跪拜 Guibai
← Back to the summary

Why NVIDIA's Board Closed the Door on Jensen Huang's Return: The Governance Machinery in Plain Sight

Core conclusion: Jensen Huang's return is not refused because his ability is called into question, but because the matter that triggered his resignation has not been fully investigated. On November 8, 2026, the board launched an independent investigation; on November 10, independent directors retained a former federal prosecutor and Wachtell Lipton; on November 12, Huang resigned as President, CEO, and director; on November 17, he said on social media "I am willing to return if invited," and a few hours later, board chair Harvey Jones opened the earnings call with a rejection: the board unanimously accepted the resignation and has no intention of restoring his position. This is a standard corporate governance action, not a personnel emotion. To understand it, you only need to understand three things: the board's fiduciary duty, key-person risk, and the succession process in a crisis.

Table of Contents

1. One-sentence conclusion: The refusal is about "process", not "ability" (Process, Not Ability)

Let's first clarify the most easily misread point 🎯: The board's statement contains no denial of Huang's ability. The official press release says "Jensen has had a profound impact on NVIDIA," and on the earnings call Jones only said "does not intend to restore his position," while emphasizing that the board "has full confidence" in the interim CEO.

What is actually being refused is the sequence: the investigation is not finished → responsibility is not clarified → the succession process has already started. If any of these three steps is incomplete, letting the founder return to the CEO seat would be equivalent to the board admitting "the investigation can be bypassed."

There is an easily overlooked institutional detail: Huang resigned not only as CEO, but also as President and Director. Resigning as director is the most critical point—he is no longer a board member, thus losing his seat to defend himself within the board and influence the investigation's pace. The scope of resignation determines the subsequent negotiation space. This is a clean cut, not a "leave of absence".

A metaphor 🎭: this is like a hospital not letting a surgeon who has just been complained about immediately return to the operating table. It's not denying his skill, but before the accident investigation is over, the operating table cannot be decided by the person involved. What the patient (shareholders) want is not "he is skilled," but "the process is trustworthy."

2. Compress the news into events: six nodes from November 8 to 17 (Six Nodes)

Arrange the scattered reports chronologically, and the governance logic emerges:

Time (US) Event Governance implication
Nov 8 Board announces launch of independent investigation; Huang sends an all-hands letter acknowledging personal conduct issues and says he will cooperate Board acts first, the person responds later; control rests with the board
Nov 10 Independent directors retain Joon H. Kim, former acting U.S. Attorney for the Southern District of New York, and Wachtell, Lipton, Rosen & Katz to conduct a "thorough and independent" investigation Investigative body shifts from "the company" to "independent directors + external lawyers"
Nov 12 Huang resigns as President, CEO, and director, effective immediately; CFO Colette Kress becomes interim CEO; Harvey Jones becomes the first board chair in company history Completes "person leaves, power transfers, position is filled" in one step
Nov 12 Stock opens down nearly 14% Market first applies an "uncertainty discount"
Nov 17 morning Huang says on personal social media that he is willing to return as CEO if invited Founder proactively seeks to return, throwing the ball back to the board
Nov 17 after market Q3 earnings released; on the call, Jones explicitly says "does not intend to restore his position," and says Heidrick & Struggles has been engaged to choose a permanent CEO Respond to the market with both performance and governance posture

Of the six steps, the most instructive is step 3's bundled action: resignation, interim replacement, and board chair succession happen on the same day. In crisis governance, the most expensive thing is not the mistake itself, but the "authority vacuum period"—as long as someone can ask "who is in charge now," orders, supply chain payment terms, and cloud vendor long-term contracts begin to loosen.

3. Why can the board refuse a founder: fiduciary duty and where decision-making authority lies (Fiduciary Duty)

Many people intuitively think, "He is the founder, the company is his." In listed companies, this intuition is wrong.

NVIDIA is a NASDAQ-listed company. The board is elected by shareholders and owes fiduciary duty to all shareholders, with two core elements:

Once these two duties are written down, the board's reason for refusing reinstatement becomes "boring" but very hard 🔍: restoring his position before the investigation is complete could constitute a violation of the duty of care. Conversely, if the investigation proves the allegations are unfounded, the board is still not obligated to invite him back—choosing who serves as CEO falls under business judgment, which courts generally respect.

There is also a structural change that is easy to miss: Harvey Jones becomes the first board chair in company history. This "first" itself is a signal—NVIDIA previously had no separate board chair position. In a crisis, separating the chair from the CEO is the classic checks and balances design in corporate governance. Jones is the former CEO and executive chairman of Synopsys and a managing partner at Square Wave Ventures. He is the "board's top leader," not the "business's top leader."

In other words 🎯: the referee's seat in this game had already changed hands on November 12.

4. Why the founder's halo fails in a crisis: pricing key-person risk (Key-Person Risk)

A founder is an asset to a company, but also concentration risk. In finance, this risk is called key-person risk: when a company's technology roadmap, client relationships, and supply chain bargaining power are highly tied to one person, any change in that person is converted into a valuation discount.

The market has already priced it: on November 12, the stock opened down nearly 14%. This decline is not a pricing of "Huang's ability," but a pricing of "who makes the final call in the next 12 months."

But there is a counterintuitive point here ⚠️: the stronger the founder, the more the board needs to prove in a crisis that "the company is not just one person." Because what clients and suppliers fear most is not personnel change, but "after a change, the same person is still behind the scenes pulling strings"—that would mean the governance structure is a facade, and compliance, export controls, and data security matters that require "the company's name as a commitment" would all be called into question.

NVIDIA's business structure makes this even more sensitive. Its clients are cloud providers, sovereign AI projects, and server manufacturing chains; the business is embedded with export controls and national security reviews. When the investigation involves "whether there are risks of sensitive information or national security," the board must present a structure that regulators can understand—and "the founder returning to continue as CEO" is precisely the hardest structure to understand.

A metaphor 🎭: the founder is like a car's engine, and the board is like the inspection station. However strong the engine, if the brakes fail and the inspection is not passed, the car cannot go on the road. The inspection station does not issue a certificate just because "this is the only original engine."

5. Independent director committee + external lawyers: why the investigation must be "isolated" (Independence)

The November 10 statement is only one paragraph, but it is rich in information: a committee of independent directors oversees, and external lawyers lead the investigation.

This arrangement solves a very specific problem: conflict of interest.

Wachtell, Lipton, Rosen & Katz is one of the most frequently cited law firms in U.S. M&A and corporate governance; Joon H. Kim is a former acting federal prosecutor for the Southern District of New York—this combination sends a clear signal: the investigation is to be conducted to a standard that "may involve law enforcement," not an "internal HR" standard. This also explains why the statement deliberately does not disclose interviewees, timeline, or how conclusions will be formed: once disclosed, the investigation becomes a negotiation.

💡 Takeaway for ordinary readers: to judge whether a company is professional in handling a crisis, don't look at how many statements it issues; look at whether it has handed investigative authority to people not controlled by the person involved.

6. Interim CEO system: why CFO Colette Kress (Interim CEO)

The interim CEO is Colette Kress, who has served as NVIDIA's Executive Vice President and CFO since September 2013, with nearly 25 years of financial experience at Cisco, Microsoft, and other companies before that. After her appointment, she retains the CFO role simultaneously until a permanent successor is in place.

Why does the CFO often step up in a crisis? Three reasons, in order of importance:

  1. She holds all the numbers. In a crisis, the most needed thing is to give clients, suppliers, and investors "consistent answers," and the CFO is the only person who simultaneously knows revenue structure, gross margin, inventory, and capital expenditure.
  2. She is a "process" rather than "charisma" role. The interim goal is not to open new fronts but to keep the roadmap from slipping—this is precisely the CFO's comfort zone.
  3. She will not be seen as a fait accompli of "long-term competitor". The interim CEO + CFO role is naturally transitional and does not send a signal to external candidates that "the seat is already predetermined."

Jones's exact words in the press release: "The board has confidence in Colette and the leadership team, and expects the company's operations, customer commitments, and product roadmap to continue without interruption." Note that the subject of this sentence is the leadership team, not a single person—this itself is governance language: separating company capability from individuals and re-attaching it to the organization.

7. Choosing a permanent CEO: headhunters, internal and external candidates, and the national security variable (CEO Search)

The board has engaged Heidrick & Struggles to assist in identifying a permanent CEO, considering both internal and external candidates. The headhunter's most famous tech engagement was helping Microsoft select Satya Nadella in 2013—"selecting from a person long cultivated internally"—this precedent has direct reference value for NVIDIA's current situation.

Jones's remarks on the earnings call are worth reading sentence by sentence, because they give three constraints at once:

There is also an external variable: U.S. Treasury Secretary Bessent publicly expressed a hope that NVIDIA would look for a CEO externally. This statement has no decision-making power, but it will materially affect external candidates' psychological expectations and the smoothness of regulatory communication—when regulators clearly prefer an "external candidate," the board's choice space is compressed.

So the true answer to "can Jensen Huang come back" is not "does the board like him," but: in a window where the company must prove to regulators that "governance has been reset," letting the person involved return has a cost higher than the benefit.

8. What the earnings report proves and doesn't prove: a lesson in reading 115.8 billion vs 130 billion (Reading the Numbers)

Data from the third quarter of fiscal year 2027 (ending October 25, 2026), released on November 17, is the "first answer sheet" of the "post-Huang era":

Metric Q3 FY2027 YoY Notes
Revenue $115.844 billion +103% QoQ +20%, above market expectations of ~$109.1 billion
GAAP net income $65.287 billion +105% —
Diluted EPS (GAAP) $2.70 +108% Non-GAAP $2.68
Gross margin 74.8% +1.4 percentage points Prior-year 73.4%
Next-quarter revenue guidance $130 billion +91% Corresponds to Nov 2026–Jan 2027

This table can prove: short-term operations have not broken because of the change, and this is the 17th consecutive quarter of beating Wall Street revenue expectations. What it cannot prove is: the long-term roadmap does not need the founder.

The difference between these two is precisely the core of governance analysis 🔍. Financial data is a snapshot of "the past quarter," while CEO transition affects "the pace of capital expenditure over the next three years, client long-term contracts, and the right to define the next-generation platform." So you see an apparently contradictory phenomenon: the earnings report beats expectations, yet the stock price is still digesting uncertainty—the market prices both "current-period performance" and "governance discount" at the same time, but they are not on the same time dimension.

⚠️ A common misreading: treating "$130 billion" as realized revenue. It is guidance, corresponding to the next quarter, and the +91% YoY is calculated against the same quarter a year earlier. When reading financial reports, first distinguish "already happened" from "committed," then talk about growth rates.

9. A comparison table: symptom → cause → fix (Symptom → Cause → Fix)

Symptom Cause Fix
Founder resigns, publicly says wants to return, board directly refuses Investigation not complete; restoring position would damage the independence and credibility of the investigation's conclusions Independent director committee + external lawyers produce conclusions first, then discuss personnel
Stock drops nearly 14% on announcement day Market prices uncertainty about "who makes the call," not a denial of ability Complete "resignation + interim CEO + board chair succession" in one day to compress the authority vacuum
Clients and suppliers worry about long-term contracts and roadmap interruption Key-person risk concentrated on the founder; organizational capability not made visible CFO serves as interim CEO; publicly emphasize "leadership team" rather than individual
Regulators cautious about "founder's return" Export controls and national security reviews need clear, verifiable governance structure Write geopolitical response capability into job requirements; consider external candidates
Earnings beat expectations but valuation still has discount Current-period performance and long-term governance risk are not on the same time dimension Use consecutive quarters of delivery + permanent CEO in place to gradually dissolve the discount

10. Complete runnable code: recalculate this earnings report and the break-even line (Runnable Code)

The following code is zero-dependency, copy-and-run (Python 3.8+). It recalculates all the numbers cited in this article: reverse-engineers prior-year revenue, verifies growth rates, calculates gross margin increase, and computes how much the stock needs to rise after a 14% drop to recover.

# Recalculate three things with public numbers: whether the YoY growth rate is self-consistent, gross profit increase, drawdown break-even line
# Data source: NVIDIA FY2027 Q3 earnings and company announcements (2026-11-17)

# 1) This quarter and guidance (in billions of USD)
q3_revenue = 1158.44          # Q3 FY2027 actual revenue
q3_yoy = 1.03                 # Official YoY +103%
q4_guide = 1300.00            # Next-quarter revenue guidance
q4_yoy = 0.91                 # Guidance YoY +91%

# 2) Reverse prior-year revenue from growth rate, verify consistency
q3_last_year = q3_revenue / (1 + q3_yoy)      # Reverse Q3 FY2026 revenue
q4_last_year = q4_guide / (1 + q4_yoy)        # Reverse Q4 FY2026 revenue

# 3) Gross margin increase - additional gross profit (billions USD)
gm_now, gm_prev = 0.748, 0.734                # 74.8% and 73.4%
gross_profit_gain = q3_revenue * (gm_now - gm_prev)  # Extra gross profit at same revenue

# 4) After a 14% drop, break-even line: how much of a percentage gain to return to the original point
drawdown = 0.14
rebound_needed = 1 / (1 - drawdown) - 1

# 5) Output, two decimals
print(f"Q3 FY2026 reverse revenue: {q3_last_year:.2f} billion USD")   # Should be ~570.66
print(f"Q4 FY2026 reverse revenue: {q4_last_year:.2f} billion USD")   # Should be ~680.63
print(f"Gross margin increase - additional gross profit: {gross_profit_gain:.2f} billion USD")
print(f"After {drawdown:.0%} drop, break-even requires rise: {rebound_needed:.2%}")
print(f"Next-quarter guidance QoQ: {(q4_guide / q3_revenue - 1):.2%}")

Actual output:

Q3 FY2026 reverse revenue: 570.66 billion USD
Q4 FY2026 reverse revenue: 680.63 billion USD
Gross margin increase - additional gross profit: 16.22 billion USD
After 14% drop, break-even requires rise: 16.28%
Next-quarter guidance QoQ: 12.22%

Three directly applicable conclusions 🎯:

11. Three types of companies, three succession playbooks (Three Succession Playbooks)

Abstracting from NVIDIA's situation, it can be applied to three typical scenarios. As I mentioned in What other countries are doing beyond the China-US AI race, competition in the AI industry has long been not just about models and compute; governance structure is also a competitive capability—this event is the most direct example.

First, the founder is still in place and the company is growing fast. The focus is to convert "personal capability" into "organizational capability" in advance: clarify the number-two position, transfer key client relationships from personal to company-to-company contracts, and put truly independent directors on the board. This cannot prevent a crisis, but it can shorten the vacuum period when a crisis occurs.

Second, the founder leaves for some reason and the company is in crisis. The core actions are only three, and they must be completed on the same day: the person fully exits (including the board seat), an interim leader is in place, and the board power structure is reset. Any delay turns a "personnel problem" into a "governance problem."

Third, a professional manager takes over while the founder remains a spiritual symbol. Here the biggest threat is "remote governance"—the founder influences decisions without actually being in office. The solution is to write boundaries into the system: which matters must be decided by the board, which information must be synchronized to all directors, and in what capacity the founder participates (advisor / shareholder / director).

One sentence to close the thread: Corporate governance is not insurance for good people; it is a procedure prepared for "good people turning bad" or "good people being absent." In this instance, NVIDIA did not side with personal sentiment; it sided with procedure—which is precisely the choice a company of its market capitalization must make.

References:

Final judgment: If you are watching news about a company's "founder departure," don't rush to judge whether it will collapse—first check whether the board completed "person leaves, power transfers, position is filled" on the same day. If all three are done, the short-term fundamentals are likely to hold; if any one is missing, don't trust any statement saying "all is normal."

This column, "LLM Large Model Technology Column," has gone from the Transformer attention mechanism to LangChain, RAG, and AI industry governance. Each article includes a piece of runnable code and updates weekly. To read in order, subscribe on the column page.