Chinese Tech Gurus

All twenty-one lives in Volume One open on the same chapter: family background and early years. That is the house style, and it is also the fact of the matter. Cao Dewang turned his origins into an asset. Zhang Ruimin took a sledgehammer to seventy-six refrigerators and turned the wreckage into a quality standard. Ren Zhengfei spent the company's best years talking about nothing but failure. What they had in common was that there was no way back, which is why sheer endurance counted as a method.

The reason I succeeded is that I am grateful to have started from zero. Cao Dewang, Heart Like Bodhi, 2014

The twenty-one in Volume Two started from the opposite place. Peng Zhihui gave up the top salary band of Huawei's “Genius Youth” programme. Alexandr Wang dropped out of MIT at nineteen. Liang Wenfeng walked away from a quantitative fund that was already making money. What they gave up was not poverty but a good road already paved for them. Both generations bet everything; one had nowhere to retreat to, the other chose to leave. That shift explains every difference that follows — when the price of failure drops from “you do not eat” to “you go back to work”, the size of the bet you can place changes completely.

Each life in Volume One carries a chapter on national significance, and it is largely the same story each time: some link in the chain was blocked, and someone spent years unblocking it. Wang Dongsheng in display panels, Zhu Yiming in memory, Richard Chang and Liang Mong-Sung in wafer fabrication. Their lever was scarcity — build the thing nobody will sell you, and the value lies in being the only one who has it.

Until CXMT turns a profit, I will not take a cent of salary. Zhu Yiming, CXMT prospectus and related reporting, from 2018

Volume Two pulls the lever the other way. DeepSeek gave its models away as open source and got adoption by an entire industry in return; the value is not in being the only one who has it but in everyone using yours. Liang Wenfeng says so plainly.

AI is too big a thing for anyone to monopolize; those who try will be abandoned by history, so restraint is necessary. Liang Wenfeng

The line is not clean, and the counter-example sits inside Volume Two itself. Frank Wang's DJI took the entirely closed road, taking the great majority of the consumer drone market on integration it never let out of the building — he calls it the narrow gate. Since the volume grew to twenty-one, even the open-weights camp no longer agrees with itself: DeepSeek pressed its API price down to one yuan per million tokens while Moonshot AI placed its flagship in the highest price tier in China. So this is less a generational difference than a rule about diffusion: when a technology spreads faster than it can be protected, open source becomes the stronger lever. In hardware, scarcity still works.

Volume One keeps time in decades. Liu Chuanzhi, Zhang Ruimin, Hou Weigui and Wang Dongsheng all had to walk through one or more gates that very nearly did not open, and there was no way round them. Frank Wang was already the fast one: he started in 2006 from a flat in Lianhua Beicun in Shenzhen, and it took seven years to reach roughly 70% of the global consumer drone market by the end of 2013, and twelve to reach the 90% share of 2018.

Volume Two resets the scale to three years. AgiBot was founded in February 2023; on the Omdia figures the volume cites, it shipped more than 5,100 general-purpose humanoid robots in 2025, about 39% of the world market and first place, with revenue up nearly twentyfold on the year before. The time needed to do the same thing — become the best in the world at it — really has fallen by an order of magnitude. Not because this generation is cleverer, but because the supply chain, the open models and the capital were all sitting there when they started. The preconditions the previous generation had to build, this one simply called on.

But once the volume grew to six tracks it pulled that line apart itself. Three years is the clock of software, not of everyone: Zhang Changwu took eight years from registering LandSpace in 2015 to putting Zhuque-2 into orbit in 2023, and Frank Wang's first company was founded in 2006, seventeen years before Liang Wenfeng turned to large models — yet both sit in the same volume. What sets the pace is not the generation but how much of the work has to become a physical object.

The expansion also cost Volume Two its single voice. Asked when embodied intelligence turns the corner, one track holds at least three mutually contradictory answers.

The humanoid robot's 'ChatGPT moment' may still need 2–10 years. Once technology crosses this inflection point, robots can gradually enter large-scale commercial application. Wang Xingxing, Unitree Robotics

The embodied-intelligence industry may simply not have a “ChatGPT moment.” Gao Jiyang, Galaxea

Wang He lands between them: embodied intelligence is welcoming its own moment now, and by 2028, with ten times the training data, it will truly arrive. Three people, one track, one volume, three timetables. The five-person Volume Two read like a judgement; the twenty-one-person one reads like an argument — which is closer to what this actually looks like right now.

So the accurate statement is not that this generation moved ten times faster, but that its starting line was moved forward by the one before it — and moved by wildly unequal amounts depending on the track. The closer to software, the further it moved; the more the work has to become a rocket, a fusion device or a chip, the less it moved, and the more of Volume One's decade-long patience survives intact.

The starting point changed, the lever changed, the clock changed. One thing both generations describe in almost identical terms: deferring the payoff. Zhang Yiming makes it a test for judging people; Duan Yongping makes it a choice; Frank Wang makes it a gate.

Half the problems in many people's lives come from a lack of delayed gratification. Zhang Yiming

The gate that leads to success is narrow, and few find it. DJI's secret is precisely that it entered the ‘narrow gate.’ Frank Wang, DJI

Volume Two's own synthesis counts the instances: Frank Wang gave up building cars; Liang Wenfeng gave up 3D, video and world models; Peng Zhihui gave up Huawei's 2.01 million yuan salary; Demi Guo gave up the banner of artificial general intelligence; Zhang Changwu gave up non-reusable solid rockets after the failure of Zhuque-1. Five renunciations on five entirely different chains, sharing one arithmetic: shrink the boundary to buy depth. Every one of them happened at a moment when the company was not short of opportunities.

Peng Zhihui states the other half of the same thing — it is not only about holding on, but about knowing what to let go of: many people read persistence as never giving up, but selective abandonment matters just as much. What the two generations actually share is not hardship. It is a willingness to do one thing for a very long time, and to keep turning down everything else in order to do it. That is the one thing worth taking away from reading the two volumes together.