Jensen Huang's Resignation: What Changes at Nvidia

Analysis Written Before the Resignation, Analysis Written After

On November 12 (local time), Nvidia’s board announced Jensen Huang’s resignation. He stepped down from his roles as president, CEO, and director, effective immediately. CFO Colette Kress will serve as interim CEO while continuing as CFO, and the board has said it will consider both internal and external candidates for a permanent CEO.

The announcement didn’t state a reason for the resignation. Here’s what led up to it. On November 8, Huang acknowledged an extramarital relationship in a message to employees, and that same day the board said it was launching an independent investigation into reports that he had been targeted by foreign intelligence activity. At the time, the board said it had no reason to believe national security or company secrets had been compromised. On the 11th, The Wall Street Journal reported that the board was preparing to remove him, followed by a New York Times report on another relationship with an executive who reported directly to Huang. According to a U.S. Securities and Exchange Commission (SEC) filing, the board accepted the resignation on the 11th.

Over these few days, I read two analyses. One was “What Happens to Nvidia If It Loses Huang,” published by Tencent Technology (腾讯科技), a Chinese outlet. It was written before the board’s decision and identified Kress stepping in as interim CEO as the most likely scenario — which is exactly what happened a few hours later. The other was a five-point analysis that TF International Securities analyst Ming-Chi Kuo posted on X after the resignation was announced.

Both pieces start from the same premise: the business fundamentals aren’t seriously shaken. Where they diverge is on where the change will show up first. Tencent Technology focused on the organization and decision-making speed, while Kuo focused on advance orders in the supply chain and demand from China. I checked both pieces’ claims against the official announcement, the SEC filing, and how the market reacted on the 12th, point by point. I won’t go further into the personal-life controversy itself.

Why a Founder Could Still Be Removed by the Board

The Tencent Tech piece starts with ownership. Per Nvidia’s 2026 proxy statement, Huang’s beneficial ownership sits at roughly 870.6 million shares, or 3.58% of the total (as of March 23, 2026). BlackRock (7.43%) and Vanguard (7.31%) each hold roughly double that. There’s only one class of common stock, one vote per share. There’s no dual-class voting structure¹ that stacks extra votes behind the founder.

The board’s makeup points the same way. As of July, there were 11 directors, and only one of them — Huang — came from inside the company. The chairman’s seat was vacant, with an independent lead director filling that role instead. The founder has led the company for 33 years, but in terms of votes and seats alone, this was a company where a board majority could replace the CEO. The piece frames this as Nvidia being, in substance, a professionally managed company — and notes that Apple, Microsoft, Amazon, and TSMC all grew larger after their founders left.

A while back, in Issue 349, I wrote about how Anthropic, SpaceX, and Meta each chose structures that protect founder control. Nvidia sits on the opposite end of that spectrum. Had the same thing happened to a founder holding dual-class voting power, the board would have had far fewer tools at its disposal.

What Stays: The Roadmap, CUDA, and 42,000 People

The core argument of the Tencent Tech piece is that we should separate Huang’s personal role from the company’s competitiveness. Nvidia’s edge, it argues, comes from over 30 years of GPU-focused expertise, the CUDA² ecosystem, and its R&D organization working together—and the product roadmap is already mapped out several years ahead. The people stay, too. Chief Scientist Bill Dally, CTO Michael Kagan, and Ian Buck, the creator of CUDA, are all still there. As of the end of fiscal year 2026 (January 25, 2026), the company has about 42,000 employees, of whom 31,000 are in R&D.

The logic on the customer side is the same. Microsoft, Meta, Google, and Amazon buy Nvidia GPUs because of their own AI investment plans, the chips’ performance, and software built around CUDA. A change in CEO is a weak reason to cut back orders.

On the pace of technological progress, the piece diverges from the assessment Ming-Chi Kuo gave before his resignation. Kuo believed Huang’s value lies in being able to see strategy, organization, technology, and supply chains all at once and make fast decisions. The piece accepts that assessment but disagrees that Huang’s departure would slow the industry’s technological generational shift. Just as a server rack drawing tens of kW—or pushing past 100kW—eventually forces a change in power delivery design, the forces driving generational change are demand and physical limits, not any one person. As evidence, it points to AMD, Broadcom, and Google’s TPU³ all moving along the same technical path.

Board Chair Harvey Jones said much the same in his statement, expecting operations, customer commitments, and the product roadmap to continue without interruption. Up to this point, the piece and the company are saying the same thing.

What’s Changing: A Reporting Line That Used to Converge on One Man

What’s at stake is the org chart. Nvidia has no Chief Operating Officer. According to an org chart Business Insider reported in October 2025, 36 executives report directly to Huang — down from 55 before that. Depending on the outlet, the number is sometimes cited as around 60. Whatever the figure, this was a structure built around one person standing at the center to make decisions fast, and there was no designated No. 2 to step into that seat if he were gone.

That’s why Tencent Tech predicted the company would need a reorganization — installing a new COO and cleaning up the reporting lines. The announcement on the 12th contains what reads like the first step in that direction. The board named Harvey Jones, a director since 1993, as the company’s first-ever board chairman. No COO position or reporting-line overhaul has been announced yet.

Kress has served as CFO since September 2013, and before that spent roughly 25 years in finance at companies including Cisco and Microsoft. Tencent Tech’s assessment is that her familiarity with key customers and Wall Street makes her well-suited to oversee the transition. It’s also conceivable that internal technical staff hold down product and customer relationships while a permanent CEO is brought in from outside. Several names are circulating as candidates, but none have been confirmed.

On the question of when this reorganization will actually reach the product line, Ming-Chi Kuo offered a concrete year. His view: the appointment of a permanent CEO and the organizational changes that follow won’t start affecting product development and competitiveness until 2028 at the earliest. That lines up with Tencent Tech’s own account — that the roadmap is already locked in for several years out.

There’s one precedent worth noting. In June 2018, when Intel CEO Brian Krzanich stepped down over a violation of company policy, the company installed CFO Bob Swan as interim CEO and said it would search for candidates both internally and externally. Seven months later, the person who became permanent CEO was Swan himself. The lesson: an interim arrangement can stretch on for months, and the person sitting in the interim seat sometimes just stays there.

The Signal Supply Chains Read First: Pre-Orders and Spec Upgrades

Ming-Chi Kuo devoted a separate section to the supply chain. In his analysis, Huang has pushed suppliers hard in two directions: locking in pre-orders that reserve future capacity, and pushing product specs upward. According to his comments as compiled by Anue (鉅亨網), a Taiwanese economic media outlet, Huang has a track record of converting whatever new technology his engineers come up with into a product the company commits real resources to ship. Suppliers took that commitment as a promise of future orders — and expanded capacity and raised specs ahead of time on the strength of it.

That’s why Kuo says the thing to watch is whether expansion plans and new-technology development shift across the supply chain. Changes, he argues, may show up in pre-orders and spec upgrades well before they show up in the product roadmap. He also wrote that these two factors are the core reason Nvidia’s supply-chain partners command such high price-to-earnings ratios⁴.

This is where the two analyses seem to diverge. Tencent Technology argued that the earnings of supply-chain companies like TSMC, SK Hynix, Micron, Foxconn, and Quanta track Nvidia’s shipment volumes more than they track any individual CEO. Kuo argued that those companies’ lofty valuations depend on Huang’s style of pre-ordering. But the two claims are measuring different things. Orders already placed show up in earnings as shipments go out; share prices, by contrast, price in the next round of expansion and the next round of specs in advance. The former rests on a roadmap that’s already locked in; the latter hinges much more on who’s willing to commit to orders, and how boldly.

A Taiwanese supplier executive quoted in a Reuters report that Anue cited made a similar point: if the new CEO slows down on pre-orders, suppliers may slow down their own expansion in response. Anue flagged advanced packaging schedules, liquid-cooling rack material procurement, and high-speed interconnect component timelines as the places where that signal would show up first.

Where the Market Priced It In on the 12th

On the 12th, Nvidia fell as much as 14.35% intraday on the New York exchange before closing down 4.34%. The closing price was $166.19, erasing roughly $182 billion in market cap for the day. Market cap briefly dipped below $4 trillion during the session but climbed back above that line just before the close. Data also showed that individual investors’ net buying in early trading was the highest on record.

Ming-Chi Kuo saw the resignation as a potential one-shot clearing of the overhang. With Huang having fully severed ties to the company, any future congressional summons or follow-up investigation is likely to weigh less heavily on Nvidia’s stock going forward, he argued. That said, he added it’s too early to judge how Huang’s subsequent moves might affect the company until the investigation’s findings are out.

On who benefits, the two analyses converge. Tencent Technology pointed to Broadcom and Google, noting that Nvidia’s management uncertainty coincides with a moment when Big Tech is ramping up its own custom chips (ASICs⁵). Kuo named Google’s TPU as the biggest likely beneficiary. While succession remains uncertain, TPUs could win more trust from customers and supply-chain partners, he said, and in the short term attention may shift to MediaTek — the Taiwanese chip design firm that’s a core TPU partner — and to Broadcom. According to a summary by Juxingwang (a Chinese tech news outlet), he explained that large customers looking to reduce reliance on a single supplier might shift some training and inference workloads to Google Cloud’s TPU, then decide on scaling up based on cost and performance.

Based on the closing prices on the 12th, AMD and Broadcom each rose roughly 6%. Tencent Technology flagged the emerging cloud providers that have leaned on Nvidia’s investment and backing — the so-called “neoclouds”⁶ — along with AI startups as the group most exposed, and indeed CoreWeave, an Nvidia-backed company, fell 6.86%. That said, this should be read as a single day’s reaction: CoreWeave had jumped more than 27% the day before on an earnings beat, and the same day also saw a sharp drop in another neocloud, Nebius, muddying the picture. MediaTek trades on the Taiwan exchange, so as of this writing I haven’t been able to confirm how it reacted.

The arrangement in which chipmakers help fund their own customers was covered in Issue 262 and Issue 341. The volumes written into purchase orders remain contractual commitments, but who gets invested in and whose loans get guaranteed is decided case by case by management. That’s exactly why Tencent Technology drew a distinction between large customers and emerging players.

The Variables Outside Both Analyses: The Investigation and China

The Tencent Tech piece attached a condition to its conclusion. It set aside the scenario where things escalate further — say, if an export-control violation is confirmed, bringing fines and compliance monitoring — and stated that no such evidence exists right now. The November 12 announcement makes no mention of how the independent investigation turned out. The board said on the 8th that it had no reason to suspect wrongdoing, but that it would review all available information anyway; the results were never disclosed. There are also reports that a shareholder class action has been filed.

Guo Mingchi treated China as its own separate item. If you count the chips and racks that get resold through third countries and ultimately end up with Chinese customers, he argues, China’s share of final demand in Nvidia’s AI chip revenue could be higher than the company’s officially disclosed figures. He also noted that this share is hard to verify from public data. This is exactly why some investors worry that tightened U.S. export controls could cause Nvidia’s revenue to drop sharply. Guo’s view is that even if regulations tighten further, final demand from China won’t disappear. As long as overseas resale channels remain open, Chinese buyers can still obtain chips or racks through roundabout routes — even at several times the normal price.

This point intersects with another variable. Before the resignation announcement, JPMorgan had assessed that the risk was rising of new guidance further restricting direct and indirect shipments to China on national security grounds. Geohyeongmang, a Chinese tech news outlet, also cited a Baird analyst’s estimate that if the U.S. blocks H200 exports to China again, it could trigger roughly $5 billion in asset write-downs. Guo Mingchi’s claim that demand persists rests on the assumption that resale channels stay open. And in this incident, where the chips ultimately end up is precisely what’s drawn U.S. attention.

The next checkpoint is November 17. Nvidia will report its Q3 FY2027 earnings as scheduled. Cress, who once fielded questions as CFO explaining the numbers, will now face them as CEO. How much the earnings call touches on reorganization plans, the CEO succession timeline, policy on supplier prepayments and investment guarantees, and China-related revenue will give us fresh material to test both analyses against.

Huang stepped down as CEO and director but remains a shareholder holding 3.58%. The announcement says nothing about what role, if any, he’ll play going forward.

Oswarld’s Lens

I agree with the broad thrust of both analyses. Chip design, production, and customer allocation all run on multi-year schedules—one person’s departure doesn’t halt next quarter’s shipments.

What catches my eye first are the decisions that never make it onto a schedule. The pre-orders Ming-Chi Kuo flagged are exactly that kind of decision, and so is a chip supplier backing a customer’s financing with investment or guarantees. When I get asked to review an AI data center deal these days, I look past the site or the GPU count first and ask whether there’s a confirmed tenant, and whether lenders are moving on that tenant’s credit. When a supplier decides on capacity expansion, it’s ultimately asking the same question: who’s going to buy this volume, and can that promise be trusted. Roadmaps get written down, but that promise has always been made person to person, deal by deal. I think which way that judgment tilts during a stretch when the CFO is also serving as CEO matters more for the companies orbiting Nvidia than for Nvidia itself.

A structure where 36 people reported to one person had the advantage of fast decisions, but the structure itself was built around that one person. Whoever succeeds him will struggle to run it the same way. Add a COO and you add a layer, and decisions could slow down—I read that as the actual substance behind the “decision speed” that Tencent Tech was talking about.

Closing

Among the predictions made in analyses written before the resignation, the interim CEO call turned out right. The analyses that came after the resignation widened the scope of what to check, from the org chart all the way to suppliers’ purchase orders. But the investigation’s findings and the China-route variable remain outside both pieces.

---

💬 Reader, have you ever worked at an organization where decisions all funneled through a single CEO? Tell me in the comments what was the first thing that ground to a halt when that person stepped away.

📨 If you know someone who wants to look at Nvidia news from an angle other than the stock price, pass this piece along to them.

References & Further Reading

• NVIDIA, “NVIDIA Announces CEO Transition”, 2026.11.12. : The primary source for the resignation, the interim CEO appointment, the naming of the first board chair, and the earnings schedule.
• NVIDIA, “Statement from NVIDIA Board of Directors”, 2026.11.8. : Contains the launch of the independent investigation and the board’s reasoning at the time.
• NVIDIA, “Form 8-K”, 2026.11.12. : Confirmed the date the board accepted the resignation (November 11) and the year Jones’s board tenure began.
• NVIDIA, “2026 Proxy Statement”, 2026.5.12. : Source for board composition, voting rights, and shareholding structure.
• NVIDIA, “Form 10-K, Fiscal 2026”, 2026.2.25. : Confirmed headcount and R&D staff size.
• Tencent Tech, “失去黄仁勋，英伟达会怎样？”, republished via Futu News, 2026.11.12. : The pre-resignation analysis I contrasted in this issue.
• Ming-Chi Kuo, “重點分析 Nvidia CEO 黃仁勳辭職”, X, 2026.11.12. : The post-resignation analysis I contrasted in this issue. The follow-up thread includes the China-demand item.
• CnYes, “黃仁勳請辭輝達利空出盡？郭明錤點名Google TPU最大受惠”, 2026.11.13. : Confirmed Kuo’s explanation of the pre-orders, remarks from Taiwanese supplier executives, and Baird’s estimates.
• CNN, “Nvidia CEO Jensen Huang resigns after mounting scandal”, 2026.11.12. : Source for the intraday and closing declines, the drop in market capitalization, net retail buying, and CoreWeave’s closing move.
• CNBC, “Nvidia stock closes down over 4% after CEO Jensen Huang announces resignation”, 2026.11.12. : Confirmed the recovery to a $4 trillion closing market cap and the content of The New York Times’s November 11 report.
• SBS, “Nvidia’s Market Cap Shrinks Further, Down 30% from Last Month’s Peak”, 2026.11.13. : Confirmed the closing price of $166.19.
• Reuters, “Global AI stocks sink as Nvidia CEO’s exit deepens sector concerns”, republished via Kitco, 2026.11.12. : Confirmed CoreWeave’s gain the previous day.
• NBC News, “Nvidia stock falls after Huang resigns, weighing on broader market”, 2026.11.12. : Confirmed JPMorgan’s assessment of China shipment-restriction risk and Nebius’s sharp decline.
• Fortune, “2 scandals, 1 destination: why Nvidia can’t shake China smuggling allegations”, 2026.11.10. : Referenced for U.S. concerns over the chips’ ultimate destination.
• Fortune, “Report: Nvidia preparing to oust CEO Jensen Huang”, 2026.11.11. : Referenced for the structural gap left by the absence of a COO and board chair, and the timeline of The Wall Street Journal’s reporting.
• Al Jazeera, “Nvidia CEO Jensen Huang steps down amid spy scandal and market loss”, 2026.11.12. : Source for the reporting on the shareholder class-action lawsuit.
• Intel, “Intel CEO Brian Krzanich Resigns, Board Appoints Bob Swan as Interim CEO”, 2018.6.21. : The precedent of a CFO stepping in as interim CEO.
• Kwangseob Ahn, “AI Data Centers’ Essence… A Complex Four-Variable Function”, ZDNet Korea, 2026.9.19. : Source for the data-center evaluation framework I used in “Oswarld’s Lens.”

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.

📝 Glossary

Footnotes

1. Dual-class voting rights: A structure that grants some shares more voting power than others, letting founders control shareholder votes even with a small ownership stake. ↩

2. CUDA: A software toolkit for writing and running computing programs on NVIDIA GPUs. Much AI software is built around it, so porting to other chips requires rework. ↩

3. TPU: A chip Google designed in-house for AI computation. Outside customers can also rent it through Google Cloud. ↩

4. Price-to-earnings ratio (P/E): Share price divided by earnings per share. A higher ratio means the market is pricing in stronger future earnings growth. ↩

5. ASIC: An application-specific integrated circuit designed for a particular purpose. The AI chips Big Tech companies build to fit their own services fall into this category. ↩

6. Neocloud: A newer cloud provider that specializes in renting out GPU capacity for AI computation. CoreWeave is a prominent example. ↩