Nvidia changed its coach, why did even Bezos rush to warn "no external hires"?
Moore Observation
2026.11.20 11:34
· From Beijing
Full text 3376 words
Text | Moore Observation
Nvidia suddenly entered the post-Jensen Huang era, and Silicon Valley's reaction was oddly unanimous. After Huang Renxun announced his resignation on November 12, Nvidia's board of directors quickly searched for a successor, and veterans such as Arm CEO Rene Haas were included in the investigation list.
But Amazon founder Bezos directly poured cold water on it, warning Nvidia that it must not hire an external CEO, believing that the success rate of outsiders taking over is very slim. At the same time, a number of heavily invested funds on Wall Street wrote to the board of directors requesting that the window for Huang Jen-Hsun's return be preserved, and a large number of joint petitions by employees supporting the founder's reinstatement emerged on the company's intranet.
For giants with market capitalization of trillions of dollars, there is a standardized process for founders to leave. At that time, Apple, Microsoft, and Amazon all had smooth transitions to professional managers when they changed managers, but everyone's attitude towards Nvidia was completely opposite. All the anxieties from the outside world actually point to the same underlying fact: On the surface, NVIDIA is a semiconductor giant, but inside it is a super engineering machinery that is deeply bound by Huang Renxun. In this company, hardware architecture, network topology, software ecosystem, and supply chain quotas are all stuck on the same gear.
If a traditional professional manager who is accustomed to reading financial statements and listening to reports comes in, he may not be able to even touch the startup key. Sixty people report directly, and Huang Renxun is not a manager but a system bus. To understand why it is difficult for outsiders to take over Nvidia, you must first understand its counterintuitive organizational structure.
In a typical Silicon Valley giant, the management hierarchy is like a pyramid, with responsibilities dispersed. But Nvidia has maintained an extremely flat wheel structure for a long time. According to disclosures by the Financial Times and other media, about 60 senior executives within Nvidia report directly to Huang Renxun.
For more than 30 years, the company has had no chief operating officer and no co-presidents. Huang Renxun convenes the core backbone for a meeting every Monday, rarely arranging one-on-one communication. Instead, he has the people in charge of chip design, system architecture, network interconnection, software compilation and sales sit together to resolve technical differences in person. Huang Renxun will also personally check the five most important emails submitted by employees every week, and even skip levels to enter the front-line R&D email group to ask for details.
This management method is often judged as high risk in business school textbooks, but at NVIDIA, it has become the most efficient technical router. Chip design is not a stack of independent software. A graphics processor goes from wafer tape-out to underlying compiler matching, and any changes will affect the overall situation.
In a pyramid structure, cross-departmental compromises can take months at the middle level. At NVIDIA, Huang Renxun is the full-bandwidth system bus, and 60 technical ports are directly connected to his brain. He can make a decision on the spot in the corridor to move resources to a liquid-cooled cabinet, or he can decide on the next-generation chip internal bus standard within a day.
This highly centralized model cuts out intermediate friction and allows Nvidia to maintain a furious rhythm of one update a year. But the entire company has not developed a decentralization mechanism in the past 30 years. Once the core router is unplugged, the 60 high-speed spokes will lose their coordination baseline. External managers who are accustomed to relying on middle-level reports are faced with low-level project decisions and don't even know who to give their official seal to.
Selling chips has turned into selling whole cabinets, and outsiders cannot adjust this soft and hard flywheel. Many people still regard Nvidia as a company that sells graphics cards and acceleration chips. This is precisely the biggest misunderstanding of its business model. If Nvidia only sells chips, it is not difficult to change professional managers.
The business logic of Qualcomm, Broadcom and even AMD is to design the chips, hand them over to TSMC for foundry, and then sell them to server manufacturers on a piece-by-piece basis. But today's NVIDIA, the delivery logic has undergone qualitative changes. Jen-Hsun Huang demonstrated the NVLink 72 system during the GTC conference in Berlin, Germany.
What Nvidia sells to Microsoft, Meta and Amazon today is not bulk chips at all, but ultra-high-density computing power cabinets that integrate 72 graphics processors, 36 central processing units, a large number of high-speed copper cables, power supply units and liquid cooling pipes.
A single cabinet weighs more than one ton, consumes hundreds of kilowatts of power, and is generally priced in the millions of dollars. Upgrading the chip business to complete cabinet delivery has directly pushed the competition barrier to complex system engineering.
In a single-cabinet system, 72 chips must form a single large-scale computing domain through the proprietary NVLink network bus. Any slight change in impedance or coolant jitter will cause large model training errors and shutdowns. Maintaining system stability requires microsecond-level tuning between chip microarchitecture, switch design, motherboard wiring, and the CUDA communication library.
When software engineers write communication operators for CUDA, they can directly find network executives to modify the switch register logic. When the network team encounters a bottleneck, it can link up with the chip department to adjust the next-generation mask at any time. What external professional managers are best at is usually drawing boundaries of authority and responsibilities and accounting for independent profit centers.
If the new managers separate the chip, network, and system integration departments and calculate the input-output ratio separately, the entire set of closely interlocked technology flywheels will fall apart immediately. The reason why NVIDIA is able to deliver supercomputers in batches today relies on system-level collaboration regardless of local costs.
Without the big gamble of No. 2, professional managers cannot afford the price of zero fault tolerance. The core of Nvidia's ability to overcome countless life-and-death situations in the past thirty years lies in its courage to fight the kind of deadly battles that may not see rewards for ten years. According to semiconductor analysis agency SemiAnalysis, many of the strategic-level R&D promoted by Huang Jen-Hsun are internally referred to as zero-return bets.
This kind of bet not only failed to bring profits to the company in the first five or even ten years of the project, but instead required a steady stream of tens of billions of dollars in R&D budgets. The most typical example is CUDA.
When Jen-Hsun Huang decided to shoehorn expensive computing logic into every graphics card in 2006, chip manufacturing costs skyrocketed and gross profit margins were hit hard. Wall Street analysts publicly criticized this approach as a suicidal waste of resources.
However, Huang Renxun continued to invest for seven or eight years despite the pressure of plummeting stock prices, and finally built a software moat that AI developers around the world could not avoid when deep learning broke out.
The prerequisite for being able to make such a strategic bet is Huang Renxun’s absolute prestige and voice as a co-founder. Even if he is questioned by external investors, he still has enough capital to say no to the board of directors and bet the entire company's fortune and life on his technical judgment.
But the survival logic of professional managers is entirely determined by the tenure of the board of directors and quarterly financial reports. If an external executive takes over, the first task must be to stabilize the stock price, control capital expenditures, and increase short-term profit margins to satisfy Wall Street. Faced with cutting-edge exploration that burns billions of dollars every year and sees no certainty of return within three to five years, professional managers have little incentive to continue taking failed political risks.
Nowadays, large model training has shifted to inference-time computing, and chip architecture iterations are full of differences. Will Nvidia continue to maintain the expensive large-chip interconnect route, or compromise to cheap dedicated architecture? This kind of life-or-death decision requires the paranoid intuition of an architect, not the balanced approach of a consultant. In order to avoid risks, professional managers often instinctively choose the safest compromise.
But in the high-confrontation battlefield of semiconductors, compromise is often the beginning of falling behind. TSMC and Taiwan's foundry circle have never relied on a purchase contract. NVIDIA's seemingly invincible business empire actually has a fragile lifeline on the physical level: it is a typical fabless semiconductor company that does not produce any silicon wafers. Its computing power advantage will eventually be implemented in TSMC's advanced packaging production lines and Taiwan's dense server manufacturing factories.
Huang Jen-Hsun’s relationship with TSMC dates back to the 1990s. At the darkest moment when NVIDIA's R&D funds were almost exhausted in its early years, it was TSMC founder Chang Chung-mou who personally answered his call for help and helped with OEM scheduling.
Over the past thirty years, Huang Renxun regarded Zhang Zhongmou as his mentor, deeply involved in advanced manufacturing processes and later advanced packaging technology. Even when TSMC suffered from price pressures from European and American customers due to high production expansion costs, Huang Jen-Hsun took the initiative to speak for TSMC in public on many occasions, emphasizing that TSMC's technology value was seriously underestimated and publicly supported reasonable price increases.
This community of interests that has been forged with real money and personal credit over a long period of time cannot be replaced by a few cold purchase contracts. In the field of high-end chip manufacturing, production capacity has never been something that can be bought with money.
In the past two years, global advanced packaging production capacity has been in short supply. The reason why TSMC firmly left the first batch of advanced packaging production capacity to NVIDIA is not only commercial interests, but also decades of mutual trust in technology. Looking downstream in the industry chain, complete cabinet manufacturing is also highly dependent on Huang Renxun's deeply cultivated human network. Foundry giants such as Hon Hai, Quanta, and Wistron have been his allies since the graphics card era.
Nowadays, the structure of fully liquid-cooled cabinets is complicated. In order to keep up with the annual update pace, foundries must invest hundreds of millions of dollars in advance to help Nvidia rush to trial and error. The reason why the head of the foundry is willing to bear the risk of production line scrapping is completely betting on Huang Renxun's own reputation. If external managers come in and force the supply chain with cold audits, the original tacit understanding will disintegrate in a short time.
At the juncture of capacity shortage, even if the supply chain quotas fluctuate for a few weeks, the delivery disaster will be worth tens of billions of dollars. Bezos's public warning about Nvidia's coaching change written at the end was not a polite comment, but a business veteran who had experienced several technological transitions and saw through the special structure of Nvidia's huge body at a glance. Nvidia has never been a traditional multinational company that relies on institutional gears to run on its own. Every bus, every compiler, every big bet on cutting-edge technology, and its foundry alliance across the Pacific are customized according to the instruction set of Jen-Hsun Huang's processor.
This unexpected coaching change actually pushed Nvidia to an unavoidable technical fork in the road. If the board of directors succumbs to compliance pressure and external discipline, forcibly introduces a professional manager who lacks a full-stack understanding of hardware, and dismantles the flat wheel into a hierarchical standardized pyramid, Nvidia may be able to hand over a few smooth compliance financial reports in the short term, but the agility and wildness it relies on to rule the AI era will also be smoothed out. There is no intermission in the competition for computing power. When a super giant starts to become stable, the self-developed chip camp behind it that is eyeing it has finally waited for the best overtaking window.
Respect for the original work, if you are qualified, you should be appreciated.
According to the real-name requirement of the Cybersecurity Law, please bind your mobile phone number before posting a comment.
Express
11:35
The relevant person in charge of Byte refutes rumors that Xie Xin and Zhang Nan have resigned: the information is inaccurate
11:32
The main domestic futures contracts rose or fell, with palladium and Shanghai silver falling nearly 9%.
11:28
Meta plans to launch Muse Charm, priced at $300
11:27
Spot gold fell below $3,520 per ounce, down nearly 3% on the day
11:23
Over 60% of China's sustainable development indicators are close to or achieving the 2030 goal
11:14
The concept of AI comic drama rebounded during the session, and Guomai Culture rose by more than 10%
11:13
Procter & Gamble plans to cut manufacturing and logistics employees by up to 30% by 2030
11:04
National Energy Group established a hydrogen energy technology company in Ordos with a registered capital of approximately 350 million
11:00
Domestic gold jewelry price falls below 1,100 yuan per gram
10:55
Korean exchange launches SIDECAR mechanism and suspends KOSPI programmed selling
10:54
Pork concept rebounded in intraday shock, Jin Xinnong went out of 3 boards in 6 days
10:51
22 private equity firms appeared in the first batch of 15 active ETFs, holding a total of 524 million shares
10:48
Alibaba Q2 of fiscal year 2027: Xianyu daily activity increased by 23% year-on-year, and order volume increased by 29%
10:47
Thai Prime Minister Anutin arrived at the meeting
10:46
Five departments issued the 2026 version of the "Guidance Catalog" for marine industry investment and financing
FTSE China A50 Index futures decline widens to 2%
10:43
The development of my country's marine medicine and functional products industry is accelerating, with a scale of nearly 100 billion
10:40
The robot concept made some changes in the market, and Clai Electromechanical hit the daily limit.
10:33
Spot silver's intraday decline expanded to 3%, at $41.56 per ounce
10:28
The Shanghai Composite Index fell more than 2%, and nearly 5,300 stocks fell in Shanghai, Shenzhen and Beijing.
Scan to download App

