Wang YajunWithout love there is no innovation; fear is the enemy of innovation
In his last interview with the Stripe founder, Jensen Huang talked about one of the company's values being love and care. Wang Yajun says this isn't contradictory with rationality: innovation means facing the unknown and uncertainty, and it will fail many times before it succeeds; if innovation is a company's core capability, you must know how to manage failure, how to accept failure, and the cost of failure can't be too high. He uses the metaphor of a child holding a cup: you can give them an unbreakable cup, or you can give them a breakable cup so they learn, but if they break three or five cups you'll be exhausted — without love you can't do this. Many companies manage through fear and only add a bit of care when filling out ESG forms; the underlying logic isn't connected.
— Wang YajunA company isn't a barrel; it's a bamboo forest
Wang Yajun proposes an alternative to the barrel theory: the bamboo forest. Bamboo is segmented both above and below ground, but the roots below are all connected; a bamboo grove you see might be the life of a single bamboo that has grown dozens of stalks. He personally experienced trying to dig out a small bamboo grove and could only use a shovel — it was agonizing. Mapped to an organization, 60 people reporting directly to the CEO means the horizontal alignment between teams is a bamboo-forest-style organic connection, not a barrel — each function is a stave, hooked together with a few hooks, with a big board at the bottom as the CEO. He says horizontal alignment has a methodology; it can be taught, learned, and grown sustainably.
— Wang YajunIn their own words · checked verbatim
He said it's like in an Excel spreadsheet, you have a number, this number is a person's pay, and then you click the bottom-right corner and drag it straight down to everyone's spot — meaning everyone's pay can be identical to the dollar or the cent, and these 60 people's pay is the same.
他说就像在Excel表格里面 你有一个数 这个数是一个人的薪酬 然后你把他的右下角按一下 往下直接拉 拉到所有人的那个地方 意思是说 每个人的薪酬都可以精确到 一美元或一美分地的一样 这60个人的薪酬是一样的
Wang Yajun8:16
This is learning, this is not a criticism. Don't treat it as me criticizing you or reprimanding you, right? If you think of it as learning, then besides learning from your own mistakes, the best thing is to learn from other people's mistakes, right?
这是学习 这不是一个批评 大家不要把它当成一个 我来批判你批评你 对吧 如果你把它想成一个学习 那你除了从自己的错误里面 学习最好的是从别人错误里面 学习对吧
Wang Yajun10:17
What does Conway's Law say? It says an organization's communication structure will determine its product structure — its organizational structure, not just its architecture. How the organization operates, how information is transmitted, will influence and determine its product architecture.
康威定律是说什么呢 就是一个组织的它的沟通结构 会决定了它的产品结构 就它的组织结构 不只是架构 这是组织是怎么玩的 信息如何传递的 会影响 会决定它的产品架构
Wang Yajun14:19
So he'll say, I will torture you into greatness — I will torture you into greatness. And greatness really is tortured out of you.
所以他会说 I will torture you into greatness 我要把你折磨成伟大 那伟大真是被折磨出来的
Wang Yajun40:31
He only does the kind of thing others can't do. He doesn't look at Market Share, and he goes after that Zero Billion Dollar Market — a market that doesn't exist yet. I want to create a new market, so I don't talk about Market Share. Market Share is for the losers.
他只做那种别人做不了的东西 他不看Market Share 而且他去做那种Zero Billion Dollar Market 现在还没有的市场 我要去创造出新市场 所以我不讲Market Share Market Share is for the losers
Wang Yajun42:33
An enterprise capable of innovation — if there is no love in this enterprise, if no one talks about love, about care — it will not have innovation. Fear is the enemy of innovation.
一个能创新的企业 如果在这个企业里面没有爱 大家不会去讲爱 关爱 他是不会有创新的 恐惧是创新的敌人
Wang Yajun48:36
The roots underneath are all connected. The structure of bamboo below the soil and above the soil is all bamboo nodes, all segmented, but underneath it's all connected. So when you see a bamboo grove, it might be the life of a single bamboo that has grown dozens of stalks, but underneath they're all connected.
它底下的根全连在一起的 竹子在土下面的结构和在土上面的结构 都是竹节 都是一节一节的 但它底下它是连在一起 所以当你看到一个竹林的时候 它还有可能就是一个的竹子的生命 给它长出了几十根竹子 但底下都是连着的
Wang Yajun1:12:45
Figures
| Number of executives reporting directly to Jensen Huang | 60 | 4:13 |
| Number of Nvidia employees | 28,000 | 56:43 |
| Nvidia market cap | $2.62 trillion | 56:43 |
| Employee emails Jensen Huang reads daily | hundreds | 35:27 |
| Long-term market cap difference between high-trust and low-trust companies | 5x to 10x | 1:09:45 |
Glossary
- Conway's Law
- An organization's communication structure determines its product architecture.
- context, not control
- A management philosophy proposed by Zhang Yiming: leaders provide the information needed for decisions rather than issuing orders.
- Zero Billion Dollar Market
- A new market that doesn't exist yet and needs to be created.
- chiplet
- A modular chip design that breaks a chip into multiple modules assembled together, rather than one monolithic design.
How to listen
Founders, CEOs and HR leaders building or restructuring their organizations; investors curious about Nvidia's management methodology but who have only seen scattered anecdotes.
The final extension on the trust formula and the bamboo forest metaphor can be skimmed; the first eight points are enough.