Nvidia’s board accepted Jensen Huang’s resignation as president, chief executive and director on 11 November 2026, effective immediately, according to the company’s Form 8-K. Colette Kress, the chief financial officer, is interim chief executive while the board searches inside and outside the company. Harvey C. Jones, a director since 1993, became Nvidia’s first board chair. The filing gives no reason.
For anyone who builds on Nvidia, the short version is that nothing you run changed this week. CUDA, the drivers and the Blackwell systems already in data centres are what they were on Friday. What changed is the risk around them. A company whose founder held the roadmap, the supplier relationships and the public story for 33 years now has an interim chief executive, a new chair, an independent investigation and possible hearings in Congress, and its shares had their worst week since April 2025.
I run Crimson Umbrella, and the AI systems we build mostly call models that run on Nvidia chips in somebody else’s cloud. What follows separates what the record shows from what has only been reported, then sets out what I’d do about it as a buyer of compute and as someone who runs an engineering team.
Nvidia CEO resignation: what the filing says, and what has only been reported
The primary record is short. The 8-K and the press release attached to it say the board accepted the resignation, appointed Kress as interim chief executive while she stays chief financial officer, and made Jones chair. Jones said “The Board has full confidence in Colette and the leadership team” and that operations, customer commitments and the roadmap should continue without interruption. Nothing in either document explains why Huang left, and neither carries a statement from him.
Three days earlier, on Sunday 8 November, the board had announced an independent investigation “in light of recent reporting”, run by independent directors with outside counsel. In that statement the board said it had “no reason to believe that either national security or NVIDIA secrets were compromised”, and that Huang supported the review.
The rest comes from the press, and it should be read that way. NBC News reported an internal memo in which Huang wrote “I was unfaithful to my wife”. NBC also reported that the Financial Times and The New York Times had first described a US government memo alleging a relationship with an executive at Megaspeed, a Singapore buyer of Nvidia chips, and alleging links between that executive and Chinese intelligence. CBS News reported that Megaspeed is under investigation in the US and Singapore. China’s government called the allegations “baseless and purely slanderous smears”, ABC News reported. The New York Times and Reuters reported relationships with Nvidia employees, which other outlets say they have not verified. Al Jazeera noted there is no established evidence that Huang knowingly assisted Beijing. A company spokesperson told NBC he will not work with Nvidia in any capacity going forward.
| Date | What happened |
|---|---|
| Sunday 8 November | The board announces an independent investigation with outside counsel |
| Monday 9 November | NVDA falls more than 4% to $178.67 on a day the Nasdaq rises 2.7%, per ABC News |
| 10 November | The board hires a former acting US attorney and his firm to lead the review, NBC reports |
| 11 November | The board meets and accepts the resignation, the event date in the 8-K |
| 12 November | The 8-K and press release go out; NVDA closes down 4.34% after falling as much as 14% during the day |
| 13 November | NVDA falls another 2.01% and closes below $4 trillion in market value for the first time since July 2025 |
| 17 November | Third-quarter results for fiscal 2027, due at about 1:20pm Pacific |
Washington has its own timetable. CNBC reported that Brian Mast, who chairs the House Foreign Affairs Committee, said Huang had agreed to appear, likely before Thanksgiving, and that Senator Tim Scott asked Huang and the commerce secretary to testify in December. None of that changes a product, but every hearing date is another headline in the middle of a chief executive search.
What happened to Nvidia’s stock
The shares fell about 13% in the week to 13 November, CNBC reported: the biggest weekly drop since April 2025 and a sixth straight weekly loss, the longest run since 2012. The stock was already sliding. By CNBC’s count it was down about 30% from early October, so the scandal arrived on top of a market that had begun to question how long AI spending would grow this fast.
Analysts moved quickly, and not all in one direction; none of what follows is investment advice. Bernstein cut its price target to $200 from $400, CNBC reported. Truist downgraded the stock to hold and cut its target to $174 from $346, and Craig-Hallum downgraded to hold at $180 from $275, Yahoo Finance reported. D.A. Davidson cut its target to $200 from $300 and kept a buy rating. Wedbush’s Matthew Bryson wrote that “nothing has changed” and expects results above forecasts, because supply, not demand, limits what Nvidia ships.
Other companies’ bosses have left over personal conduct before, and their shares barely moved on the day:
| Company and year | What happened | First-day move |
|---|---|---|
| Intel, 2018 | Brian Krzanich resigned over a past relationship | About -2% |
| McDonald’s, 2019 | Steve Easterbrook was dismissed over a relationship | More than -2% |
| BP, 2023 | Bernard Looney resigned over undisclosed relationships | Up to -1.7% |
| Nestlé, 2025 | Laurent Freixe was dismissed over a relationship | About -0.8% |
The figures come from an MLQ compilation that mixes opening and closing moves, so treat them as rough. Nvidia’s fall was larger, and its stock was already sliding on doubts about AI spending. But none of those companies had a founder who had run it for its whole life, no board chair and no named successor, and none had a foreign intelligence question attached.
How much of Nvidia was one person
Huang founded Nvidia in April 1993 with Chris Malachowsky and Curtis Priem. It shipped the GeForce 256 in 1999, marketed as the first GPU, and released CUDA in 2007, which turned a graphics card into a general-purpose parallel computer and later into the default hardware for training neural networks. Revenue went from $4.13 billion in fiscal 2014, when Kress became chief financial officer, to $215.9 billion in fiscal 2026, by Al Jazeera’s count.
The latest quarter shows the scale. Nvidia reported second-quarter fiscal 2027 revenue of $96.2 billion, up 106% on the year, with $89.0 billion from the data centre and a 75.0% gross margin. It guided the third quarter to $108.0 billion, plus or minus 2%, assuming no data-centre compute revenue from China.
Governance is where the founder’s shadow shows. The company had never had a board chair until this week, and it had no named successor; whether it kept a confidential succession plan is unclear. People who know the industry told CNBC that part of Huang’s job was persuading suppliers to commit capacity years ahead. A chief financial officer can run a company well; that is a different job from talking TSMC into a packaging line. Kress has been chief financial officer since 2013 and knows every number in the business, so the near-term risk is less about running Nvidia and more about who decides its next bet.
Rubin, CUDA and the supply chain: what could actually change
Nvidia said in August that Vera Rubin, the platform after Blackwell, was in full production. The board says the roadmap continues without interruption. Analysts disagree about the years after that. SemiAnalysis founder Dylan Patel warned, as paraphrased by TokenPost, that Rubin manufacturing problems could take longer to solve without Huang, and that later parts such as Rubin Ultra and Feynman could slip. His colleague Jordan Nanos countered that near-term delivery plans and supply chains were already arranged. Both views are judgements. No delay has been announced.
The constraint was never the chip design alone. Advanced GPUs depend on TSMC’s CoWoS packaging and on stacks of high-bandwidth memory, both scarce, and those allocations are negotiated years ahead. China is the other variable. US policy allowed sales of H200 chips to approved Chinese customers from December 2025 with a 25% tariff, and Chinese firms reportedly ordered more than two million of them, Al Jazeera reported. A board facing congressional questions about export compliance will be more careful about that business, which is a revenue question first and a supply question for everyone else.
For a team writing CUDA kernels or serving models, none of this touches code. Drivers, libraries and the APIs you call are stable products with thousands of engineers behind them. What could move are prices, allocation of scarce parts, and how fast new generations arrive, which matter if your 2027 plan assumes Rubin-class capacity at a certain cost.
What I’d do this quarter if you buy GPU compute
Don’t migrate anything because of a headline. Moving a production AI system off CUDA costs months, and the chance that Nvidia’s software stops working is close to nil. The useful work is smaller and cheaper.
Make sure you could move if you had to. Keep model code on standard frameworks, put inference behind an interface your application owns, and run your evaluation set on a second provider or accelerator once a quarter, so the cost of switching is a number you know rather than a guess. We wrote about why the evaluation set comes first and why the KV cache, not the GPU, is usually the bill.
Read your capacity contracts. If you reserve GPU time through a cloud provider, check what the contract says about price changes and substitution when a generation slips. Then watch the results on 17 November: guidance for the fourth quarter, anything said about Rubin timing and supplier commitments, and whether the interim chief executive answers roadmap questions or leaves them to engineering leads. That call will tell you more than this week’s coverage did.
If you need help with any of this, it is the work our AI and automation and data engineering teams do: inference that can run on more than one provider, evaluation harnesses, and cost work on token-heavy systems.
The key-person lesson is for your team too
Nvidia’s problem is a large version of one most engineering teams have. One person holds the roadmap in their head, knows why the system is built the way it is, and is the only one a key supplier will call back. Nobody plans for that person leaving, because the person who would write the plan is the one it is about. Then they leave, for a good reason or a bad one, and the company finds out how much was written down.
You can test for it in an afternoon. Ask a new engineer to set up your main system from the documentation alone. Ask who else could approve a production change this week, and who else could explain the architecture to a customer. On our own projects we hold ourselves to a simple bar: a new engineer clones the repository and has it running locally inside an hour, from the README alone. Code, infrastructure and documentation sit in the client’s accounts, a named senior lead owns each engagement, and written handovers record each decision so it doesn’t live only in one head.
If your team depends on one person more than you’d like, an architecture review is a fast way to find where, and a part-time technical lead is a lighter way than a full-time hire to add a second senior person who knows the system. Both start with a twenty-minute call.
Questions people are asking
Did Jensen Huang resign as Nvidia CEO?
Yes. Nvidia’s board accepted Jensen Huang’s resignation as president, chief executive and director on 11 November 2026, effective immediately, according to Nvidia’s Form 8-K filed on 12 November. He had led the company since founding it in 1993, and a company spokesperson said he will not work with Nvidia in any capacity going forward.
Why did Jensen Huang resign?
Nvidia has not said. Its 8-K and press release give no reason. On 8 November the board announced an independent investigation in light of press reporting, and NBC News reported an internal memo in which Huang admitted an affair; other reports remain unverified. Nvidia’s board said it had no reason to believe national security or company secrets were compromised.
Who is Nvidia’s CEO now?
Colette Kress, Nvidia’s chief financial officer since 2013, is interim chief executive and keeps the finance role while the board searches internally and externally for a permanent chief executive. Harvey C. Jones, a director since 1993, became Nvidia’s first board chair on 11 November 2026.
How much did Nvidia stock fall after Huang resigned?
NVDA closed down 4.34% on 12 November 2026, after falling as much as 14% during the day, and fell another 2.01% on 13 November to close below $4 trillion in market value for the first time since July 2025. CNBC put the weekly fall at about 13%, the largest since April 2025, and the decline since early October at about 30%.
Does Huang’s departure affect CUDA or the Rubin roadmap?
Not directly. CUDA, Nvidia’s drivers and its shipping Blackwell systems are unchanged, and Nvidia said in August that Vera Rubin was in full production. The board says the roadmap continues without interruption. Some analysts worry later generations could slip without Huang’s influence over suppliers, but no delay has been announced.
When does Nvidia report earnings?
Nvidia reports third-quarter fiscal 2027 results on Tuesday 17 November 2026, with the release at about 1:20pm Pacific and a call at 2pm Pacific. It guided revenue to $108.0 billion, plus or minus 2%, in August. The call is the first under the interim chief executive.
What is key-person risk?
Key-person risk is the danger that a business depends on one individual’s knowledge, relationships or decisions so heavily that their departure disrupts it. Nvidia, with a founder of 33 years, no board chair until November 2026 and no named successor, is a large example. The same risk exists in small engineering teams when one person holds the architecture in their head.
Saurav Raj Joshi is the founder and CEO of Crimson Umbrella Technologies and a principal data and BI engineer. You can find Saurav on LinkedIn.