Jensen Huang Resigns From Nvidia After 33 Years; Tech Leaders Demand Reinstatement and Float Sabotage Theory

Jensen Huang Resigns From Nvidia After 33 Years; Tech Leaders Demand Reinstatement and Float Sabotage Theory

On November 16, 2026, Nvidia announced the resignation of CEO Jensen Huang, ending a 33-year tenure that began when he co-founded the company in 1993; after-hours trading swung sharply on the news. Within days, prominent Silicon Valley investors, AI founders and early Nvidia shareholders publicly called for his reinstatement and amplified an unverified claim that the scandal was a deliberate plot against him. This piece argues that whether or not the allegation holds, Huang's departure moves

The timeline is tight. On the evening of November 16, Nvidia disclosed in a press release that Huang had stepped down as CEO and from the board, effective immediately. The statement gave no specific reason beyond "disagreements with the board over governance matters." The Verge AI broke the story at 23:17 that night, describing a move that sent shockwaves through the tech industry after more than three decades. Huang conceived the company in 1993 with Chris Malachowsky and Tench Coxe at a Denny's in San Jose, took it public in 1999, and launched CUDA in 2006, turning the GPU from a graphics chip into a general-purpose parallel computing platform that became the de facto standard for AI training.
Within days, former board members and founders of several AI unicorns spoke out on X and on podcasts, pressing the board to reconsider. The faster-spreading narrative was a second one: that

Huang's exit is magnified because of where Nvidia sits. In fiscal 2025 the company reported $130.5 billion in revenue, $115.2 billion of it from data center, roughly 88%; GAAP gross margin was about 75%; and in October 2025 it became the first company to cross a $4 trillion valuation. More than 80% of global AI training and inference runs on the CUDA ecosystem, and the capex plans of the four largest cloud providers are anchored to Nvidia's product roadmap. A CEO change at such a company is not an internal affair; it is a scheduling problem for all AI infrastructure.
The deeper backdrop is structural tension in governance. Nvidia is a textbook founder-centric company: Huang was

Liu Gong's take: within 12 months, Nvidia is unlikely to be run again by a single founder-figure who combines technical authority with external storytelling; the board will likely settle on an interim structure of co-CEOs plus a strong CTO, and that structure will slow the definition of at least one product generation.
From a technical standpoint, the value of a founder-CEO is not day-to-day management but roadmap trade-offs. Nvidia has compressed architecture cycles to roughly one year — Hopper, Blackwell, Blackwell Ultra, Rubin — each carrying simultaneous bets on process node, packaging and interconnect. Those bets need one person to decide and absorb the consequences; committees drift toward the safer option. In industrial machine vision, where Liu Gong works, the same pattern shows up: once an algorithm roadmap shifts from point breakthroughs to system engineering, dispersing decision rights reliably stretches delivery cycles, because every step waits for consensus.
On the business side, the real risk is not the share price but the ordering rhythm of customers. By Nvidia's own framing, GB200 NVL72 delivers roughly 30x the inference performance of H100 — but that compares a 72-GPU rack against an 8-GPU server. The cost is power: a rack moves from roughly 10 kW to roughly 120 kW, making electricity and liquid cooling the binding constraint. When procurement shifts from buying cards to buying racks plus power plus thermal design, the counterparty set expands from Nvidia alone to Nvidia, the ODM and the data-center operator. With no permanent CEO, signature cycles on those three-party contracts lengthen. Liu Gong predicts that within six months, at least two hyperscalers will shift part of their 2027 volume to in-house silicon or a second supplier as a hedge against governance uncertainty.
For the industry, the spillover is a repricing of the founder premium. For a decade, public markets paid up for founder control because it bought long-termism; when one person becomes a systemic risk, that premium turns into a discount. TSMC's 2018 handover from Morris Chang to a dual-leader structure, which carried the company through 3nm and 2nm, is the closest reference — with one difference: TSMC's roadmap is tightly constrained by customer demand, while Nvidia's is self-defined, and a self-defined roadmap depends far more on one decision-maker's appetite for risk.

  • Tenure: Huang ran Nvidia from April 1993 to November 2026, about 33 years and 7 months — longer than Bill Gates's 25 years at Microsoft and Jeff Bezos's 30 years at Amazon.
  • Market value: Nvidia became the first company to cross a $4 trillion valuation in October 2025; at that scale, a 1% move in the stock equals roughly $40 billion.
  • Revenue mix: fiscal 2025 revenue was $130.5 billion, of which $115.2 billion came from data center, about 88%; gaming contributed roughly $11.4 billion, about 9%.
  • Margins: fiscal 2025 GAAP gross margin was about 75%, up roughly 2 points from 72.7% a year earlier — rare for a hardware company without monopoly protection.
  • Position: Nvidia held about 82%

For the industry, the near term of three months brings a holding pattern in supply chain and orders, with cloud providers demanding firmer roadmap commitments and delivery guarantees. Over 12 months, if a successor cannot hold the annual architecture cadence, AMD's MI line and customer ASICs gain a real share window, and the decline in cost per unit of AI compute may slow.
For ordinary users, the next three months should leave consumer graphics launches and pricing largely untouched, since gaming is only about 9% of revenue. Over a year, cloud inference price wars may cool because of supply uncertainty, narrowing the expected drop in AI service prices and leaving smaller developers with

Questions worth exploring:

  • Whether founder-centric governance still works for compute-infrastructure firms, contrasting TSMC's dual-leader model with Nvidia's single-decider structure
  • Whether CUDA's lock-in survives independently of the CEO, and where switching costs for developers actually end
  • The disclosure boundary around board governance disputes, and how 8-K and press-release language shapes how investors price uncertainty

Further reading: 2026 Frontier LLM Comparison: A Practical Guide — a durable, evergreen reference. Bookmark it.

This item comes from The Verge AI (published 2026-11-16 23:17:34). Wenfeng Radar provides a Chinese/English summary and commentary on overseas AI developments; the original content is owned by its author.


Daily aggregation of overseas AI developments with in-depth analysis. Bookmark us to never miss an important signal; back to homepage.

Republishing: feel free to quote or repost with a link back to the original article on Wenfeng Radar. · Business & partnership: see our media kit
Daily briefing + Sunday review · Unsubscribe anytime by replying · Your email stays private

Leave a Reply

Your email address will not be published. Required fields are marked *

中文EN
Contact · Privacy · Disclaimer · AI Disclosure · Weekly Review · Media & Business
Email: wenfengleida@wenfengleida.cn