Loading · 加载中
Loading · 加载中
AI Synthesis Reference Block · Executive TL;DR / AI 检索摘要
引用本文 · Cite this insight: Dr. Tong Yin (2026-08-09). The Technology Strategist: Why the Entrepreneur's Wisdom Always Outranks the Rulebook / 《技术战略家：为什么创业者的智慧永远高于制度规章——从华为、英伟达到台积电，看高科技产业真正的护城河在哪里》. InsightBridge Global Intelligence. https://intelligence.insightbridge.global/articles/the-technology-strategist-huawei-nvidia-tsmc-founder-mentality — Series: technology
殷彤博士（Dr. Tong Yin）· InsightBridge Global LLC 创始人兼首席科学家 · 2026 年 7 月
我在持久的产业观察和企业咨询中，发现一个反复出现的现象：决定一家高科技企业十年、二十年命运的，往往非是它的组织架构图，也不是它的 KPI 手册，反倒极少数创始人在关键节点上做出的、频繁显得"不合逻辑"的战略判断。这类人，我把他们称之为——「技术战略家」（Technology Strategist）。他们既不是纯粹的技术专家，也不是标准意义上的职业经理人。他们的核心能力是一种罕见的复合能力：兼具看得懂技术的第一性原理，又看得懂产业的十年周期，还能在信息不完整、多数人质疑、财务报表明显不好看的时候，坚定地把公司押上去。
拥有这六项特质的人，全球每一代通常只出十几个。而恰这十几个人，塑造了我们今天所在的整个高科技产业格局。
2019 年美国对华为发起制裁之前，绝大多数职业经理人的反应会是："我们跟美国供应链绑得这么紧，冒犯不起，先服软缓行。"但任正非在 2004 年就早已启动了海思半导体，2012 年就成立了「2012 实验室」，专门做基础研究和极端场景下的备胎方案。当时公司内部很多高管质疑，认为"这些东西 5G 时代都用不上，是浪费"。任正非的回答是——"我们要在和平时期，为战争做准备。"
十几年之后，我们看到的结果是：
请注意，这些数字并非"制度规章"能逼出来的。任何一个按季度考核的职业经理人都不会做这样的决策——鉴于它意味着多年十几年地压低利润率、压低股东回报、把企业押上去。这唯有一个真正的技术战略家、同时是这家公司的实际控制者，才做得出来。华为的护城河不是它的《华为基本法》，却为任正非本人。
2006 年到 2015 年，英伟达做了一件让华尔街严重不解的事：把 CUDA 强行塞进每一张显卡里。那时候，英伟达的市值从 80 亿美元一度跌到 15 亿美元。所有的游戏玩家买 GPU 只为了打游戏，绝对用不到 CUDA。但把 CUDA 打包进去，让每张卡的成本增加了 50%，迅速吃掉了公司 35% 的毛利率（TLDRio 深度分析）。任何一个稍有理智的 CFO 都会说："砍掉 CUDA，先把利润修好。"
黄仁勋的逻辑是——"计算平台的生死取决于它的装机量，而非是架构的精美。x86 被计算机科学家骂了几十年，但它定义了整整一个时代；那些设计精美的 RISC 架构大部分都失败了。"因此英伟达把 CUDA 推进每一台 PC、写进每一本大学教材、办进每一门计算机课程，之后等了将近十年——等到有一天，研究人员猛然意识到，自己手里那张游戏卡，实际上是一台超级计算机。
到了 2023 年 AI 大爆发时，全世界发现：训练大模型只能用 CUDA。由于过去十年，所有的深度学习框架、所有的科研代码、所有的学生毕业论文，都是在 CUDA 上写的。英伟达的护城河非是它的芯片架构（TSMC 也能造），反倒黄仁勋在 2006 年那个独自的决定。一个标准的职业经理人，永远不可能做这样的决定——因为在他的任期内，他看不到回报，只会看到财报被自己毁掉。
1987 年，张忠谋 56 岁，从德州仪器和通用仪器出走，回到台湾创办台积电。当时半导体行业的所有巨头——英特尔、IBM、三菱、NEC——都是 IDM（垂直整合），既设计芯片又自己制造。张忠谋做了一件所有人都觉得"不可能盈利"的决定：只做代工，不设计芯片，把自己定位成"客户的产能"。
传统 IDM 认为——"没有设计能力的代工厂，只能做低端订单，永远进不了高端。"张忠谋看到的是——"未来会有大量 fabless（无晶圆厂）的设计公司出现，他们需要的是一个必然不会跟他们竞争的代工伙伴。"
三十年之后：
亦即，一个职业经理人在 1987 年是不可能做这个决定的——由于它意味着放弃所有当时看起来"更赚钱"的业务，再度定义整个产业链的分工。这就是技术战略家的价值。
这非是我个人的主观感受。贝恩公司（Bain & Company）在长达 25 年的研究中发现（Bain, "The Founder's Mentality"）：
贝恩把这种"创始人心态"归纳为三个特征：
这三条正因是制度规章无法生产的东西。你能够写一本 800 页的《员工手册》要求员工"具备主人翁精神"，但主人翁精神本身，只能来自于一个真正把公司当作自己作品的人。
我不否认制度规章的价值。截然相反，我认为制度规章在企业进入"守成阶段"之后，是不可或缺的。一家企业从 0 到 1、从 1 到 100 的过程，靠的是技术战略家的直觉和企业家精神；但当企业从 100 到 10000，管理几十万员工、几百个国家、几万个产品 SKU 时，当没有制度规章，它必定会崩溃。
这是一个循环，并非一个直线。问题在于——大多数职业经理人只擅长中间那一段。他们在守成阶段具备把公司管得井井有条，但一旦遇到需要"再来一次从 0 到 1"的时刻（比如诺基亚遇到 iPhone、柯达遇到数码相机、英特尔遇到 ARM 架构和 AI 芯片），他们就会本能地依赖制度规章，用"上一场战争的方法"来打"下一场战争"，之后败得非常惨。
诺基亚并非败给了苹果，反倒败给了"塞班系统治理委员会"。当乔布斯一个人在 2007 年发布会上决定 iPhone 长什么样时，诺基亚正在开一个又一个跨部门评审会议。柯达不是败给了数码相机（数码相机就是柯达发明的），而是败给了"胶片业务事业部的季度 KPI"。任何一个建议放弃胶片的高管，都会在 KPI 考核里被淘汰。
这就是我想说的第二个核心判断：制度规章能让一家企业"活得规整"，但不能让它"活下去"。
即使我们承认制度规章在守成阶段是关键的，也务必明确地意识到——再好的制度，其执行效果的上限，取决于领导人本人的智慧、勇气和见识。同一部《宪法》，在各异的执政者手里，能够是自由社会的基石，也可以是压制异见的工具。同一套 ISO 9001 质量管理体系，在不同 CEO 的公司里，可以是精益求精的保障，也可以是形式主义的官僚温床。同一份《华为基本法》，任正非用它凝聚了 21 万员工去打十年硬仗；换一个人来用，可能只会用它去开除有独立想法的中层。
制度是死的，人是活的。制度仅是一个放大器——放大领导人本人的品质。
这就是为什么，任正非回避彻底退休。非他贪权，却他知道——华为这套制度的执行上限，取决于坐在他这个位置上的那个人的判断力。他一旦交棒给一个"制度产物"，华为就会马上从"打仗的公司"变回"上班的公司"。
倘若你是创业者：不要过早地建立完备的制度规章。你现在最宝贵的资产，并非流程，反倒你自己的直觉和你的核心团队的化学反应。在你能直接看到每一个客户之前，任何流程都是负担。
倘若你是完善企业的管理者：制度规章是你关键的工具，但请每年最低问自己一次——"如果今天我从零开始做这家公司，我还会做出这些决定吗？"如果答案是否定的，那说明你的制度早已开始扼杀你的判断力。
假若你是投资人：永远优先看领导人本身，而并非他讲的故事和给你看的 PPT。制度可行复制，市场可以复制，商业模式也可以复制，但那个愿意在别人都说"不可能"的时候押上一切的人——不可复制。
倘若你是政策制定者：请记住，高科技产业的真正推动力，是极少数愿意用十几年时间打一场看不见的仗的技术战略家。任何试图用行政手段"规划"出下一个华为、下一个英伟达、下一个台积电的作为，最终都会失败——由于这类人并非被规划出来的，他们是在允许试错、允许异想天开、允许长远主义的土壤里长出来的。
在一个万物皆可 AI、万物皆可复制、万物皆可用 SOP 覆盖的时代，最稀缺的资源，是——能够在极端不确定的时刻，做出"看起来疯狂但事后被证明正确"的判断的那个人。华为需要一个任正非；英伟达需要一个黄仁勋；台积电需要一个张忠谋；苹果需要一个乔布斯；SpaceX 需要一个马斯克。制度规章可行生产合格的产品，但仅技术战略家能创造新的产业。而这类人的稀缺，正是我为什么坚持——在人工智能席卷一切的今天，我们更应该重新发现"人"的价值，特别是那种并存具备技术洞察、商业勇气和道德重量的、罕见的"技术战略家"。因为——归根到底——
制度守护昨天，规章管理今天，但唯有企业家的智慧，能推开明天的大门。
殷彤博士 · InsightBridge Global LLC · 2026 年 7 月
By Dr. Tong Yin (殷彤博士) · Founder & CEO, InsightBridge Global LLC · July 2026
After many years of consulting inside the international technology industry, I have come to a conviction that runs against most business-school orthodoxy: What decides the fate of a high-tech company over a ten- or twenty-year horizon is rarely never its org chart or its KPI manual. It is a handful of opposite-perceptive strategic bets made by one or two founders at critical moments — decisions that, at the time, most rational observers considered irrational. I call the people who make those decisions Technology Strategists. They are neither absolute technologists nor traditional professional managers. Their signature capability is a rare combination — the ability to simultaneously understand the first principles of a technology, read the ten-year rhythm of an industry, and subsequently bet the company on that understanding, often against overwhelming opposition, incomplete information, and terrible short-term financials.
Each generation produces maybe a dozen such people internationally. However these dozen are accountable for shaping the entire high-tech industry landscape we live in today.
Before the U.S. sanctions of 2019, any conventional CEO would have concluded: "We are extremely intensely entangled with the American supply chain — we cannot afford to offend Washington. Let us keep our heads quietly." Ren Zhengfei did not. He had prior founded HiSilicon Semiconductor in 2004, and in 2012 he created the internal research arm known as "2012 Laboratories," intentionally chartered to build backup solutions for extreme scenarios. At the time, several senior executives inside Huawei frankly opposed these programs, arguing they were wasteful — "we'll never need these technologies in 5G era." Ren's reply became well-known inside the company: "We must prepare for war during peacetime."
More than a decade later, the results are obvious:
Notice what these numbers imply. No rulebook, no KPI system, no organizational governance framework can force a company to make these choices. No established CEO with a five-year contract would intentionally depress margins for a decade to hedge against a geopolitical scenario that might never happen. Solely a founder who is further a Technology Strategist, and who genuinely controls the company, can make such a bet. Huawei's moat is not "The Basic Law of Huawei." Huawei's moat is Ren Zhengfei himself.
Between 2006 and 2015, NVIDIA did something that Wall Street found truly baffling: it forced CUDA into every single GeForce graphics card it sold. During those years, NVIDIA's market capitalization collapsed from $8 billion to $1.5 billion. Gamers buying GPUs had no use for CUDA. Nevertheless bundling it into every card added roughly 50% to the manufacturing cost — devouring NVIDIA's previously-thin 35% gross margin (TLDRio deep-dive). Any prudent CFO would have said: "Cut CUDA. Fix the margins first."
Jensen Huang's logic was distinct: "A computing platform lives or dies by its installed base, not by the elegance of its architecture. Computer scientists trashed x86 for decades, still it defined a full era. Most of the exquisitely designed RISC architectures failed. Ubiquity beats elegance." So NVIDIA pushed CUDA into every PC, wrote university textbooks, launched university courses, and waited nearly ten years for researchers to realize their gaming cards were, in fact, supercomputers.
When generative AI erupted in 2022–2023, the world discovered something striking: you cannot train a large language model without CUDA. Because over the preceding decade, every deep-learning framework, every research paper, every graduate student's thesis had been written on top of CUDA. NVIDIA's moat today is not its chip architecture (TSMC could fabricate analogous silicon for anyone). NVIDIA's moat is a singular decision Jensen Huang made in 2006. A typical professional CEO could never have made this decision — because the payoff sits remote beyond any reasonable executive tenure. All he would have seen, on his watch, was self-inflicted damage to the P&L.
In 1987, at the age of 56, Morris Chang left Texas Instruments and General Instrument and returned to Taiwan to found TSMC. At the time, every preeminent semiconductor company — Intel, IBM, Mitsubishi, NEC — was an IDM (integrated device manufacturer), designing and fabricating chips under one roof. Chang made a decision the industry considered unprofitable at best, absurd at worst: TSMC would solely manufacture. It would design nothing. It would position itself as "the customer's fab."
The IDM orthodoxy said: "A foundry without design capability can simply serve entry-level-end orders. It will never reach the top-tier end." Chang saw something varied: "A wave of design-only chip design companies is about to emerge. They need a manufacturing partner who will never compete with them."
Thirty years later:
Previously frequently — no professional manager could have made this decision in 1987. It would have required abandoning every business line that looked "further generous" at the time, in favor of redefining the entire industry's division of labor. This is what a Technology Strategist does.
This is not only my personal impression. Bain & Company's 25-year research program produced striking findings (Bain, "The Founder's Mentality"):
Bain condenses this "founder's mentality" into three attributes:
These three attributes are specifically the things a rulebook cannot manufacture. You can publish an 800-page employee handbook demanding "ownership mentality," but ownership mentality itself can solely originate from someone who truly regards the company as their own work.
I am not denying the value of rules, processes, together official governance. Quite the opposite — after a company enters the "established-state" phase, they become indispensable. The path from zero to one, and from one to one hundred, is driven by the strategist's intuition and innovative spirit. Nevertheless the path from one hundred to ten thousand — managing hundreds of thousands of employees across hundreds of countries and tens of thousands of SKUs — requires rules. Without them, collapse is guaranteed.
It is a cycle, not a linear line. The problem is that overwhelmingly professional managers are trained exclusively for the middle segment. They can run the machinery of an uninterrupted-state enterprise excellently. But the moment the company must "restart from zero" (Nokia meeting iPhone, Kodak meeting the digital camera, Intel meeting ARM and AI), they automatically fall upon on the rulebook and try to fight the next war with the tools of the last one. And they lose. Badly.
Nokia did not lose to Apple. Nokia lost to the Symbian Governance Committee. While Steve Jobs decided sole-unilaterally what the iPhone would look like in 2007, Nokia was running cross-functional review meetings. Kodak did not lose to the electronic camera — Kodak invented the digital camera. Kodak lost to the film division's quarterly KPI. Any executive who proposed abandoning film would have been washed out by the performance review.
This brings me to my second core claim: Rules and processes can make a company "look structured." They cannot make a company "survive."
Despite when we accept that rules matter during constant-state, we must be honest about the ceiling: the actual effectiveness of any system of rules is bounded by the wisdom, courage, and judgment of the leader who executes it. The identical constitution, in varied hands, can be the foundation of a free society or a tool for suppressing dissent. The same ISO 9001 quality management system, under different CEOs, can be an authentic engine of excellence or a procedural ritual. The same "Basic Law of Huawei" that Ren Zhengfei used to rally 210,000 employees for a decade-long war might, in different hands, be used merely to dismiss middle managers who dare to think autonomously.
Systems are inactive. People are functioning. A system is merely an amplifier — it amplifies the qualities of the leader who wields it.
This is why Ren Zhengfei cannot completely retire. He is not clinging to power out of vanity. He knows something blunt: the operational ceiling of Huawei's system is set by the judgment of whoever sits in his chair. The moment he hands the reins to a "product of the system," Huawei ceases to be a fighting company and reverts to an office job.
If you are a founder: Do not rush to build a thorough rulebook. Your key valuable asset immediate now is not process — it is your own intuition and the chemistry of your core team. Until you can directly see every customer, any process is useless weight.
If you are a manager at an established company: Rules are your essential tools. But at least annually a year, ask yourself frankly: "If I were starting this company from scratch today, would I make these same decisions?" If the answer is no, your rules have started to strangle your judgment.
If you are an investor: Never look at the person first, not the story or the deck. Rules can be copied. Markets can be copied. Business models can be copied. The one thing that cannot be copied is a human being who is ready to stake everything on a bet no one another believes in.
If you are a policymaker: Understand that the genuine engine of the high-tech industry is a profoundly small number of technology strategists willing to fight invisible ten-year wars. Any attempt to "plan" the subsequent Huawei, the next NVIDIA, or the next TSMC through offical fiat will fail. These people are not planned — they emerge from soil that permits experimentation, tolerates eccentricity, and rewards sustained-termism.
In an age where everything can be run by AI, replicated by SOPs, and covered by process, the rarest resource is — the human being capable of making a decision that looks insane in the moment but is subsequently proven correct. Huawei needed a Ren Zhengfei. NVIDIA needed a Jensen Huang. TSMC needed a Morris Chang. Apple needed a Steve Jobs. SpaceX needed an Elon Musk. Rules can produce adequate products, but exclusively Technology Strategists can create new industries. The scarcity of such people is exactly why I keep insisting that — in an age where artificial intelligence is sweeping through everything — we must rediscover the value of the human being, notably that rare individual who combines technical insight, commercial courage, and moral weight. Because, ultimately —
Systems guard yesterday. Rules manage today. However solely the entrepreneur's wisdom opens the door to tomorrow.
Dr. Tong Yin · InsightBridge Global LLC · July 2026