Jensen Huang resigned as Nvidia's CEO on 12 November 2026, days after admitting an affair with a woman US officials reportedly suspect of working for Chinese intelligence and a day after a second relationship, with a direct report, was reported. This extension post combines the ChinaTalk emergency podcast with the latest reporting to ask what comes next: for Nvidia's control structure, for Huang's chances of a comeback, for Washington, for Musk and SpaceX, and for the balance between the US and China in AI.
- How to read this extension
- 1. Executive summary
- 2. Nine days that ended a 33-year run
- 3. What the ChinaTalk panel said, chapter by chapter
- 4. It is economic warfare, and smuggling is the front line
- 5. What is ahead for Nvidia: the control problem
- 6. Can he come back? Jobs, Altman and the limits of the analogy
- 7. Washington: Trump, Lutnick, Bessent and the Democrats
- 8. Could Jensen end up at SpaceX, and would Musk be fine with it?
- 9. The circular dependency: Nvidia as the “central bank of AI”
- 10. Who gains and who loses: AMD, Google, Amazon, Chinese models, OpenAI, Anthropic and the hyperscalers
- 11. China versus the US: the chip gap closes, the model gap is contested
- 12. What to watch next
- 13. Takeaways
How to read this extension
This post extends the Nvidia chapter of Gen AI – The New Reality (book pp. 22–29, with related material on pp. 42, 73, 97–98 and 117). When the book was written in early 2024, Nvidia was “Godfather of AI”, a US$1.12 trillion company with a CUDA moat, a Taiwan link and a long list of risks. Two years on, the biggest risk on that list was one the book never put on a slide: the founder himself. On 12 November 2026 Jensen Huang resigned as president, CEO and director of the company he co-founded in 1993 [1].
The post does three things. First, it summarises the ChinaTalk emergency podcast recorded in the shock of the resignation and pulls out its best arguments, with short quotes [2]. Second, it tests those arguments against the newest reporting, which runs to 13 November 2026. Third, it answers the questions a business reader will actually ask: can Jensen come back, who controls Nvidia now, what did Trump, Lutnick and Bessent say, could he end up at SpaceX, and who gains and who loses across AMD, Google, Amazon, OpenAI, Anthropic, the hyperscalers and Chinese open models.
Two housekeeping notes. First, the podcast transcript does not label speakers, so quotes are attributed to “the panel”, or to Dylan or Jordan where the conversational cues make that clear. Second, much of what is reported about Huang is allegation. He has not been charged with any offence, Nvidia has said it had no reason to believe national security or company secrets were compromised, and the Chinese government rejects the espionage claims [3,4]. Where this post says “reported” or “alleged”, it means exactly that.
1. Executive summary
Jensen Huang is out, the AI trade has lost its figurehead, and Washington has just been handed the political momentum it needed to tighten chip controls. Those three facts drive everything else in this post.
- What happened. Huang resigned on 12 November after he admitted an affair with Alice Huang, a former Megaspeed executive whom US intelligence reportedly suspects of working for China’s Ministry of State Security, and after the New York Times reported a relationship with Kimberly Powell, a senior Nvidia executive who reported to him [5,6]. CFO Colette Kress is interim CEO while the board searches; Harvey C. Jones becomes Nvidia’s first-ever board chair [1].
- Why it matters more than a scandal. Nvidia is the “central bank of AI”: it supplies the chips and also finances the customers who buy them, through equity stakes, lease guarantees and off-balance-sheet backstops [7,8]. A founder who personally ran roughly 60 direct reports and the key relationships has left with no named successor [3].
- The market reaction. Shares opened about 13% lower on 12 November, touched $148.80, and closed at $166.19, down 4.3%. The stock is roughly 30% below its early-October record, when the company was worth about $5.8 trillion [9,10,11].
- The podcast’s core argument. Export controls should target chipmaking tools (the MATCH Act) more than remote cloud access (the RASA bill), because an outright ban on rental invites retaliation and more smuggling. Nvidia will be fine in the short term but faces a harder medium-term test if executives leave and decisions start going to committees [2].
- Can he return? Possible in theory, unlikely in practice. Steve Jobs and Sam Altman both came back, but neither left under a national-security cloud. Huang resigned from the board too, faces a law-firm investigation, congressional interviews and reported Justice Department probes into companies around him [4,12].
- Washington. Trump praised Huang and suggested Tim Cook as successor; Treasury Secretary Scott Bessent publicly blamed Commerce, which “has the pen”, for not blacklisting Megaspeed; Commerce Secretary Howard Lutnick will testify to the Senate Banking Committee on 10 December while advisers discuss whether he should go [13,14,15]. Democrats are pressing on oversight and on Trump’s handling of the case rather than on antitrust [16,17].
- Winners and losers. AMD, Broadcom, Google and Amazon gain narrative momentum for their own silicon, and Chinese open-weight models keep eating share. Neoclouds, Supermicro and anyone relying on Nvidia’s balance sheet are most exposed. OpenAI and Anthropic are mixed: more dependent on Nvidia hardware, but freed from Huang’s open-source and China-friendly lobbying [18,19,20].
| Indicator | Latest value | Source / note |
|---|---|---|
| Nvidia market cap at record (6 Oct 2026) | ~US$5.78 trillion | Motley Fool [10] |
| Nvidia close on exit day (12 Nov) | $166.19 (−4.34%) | CNBC [9] |
| Fall from record close | ~30% | RoboForex [11] |
| Q2 FY27 revenue / Data Center | $96.2B (+106%) / $89.0B (+117%) | NVIDIA [21] |
| Q3 FY27 guidance (reports 17 Nov) | $108.0B ±2%; no China Data Center compute assumed | NVIDIA [21] |
| Direct reports to Huang | ~60 | ABC / Reuters [3,22] |
| Huang’s stake | ~3.61%, fourth-largest holder | Reuters via Economic Times [23] |
| Supermicro smuggling case value (alleged) | ~US$2.5B of servers | AP [24] |
| SemiAnalysis estimate of restricted Nvidia hardware reaching China | >1.5 GW | Tom’s Hardware [25] |
| Chinese models, share of Vercel enterprise token traffic | 11% (Jan) to 55% (Aug 2026) | Latin Times [26] |
2. Nine days that ended a 33-year run
The speed of the collapse is part of the story. Nvidia touched a record close of roughly $5.78 trillion on 6 October [10]. By late October the New York Times had reported that Huang was the target of a Chinese intelligence operation run through his relationship with Alice Huang; Wired dates that report to just under two weeks before the board opened its investigation on 8 November [5]. Huang told employees that same Sunday that he had been unfaithful to his wife, adding that there was “no excuse” for what he did [27].
The board first stood by him, then reversed. It retained Wachtell, Lipton, Rosen & Katz, led by a former acting US attorney, to investigate; on 11 November the New York Times reported a second relationship, this time with an executive who reported to him; and by the evening the board was meeting to consider his removal [6,28]. Reuters later reported a third, unnamed relationship with another employee [29]. Nvidia announced the resignation about 42 minutes before the US market opened on 12 November [22].
Wired’s source offers a useful reading of the board’s thinking: the alleged relationship with a subordinate, together with Huang’s business dealings with Megaspeed, probably weighed more heavily than the espionage allegations themselves [5]. That distinction matters later in this post, because it separates a governance failure (which boards can act on quickly) from a national-security failure (which prosecutors and Congress will take months to unpick).
The market moved violently and then partly retraced. Shares fell as much as 14% on the open on 12 November, which would have been the largest one-day market-value loss on record had it held, then recovered to close down 4.3% [9,30]. Retail investors poured in a record US$111 million in the first half-hour of trading, according to Vanda Research [31]. On 13 November the stock slid roughly another 4%, back under $4 trillion [32]. By one count about $550 billion of market value was lost in the week and over $1 trillion across five weeks [31]; by another the week’s damage exceeded $600 billion [28]. Treat these as approximate: the numbers moved by the hour.
(book p.117)
(first close above $4T)
(record)
(pre-exit)
(exit day)
Analysts had been bracing. Several told CNBC before the resignation that losing Huang could cost the stock about 20%, and at least seven firms downgraded Nvidia in the preceding week [30,32]. Forbes contributor Peter Cohan put the value at risk from losing Huang at between $700 billion and $1.1 trillion, around $600 billion of which had already gone [33]. Against that, the fundamentals the market is pricing are still extraordinary. Q2 fiscal 2027 revenue was $96.2 billion, up 106% year on year, with Data Center revenue of $89.0 billion and a 75% gross margin [21].
(book p.25)
(Jul 2025)
(Apr 2026)
(Jul 2026)
(reports 17 Nov)
To put that in the book’s terms: page 25 celebrated $13.5 billion in a record quarter and a 141% jump in Data Center revenue between quarters. Nvidia now books roughly seven times that in a single quarter. The business the next CEO inherits is an order of magnitude larger than the one the book described, and that scale is exactly why the succession question is so uncomfortable.
3. What the ChinaTalk panel said, chapter by chapter
The podcast was recorded in the immediate aftermath of the news. The panel (the host plus Dylan and Jordan of SemiAnalysis) debates two things at once: what Huang’s exit means for US export policy, and what it means for Nvidia and AI. Here is the argument in order, with the strongest lines kept verbatim [2].
Chapters 1–2: Jensen is out, and the legislative fork
The opening is half gallows humour, half policy alarm. Dylan’s first worry is not Nvidia but the political reaction: he fears the scandal “is going to be used as cover to pass really stupid laws.” He then draws a line between two bills. The MATCH Act would extend controls on advanced chipmaking tools such as deep-ultraviolet lithography to China; the Remote Access Security Act (RASA) would close the cloud loophole that lets Chinese companies rent banned GPUs through cutouts in places like Malaysia. Dylan wants the first and is hostile to the second: “RASA’s pretty dumb bill.” He also supports Biden-style diffusion controls, and the panel notes that framework was issued but never enforced.
MATCH “is far more important for US national interest than the RASA Act.”— Dylan, ChinaTalk
Chapters 3–4: why not ban everything, and will China retaliate?
Dylan’s game-theory case is the intellectual centre of the episode. His claim: if the US bans every route to compute, including rental, China has no reason not to “blow the whole thing up”, because Chinese suppliers sit inside American supply chains. He points to the largest PCB supplier to Nvidia being a Chinese company. The optimum, he says, is a regime that minimises what China can access and maximises American oversight.
“What is the minimum amount of compute that China can have with the most amount of American oversight to where they will not blow up the supply chain.”— Dylan, ChinaTalk
“We do not have the stomach to disentangle the supply chain, very clearly.”— Dylan, ChinaTalk
His practical conclusion is counter-intuitive for hawks: “the best thing for America to do is allow remote access of GPUs because the alternatives are much worse”. Hyperscaler-controlled rental gives Washington a kill switch, keeps Chinese data visible to US intelligence, and sends enormous margin dollars to Nvidia and the US clouds. He cites roughly $7 million of Nvidia profit per megawatt deployed plus another $7 to $10 million for the cloud provider at current pricing, a back-of-envelope figure rather than a reported one.
The host pushes back on several fronts. Beijing may actually welcome a ban, because parts of the Chinese state are embarrassed that its champions want Nvidia chips and would like them forced into the domestic ecosystem. Xi Jinping also has reasons not to start an economic war with an economy that, as one speaker puts it, isn’t growing. Others counter that earlier rounds of tit-for-tat ended in compromise, that Beijing did not retaliate against the 2022, 2023 or 2024 rules, and that this administration has shown it will absorb political pain (the speaker cites the Iran war). The disagreement is not resolved, and it is the most honest part of the episode.
Chapter 5: the smuggling evidence
This is where the conversation turns from policy to Nvidia’s proximity to wrongdoing. The panel lists what was then public: Supermicro’s co-founder arrested, a Supermicro Taiwan sales manager a fugitive, and the now-famous detail of a staff member moving serial-number stickers between servers during an audit. One panelist asks how Nvidia could stop reporting Singapore revenue, which this post has not been able to verify. One panelist is careful on the legal point: the bar for personal complicity would be Huang actually knowing and acting, but “either way” the pattern around him is damaging.
“The picture of the Supermicro Taiwan guy with the hair dryer in his hand moving the stickers… It’s pretty like damning evidence that they’re aware of this.”— ChinaTalk panel
Two further observations are worth keeping. First, the panel notes that the timing coincides with the midterms: “the cover is gone”, as one speaker puts it, and Lutnick and Huang lose protection together. Second, they point out that the Biden diffusion framework was never repealed, only waved away in a footnote in May 2025. The enforcement gap, not the rule book, is the recurring theme.
Chapter 6: China is getting off GPUs anyway
Dylan and Jordan argue that policy is arriving late to a trend already under way. GLM shipped with support for seven domestic accelerators; DeepSeek published a megakernel for Huawei Ascend for its V4 model; Meituan claims it trained LongCat entirely on Ascend. One panelist’s summary is the most quotable line about the state of play: “the gap in the model capabilities is wider at this point, but the gap in chip capabilities is like as close as it’s ever been.” Both observations are borne out by the later data in this post.
Chapter 7: flip the table
Dylan’s strategic alternative is to go after the bottom of the stack rather than the top. Remove one jigsaw piece and China will engineer around it; blow up the board and it is harder. So push China up the value chain into buying expensive things it would rather make cheaply.
“Don’t let them buy chips, make them rent the chips. Don’t let them rent the chips, make them buy the AI.”— Dylan, ChinaTalk
The panel converges on equipment: expand MATCH beyond EUV so the Japanese, Dutch and Koreans cannot backfill American restrictions, and make China build lithography from scratch. The panel also stresses scale: it is not enough for Huawei to match a phone chip or a Hopper-class accelerator; the test is whether China can make hundreds of thousands of them.
Chapter 8: open models and talent
The conversation then widens to why US labs have not simply out-spent Chinese open-weight models. Dylan’s answer is blunt: “It’s not money, it’s talent.” He argues that Moonshot is, skill for skill, as good as the US labs and is held back by compute and deployment. A panelist adds that Nvidia itself spent heavily on its Nemotron open models without beating the Chinese open releases. Dylan, playing devil’s advocate for Huang’s own argument, makes a point that cuts against the hawkish framing: American inference companies earn real money serving Chinese models. “America is making so much money off of the Chinese models.”
Chapters 9–10: Nvidia without Jensen, and who runs it next
This is the most useful part for investors. One panelist separates the short term from the medium term. In the short term NVIDIA is fine: it has commitments to meet and a supply chain to execute on Rubin and Rubin Ultra, and “nobody set them up for success as well as Jensen could, but now they just need to execute on the stuff he set up.” The test comes later and it is about people. Jensen has been unusually good at keeping executives, and without him, “this is probably when the sharks are in the water”: venture capital will pay billions to anyone who wants to leave and start a chip company or a neolab.
The bear case is operational. Dylan says that Nvidia’s speed depends on Jensen being a decision engine: “Things get solved slower, things get solved by committees, things don’t get escalated up.” He recalls the story that the Volta GPU had no Tensor Cores until three months before tape-out, when Huang ordered them added. Stack up a hundred decisions like that, he says, and Feynman could slip, Rubin yield could ramp slower, and by 2029 the best chips may not be Nvidia’s. The rebuttal from the others is that this is 2026, not 2022: there are “lots of other competent teams” and Nvidia remains the leader.
“If NVIDIA slows down, it’s not like someone else accelerated.”— ChinaTalk panel
“Hardware, it progresses slower without Jensen at the helm of NVIDIA, period.”— ChinaTalk panel
On the Steve Jobs analogy the panel splits. One side argues that buying Apple after Jobs stepped back would have made a lot of money. Another replies that “I don’t think Steve Jobs exerted as much control over Apple as Jensen does over NVIDIA or Elon does over his companies”, because Jensen negotiates supply contracts himself. The panel’s dream CEOs are Lisa Su, Charlie Kawwas of Broadcom and, jokingly, Hock Tan. On Kress, one panelist doubts that any senior executive can be wholly unaware of what happened with the chips, while others note she is not the one in Asia making agreements. This post has found no report alleging wrongdoing by Kress, and the panel’s remarks should be read as speculation about congressional testimony, not as a factual claim.
Chapters 11–12: finance, sales and technology — and the nationalisation debate
One panelist closes with a governance observation that deserves to be in every management textbook: every company has a permanent tension between sales (which wants deals), technology (which wants big bets) and finance (which has to pay for it). With Kress as interim CEO, “Finance is in charge now”, which probably means a profitable stretch followed by a different company in the medium term. The final minutes become an argument over whether the US government is effectively nationalising Nvidia, a theory one panelist floats and the others reject. The pushback is that “there was like so much between don’t break the law and nationalized”, and that any of the CEOs being floated are “just logical CEOs”. It is the least evidence-based exchange of the episode and is included here for completeness; nothing in the reporting supports state control of Nvidia.
4. It is economic warfare, and smuggling is the front line
The most useful sentence in the podcast is also the simplest: the US and China are in an economic war. Dylan says he wrote that into his reading of the October 2022 chip rules and still believes it. Read that way, three things that look unrelated are one story: a founder CEO who lobbied for access to the Chinese market, a procurement chain that leaked restricted hardware, and a customer base that learned to rent what it could not buy.
4.1 What the record shows so far
The strongest public evidence of diversion is not about Huang at all. In March 2026 the Southern District of New York charged Supermicro co-founder Wally Liaw, a Taiwan sales manager and a contractor with conspiring to divert about $2.5 billion of servers containing Nvidia chips to China. Prosecutors said the group used fabricated paperwork, a pass-through company and staged non-working dummy servers to satisfy audits [24]. The contractor pleaded guilty in October; Liaw and the Taiwan manager remain charged, the manager a fugitive when last reported, and Supermicro itself is not a defendant [35].
The Megaspeed thread is newer and murkier. Megaspeed is a Singapore data-centre company that became one of Nvidia’s largest buyers in Southeast Asia. Semafor reports Nvidia sold it about $4 billion of chips [34]. The US government suspects Megaspeed of directing advanced chips to China, and Alice Huang is a former Megaspeed executive [5]. Bloomberg reports that the Justice Department is running three probes: an antitrust inquiry into Nvidia and two others on Megaspeed and Supermicro, while Manhattan prosecutors are examining whether entities linked to Alice Huang were involved in moving Supermicro hardware to China. Neither Nvidia, Huang nor Supermicro has been charged [4]. Supermicro says Megaspeed has never been its customer [36].
The scale estimates are large but come from analysts, not courts. SemiAnalysis, the firm behind the podcast guests, estimates that more than 1.5 gigawatts of restricted Nvidia hardware has reached China, excluding chips designed to comply with the rules such as the H20; Tom’s Hardware notes the firm does not disclose unit counts and says the Commerce Department has reached a similar conclusion about scale [25]. Convert that at 700 to 1,000 watts per accelerator and you get something between 1.5 and 2.1 million GPUs, which is the right order of magnitude to explain why Chinese labs have been able to train frontier models.
4.2 The legal loophole the book already flagged
Smuggling is only half the story. The other half is perfectly legal. Page 27 of the book warned that, despite controls aimed at physical exports, “most Chinese technology firms access chips virtually using services offered by cloud computing companies”, and that Nvidia had called out intentions to expand cloud supercomputing in China. That is exactly the gap RASA is meant to close. The bill passed the House 369 to 22 on 12 January 2026 and has sat in the Senate Banking Committee since; the Senate text (S. 3519) is narrower than the House version and covers only infrastructure-as-a-service [37,38]. Commerce lawyers say they lack authority to police rented compute until Congress acts, and one analyst note estimates that the compute China reportedly accesses remotely could lift its capacity by at least 60% in 2026 [38,39].
The headline case study is Moonshot. A White House science official accused the lab in July of running workloads on Nvidia GB300 servers in a Thailand data centre, and the lab’s Kimi K3 model reportedly drew on around 20,000 H200 accelerators accessed through an Alibaba compute-sharing arrangement [39]. That is a single secondary source and should be read with care, but it matches what the panel described.
4.3 Misjudgement, wilful blindness or something worse?
The question the panel keeps returning to is what Huang knew. The evidence supports three distinguishable readings.
- Misjudgement. Huang genuinely believed that selling to the world, including China’s approved customers, was in America’s interest and that diversion was rare. Reuters notes that until the Justice Department brought indictments last year, Huang and Nvidia had argued there was no chip smuggling and that the company complied with export laws [40]. Trump said Huang told him he had no idea the woman was a spy [23].
- Wilful blindness. Nvidia’s own customers and partners were the vector. A company that books $4 billion from a single Singapore buyer and sees its partner’s staff move audit stickers is, on this reading, choosing not to look. The New York Times reported that Nvidia’s general counsel intervened in a Commerce Department investigation into Megaspeed, a claim reported in a newsletter republished by Senator Banks’s office and not independently confirmed here [41].
- Active complicity. No public evidence supports this. The legal test the panel identified, Huang knowing and acting, has not been met by anything on the public record, and he has not been charged [4].
The honest position for a book that must outlive the news cycle is to say that the first reading is not disproved, the second is plausible, and the third is unproven. What is not in doubt is the governance cost. A CEO who lobbied for China access, who was the target of an intelligence operation, and whose closest commercial partners were under investigation had given his board a problem it could not contain.
4.4 The workplace affair is a separate, and fully sufficient, problem
There is a second, simpler thread, and it may have been the decisive one. A relationship with a direct report is an ethics and governance issue wherever it occurs. Huang admitted unfaithfulness to his wife on 8 November; the allegation about Powell, reported by the Times on the word of five Nvidia employees and by Reuters for a second employee, concerns someone who reported to him [6,29]. Powell has not commented publicly on the report [30]. Wired’s source says this, together with the Megaspeed dealings, likely played a larger role in the board’s decision than the espionage allegations [5]. The Times has said it has not seen the video it describes, and CBS has not verified the report [6].
5. What is ahead for Nvidia: the control problem
The book described Huang as a visionary who “invented the GPU” and bet early on CUDA (book pp. 22–23). What it could not show is how the company was actually run. The reporting of the past week fills the gap, and it explains why this transition is harder than Apple’s in 2011 or Google’s in 2019.
5.1 A flat structure with Jensen as the integration layer
Huang held the roles of co-founder, president and CEO, and ran an unusually flat organisation with about 60 executives reporting directly to him [3]. Reuters describes a leader who “dug deep into startups’ technology”, helped them win customers inside Nvidia and even joined fundraising calls, personally overseeing the 60 reports and diving into problems across the company [22]. The panel’s anecdote that SemiAnalysis has had to email Jensen to get a fix to some open-source software is the same observation from the outside: in Nvidia, escalation ends at one person.
Two structural details compound the problem. The WSJ reported that Huang once said he did not believe in succession planning [3]. And he was the only Nvidia insider on a board that had no permanent chair until Harvey C. Jones was named on 12 November [1,3]. Nvidia’s own annual report warns that it is “highly dependent” on its long-standing executive team and that failure to ensure effective succession planning could hurt long-term success [10]. The risk was disclosed. It was not mitigated.
“The secret for being CEO for this long is 1) don’t get fired and 2) don’t get bored… How long [will I remain CEO]? For as long as I deserve it.”— Jensen Huang, January 2026, quoted by Forbes [13]
Forbes contributor Peter Cohan makes the sharpest point: with 60 direct reports each running a slice of strategy, “it is hard to see how any of those direct reports … could have enough of a view of the big picture to take over” [42]. That is the cost of a flat structure. It maximises speed and information flow while the leader is present and leaves no deputy with the integrated view when he is not.
5.2 Family firm characteristics
Forbes also draws attention to the family dimensions of the company: Huang’s wife co-owns the family stake through trusts, and both of their children work at Nvidia [42]. Huang is the fourth-largest investor with a stake of about 3.61% [23]. None of this is improper, but it complicates a clean break. A departing founder with a 3.6% holding, family employees and a record of personally brokering the company’s biggest relationships will remain a gravitational presence whatever the board says.
5.3 Who could run it?
The names in circulation show how thin the bench looks. Internally, Kress, who has been CFO since 2013 and spent nearly 25 years at Cisco and Microsoft before that, is the interim CEO and keeps the CFO job [1]. Forbes lists Michael Kagan, the CTO who founded Mellanox, and Jay Puri, who retired in July, and notes that Ian Buck, who helped develop CUDA and runs hyperscale and high-performance computing, has been floated although analysts say he lacks broad operational experience. Externally, The Information floated Lisa Su of AMD and TechCrunch floated Eric Schmidt [13]. Trump named Tim Cook [43]. The panel’s own wish-list was Lisa Su, Charlie Kawwas and Hock Tan [2].
The board has said it will consider internal and external candidates and has not given a timeline [1]. In the meantime the first test is the 17 November earnings call, the first under Kress, where Nvidia has guided to $108 billion of revenue, plus or minus 2%, against a consensus of about $109 billion [44]. The quarter closed before the scandal broke, so the numbers will describe the business as it was; the guidance and tone will show how the company is coping without its founder [44].
6. Can he come back? Jobs, Altman and the limits of the analogy
The question is natural because the book itself contains the best-known recent boomerang. Page 98 records that Sam Altman was sacked as OpenAI CEO on 17 November 2023 and reappointed on 22 November 2023, five days later, after a boardroom crisis that remade the board. And Steve Jobs, pushed out of Apple in 1985, returned in 1997 and built the company that Tim Cook then ran for 15 years. Founders can come back. So why is Huang’s case different?
6.1 The three differences
- The reason for leaving. Altman’s exit was a governance dispute inside a board that could not explain itself; Jobs’s first exit was a strategic fight. Huang left after an admitted affair, a reported relationship with a direct report, and a reported intelligence operation. Those are conduct and national-security issues, and they are not resolved by winning an internal argument.
- Who holds the cards. In 2023 OpenAI’s staff and investors held the leverage and used it. At Nvidia the board moved within four days of opening its investigation, and the cards now sit with the board, a law firm, prosecutors and Congress. Huang has been asked to give a closed-door transcribed interview to three House committees on 1 December and to the Senate Banking Committee, whose chair has warned of a subpoena; Reuters reported through CNBC that he agreed to testify before Thanksgiving [12,29].
- The legal perimeter. Bloomberg reports multiple Justice Department probes touching Nvidia, Megaspeed and Supermicro, though no one at Nvidia has been charged [4]. Trump has said the FBI is looking at the romantic and business ties [6]. A CEO who returns while those matters are live is a standing invitation to subpoenas.
6.2 Scenarios for Nvidia’s leadership
| Scenario | What would have to be true | Likelihood (author’s judgement) |
|---|---|---|
| A. Internal successor stabilises the company | Kress or a technical leader such as Buck wins the permanent job; top-60 talent stays; Rubin ships on time; customers keep ordering | Most likely |
| B. External heavyweight appointed | Board concludes it needs an outsider with chip credibility and Washington standing; the names floated are Lisa Su, Eric Schmidt and Tim Cook, each with obvious obstacles | Plausible, less likely |
| C. Huang in a founder-emeritus or adviser role | Investigations clear national-security concerns; the board wants his relationships without his authority | Possible, later |
| D. Huang returns as CEO within 12 months | Investigations conclude with no findings against him; staff and major customers revolt; Washington’s posture softens | Very unlikely |
| E. Huang joins a Musk company | No formal offer or talks exist today [46]; see section 8 | Low near term |
Huang’s own insistence that he would stay “for as long as I deserve it” is a clue to the board’s calculus. The company’s independent directors were willing to remove a founder who owns about 3.6% of the stock and whom the President of the United States urged to fight for his job [29]. A board that has made that call is unlikely to reverse it within a year.
There is also a market-structure argument against a comeback. The scandal has done real damage to Nvidia’s relationships in Washington. Reuters notes that the crisis threatened to implicate administration members, including Lutnick, who pushed with Huang for relaxing chip sales, and that Huang’s influence over US export policy was part of his value [29]. A newsletter republished by Senator Banks’s office said the reports are setting back Nvidia’s policy-and-influence operation in Washington “in a very real way” [41]. A CEO whose principal political asset has become a liability is a hard return to sell.
7. Washington: Trump, Lutnick, Bessent and the Democrats
7.1 Trump: loyal until the end, then a successor suggestion
Trump was Huang’s most prominent defender. On the Sunday that Huang admitted the affair, he said Huang had “done an amazing job for our country” and that Huang had told him he did not know the woman was a spy [23]. The New York Times reported that on Tuesday he urged Huang to fight for his job [29]. Wired and Forbes report that US intelligence officials briefed the President on the relationship last year, and he did not act on it [5,18]. He has also said the FBI is looking into the relationship, while adding, “I think he’ll be in good shape” [6,10].
After the resignation he went straight to the succession question on Truth Social:
“…there is only one man to take his place – TIM COOK of Apple. A Legend replaces a Legend!”— Donald Trump, Truth Social, 12 Nov 2026 [43]
Trump described Cook as Apple’s executive chairman, and nothing in the reporting suggests Cook is interested. The suggestion is best read as a signal about the kind of successor the White House wants: a supply-chain operator and government interlocutor rather than a technologist. The point matters because the next Nvidia CEO will be chosen by a board that has to answer to Congress, not to a President who is himself under questioning about how much he knew.
7.2 Lutnick: the friend who held the pen
Commerce Secretary Howard Lutnick is the administration figure most exposed. CBS reports he was the chairman and CEO of Cantor Fitzgerald, which held about $1.2 billion of Nvidia stock before he joined government, and that he developed a warm relationship with Huang [47]. He frequently consulted Huang on AI policy and praised him as one of America’s great executives, according to a Reuters report carried by the AFR [48]. Critics point to specifics. In 2024 the Commerce Department’s Bureau of Industry and Security opened an investigation into Nvidia’s sales to Megaspeed; according to the New York Times, it was abruptly transferred to the Justice Department. Commerce has not added a company to the Entity List since October 2025, and in June it reportedly held off adding DeepSeek, memory-chipmaker CXMT and more than 100 other firms [47,48].
“The real scandal here is the lack of enforcement of export controls.”— Chris McGuire, Council on Foreign Relations, via CBS News [47]
The White House says Trump “maintains complete confidence” in Lutnick, but senior officials are discussing whether he should be pushed out, according to Reuters and CNN, and Trump has fielded calls raising alarm about him [15,29]. Lutnick has agreed to testify to the Senate Banking Committee on 10 December [14]. Republican Senator Jim Banks said that “it appears” Commerce “used loopholes” to let chip sales occur despite the President’s direction [14], and Warren has demanded that Lutnick and Under Secretary Jeffrey Kessler explain why they have not closed export-control loopholes [49].
7.3 Bessent: “Commerce has the pen”
Treasury Secretary Scott Bessent went further than any other cabinet member on Friday 13 November. On CNBC he said Treasury, War, the US Trade Representative and State had favoured putting Megaspeed on the Entity List “for months, but Commerce has the pen”, and that he was “frustrated” that leakage of Nvidia chips to China “hurts our great companies in the AI race, gives the Chinese an advantage” [14,15]. Outlets differ on whether he cited A100 or H100 chips as the ceiling the President had set. Axios reported that Bessent had confronted Lutnick the previous week; Bessent declined to comment on reported tensions between them [14,47].
This is an unusual public break. A Treasury Secretary assigning blame to a colleague, on air, for an enforcement failure is a sign that the administration has decided the scandal is best survived by finding a Commerce-shaped culprit. It also gives Congress a ready-made narrative: the policy was clear, the agency with the pen did not use it.
7.4 Congress: bipartisan, and moving
The most consequential reaction is legislative. House Speaker Mike Johnson said there is a growing consensus to limit chip exports to China, and Democratic leader Hakeem Jeffries said he would work to attach the bills to the defence bill [17,29]. Senate Majority Leader John Thune told Semafor there is a “high probability” of new restrictions, and Senator John Kennedy put the odds at “better than even”, adding, “if you’re going to do it, try not to have an affair with a Chinese spy” [34]. Three measures are in play [34,40,41]:
- Chip Security Act. Location verification for exported AI chips, to detect diversion.
- AI Overwatch Act. Congressional review of advanced-chip export licences, and codification of the President’s policy against exporting the most advanced chips to countries of concern. The Senate version, in the Senate NDAA, extends that policy for 18 months and requires licences rather than blanket approvals [50].
- MATCH Act. Pushes allies to align their chipmaking-equipment controls with US levels. A lobbyist told Banks’s newsletter it is “a tougher one” because of concerns it could aggravate the global supply chain [41].
Nvidia was the biggest opponent of the Chip Security and AI Overwatch acts, according to a former congressional staffer cited by Reuters [40]. A senior House Republican aide called the moment “a big ‘I told you so’ moment to everyone peddling Nvidia talking points” [41]. Not everyone is cheering: a policy commentary argues the three bills risk handing China the AI race by making US suppliers less reliable and alienating allies [51]. And there is no deal yet to include them in the final NDAA [34].
7.5 The Democrats: oversight first, antitrust not yet
A natural question is whether any Democrat is pushing this toward an antitrust case. In the reporting reviewed here the answer is no: the Democratic effort is about oversight, records and the President’s conduct. Elizabeth Warren says Huang’s resignation “does not absolve him, other executives, or the company” and that they “still owe Congress answers” about export-control compliance [49]. Warren and Chris Coons asked Rubio, Stephen Miller and Lutnick to preserve records on Nvidia’s chip exports [16]. Democrats are targeting not only Commerce but Trump himself for defending Huang, and Jeffries is aiming to put the chip bills in the defence bill [17]. Chuck Schumer wants all three bills passed [34]. Warren has also co-signed with Republican Banks [52], which is why this is not a partisan story.
Antitrust does exist as a track, but at the Justice Department: Bloomberg reports that one of its three probes is an antitrust inquiry into Nvidia [4]. Warren-style language about “Big Tech billionaire CEOs” suggests the political appetite is there [49], but this post found no Democratic call for a competition case. That is a gap to watch: if Nvidia’s financing web [7] becomes the focus, an antitrust framing would be a natural extension.
8. Could Jensen end up at SpaceX, and would Musk be fine with it?
The short answer: Musk has said he would be, there is no sign of an offer, and the practical obstacles are bigger than the headlines suggest.
8.1 What Musk actually said
Late on 11 November, before Nvidia announced anything, the financial blog ZeroHedge posted “Elon please hire Jensen”. Musk replied: “That would be an honor” [45]. Earlier in the week he had agreed with a Sequoia partner who called Huang “irreplaceable for America’s AI leadership”, and after the resignation he replied “True words” to a post saying one does not simply replace a founder [46,53]. On Friday he reposted the “Air” monologue about Michael Jordan with a one-word caption: “Jensen” [54]. Business Insider and others stress that this is an expression of interest, not a job offer; IBTimes reports no sign that the two have discussed an arrangement [45,46].
Investors read it as a signal. SpaceX shares rose roughly 3% to 4% on the day Nvidia fell, at a market value of about $1.96 trillion [55].
8.2 Why the pairing makes commercial sense
- Nvidia owns a slice of Musk’s empire. Nvidia holds about 122.8 million SpaceX shares, received through its investment in xAI before SpaceX acquired xAI [55], valued at about $21 billion in its August 13F filing [56].
- SpaceX is a very large Nvidia customer. SpaceX has committed to a $40 billion lease for Nvidia chips and equipment [55], told investors in August it would use Nvidia GPUs exclusively, and has set a 10-gigawatt compute target for the end of 2027, a build-out reportedly costing about $500 billion [56].
8.3 Why it is harder than it looks
- Legal overhang. Huang faces a law-firm investigation, congressional interviews and a political environment in which any employer would draw scrutiny. SpaceX is a public company with government contracts, including NASA lunar work [55]. Any role touching national-security work would raise vetting questions. That last point is the author’s inference, not a reported fact.
- Conflict of interest. Nvidia is simultaneously SpaceX’s key supplier and a shareholder. A former CEO working for the customer on the other side of supply negotiations would need careful structuring.
- Musk’s own direction. One outlet reports Musk is working to reduce reliance on third-party chips through Tesla’s custom accelerators and a long-term “Terafab” fabrication project [57]. A Huang hire would sit awkwardly with a strategy of reducing dependency on Huang’s former company.
- Precedent cuts both ways. Business Insider notes the 2010 Ellison–Hurd precedent: HP’s board forced Hurd out and Ellison hired him within weeks [45]. It shows speed is possible. It also shows the hire came from a founder-controlled company with a friendly board.
Judgement: an operating role at SpaceX or xAI within the next six months is unlikely. A looser relationship – adviser, board seat or investment partner – is more plausible after the investigations settle. Musk’s enthusiasm is real, but it is a signal about loyalty and positioning more than about an imminent appointment.
9. The circular dependency: Nvidia as the “central bank of AI”
Circular dependency is the AI-relevant theme, and it deserves its own section because it is where Huang’s exit can travel furthest beyond Nvidia. This section uses “moving into the middle” to mean Nvidia’s push beyond selling chips into the financing and cloud layers that sit between silicon and applications.
9.1 How the loop works
Huang used Nvidia’s swelling cash pile to bankroll the industry that buys its chips. Reuters lists investments spanning AI cloud providers such as CoreWeave and Nebius and suppliers to the biggest labs [22]. CNBC calls the result the “central bank of AI”: Nvidia has made hundreds of billions of dollars of commitments, including cloud-service purchases and project guarantees, and the Federal Reserve is watching the interconnected financing for the possibility that an unexpected event could trigger what Fed officials discussed in July as a “broad-based repricing of assets” [7]. Jon Peddie Research is blunter: Nvidia “is shifting from a tech company to a finance company”, with $279 billion of supply and capacity commitments, $29 billion in cloud agreements, and credit support of up to $105 billion tied to an Ohio campus for OpenAI [58].
In August Nvidia also signed memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to mobilise more than $500 billion of third-party capital, using GPUs as collateral in structures resembling real-estate or toll-road securitisations. The arrangements are preliminary [59]. The vision depends on GPUs being bankable assets, and a founder with the credibility to persuade asset managers that they are.
9.2 What the exit does to the loop
Goldman Sachs analyst James Schneider expects Nvidia to honour all announced commitments but says it “may be more difficult for the company to negotiate new multi-party investments among ecosystem partners at the same scale” [8]. The market is already pricing stress. The yield on Nvidia’s 4.5% bond due 2031 was 5.77% on Friday, nearly 6% higher on the week, and credit-default-swap spreads have widened, most in lower-rated tech names [8]. The smaller counterparties are the weak link: Nvidia had $36 billion of off-balance-sheet cloud agreements, or backstops, for neoclouds at its last quarterly filing, and CoreWeave, Nebius and IREN all fell sharply [8].
Not everyone sees a systemic threat. Former FDIC chair Sheila Bair told CNBC Nvidia is not “too big to fail in a bank sense at all”, because it is not exposed to a run and would still make a product customers want, though “Nvidia going down would be a pretty seismic event” and regulators should watch bank exposure [7]. Ben Barringer of Quilter Cheviot made the same split between short and long term: the news “doesn’t necessarily harm Nvidia in the short-term”, but “what this news threatens to do is harm the wider AI ecosystem in the longer-term” [60].
Sovereign deals are the least transparent part of the web. The Guardian notes that Huang spent 2025 persuading world leaders to develop “sovereign AI” with Nvidia chips, with deals that were not particularly transparent; Nvidia appears to have promised 260,000 advanced chips to South Korea without it being clear when they would be delivered or who would pay [61]. Those are relationship-based promises, and relationships were Huang’s personal asset.
9.3 The Anthropic IPO is the first live test
Anthropic is days from an IPO targeting about a $1.4 trillion valuation, and had said it would take $10 billion from Nvidia. Bloomberg Intelligence’s Mandeep Singh wrote before the resignation that Nvidia’s share of an IPO-day investment “could face heightened scrutiny” if Huang were removed and that a transition “could prompt Nvidia’s board to reassess the size, timing or strategic rationale of the commitment”. PitchBook’s Harrison Rolfes disagrees: “It has nothing to do with Jensen’s overall control. It’s just: ‘We made the investment, and we’re good to go’” [19]. Whichever is right, the Anthropic outcome will be read as a signal of whether the loop survives its architect.
The book anticipated the shape of this risk without naming it. Page 26 compared the AI boom with the dot-com bubble: every start-up with venture funding bought servers from Sun, storage from EMC and databases from Oracle, those three “made a lot of money for several years”, then the bubble burst and the survivors were those on cheaper open technologies. The 2026 twist is that the vendor is also the lender. In the dot-com era Sun did not underwrite its customers’ data centres.
10. Who gains and who loses: AMD, Google, Amazon, Chinese models, OpenAI, Anthropic and the hyperscalers
The right order is to identify the dependency first, then see who can exploit it. The answer is that Huang’s exit helps challengers in narrative and may help them in substance, but nobody escapes Nvidia’s gravity in the next two years.
10.1 The scale of the dependency
Estimates of Nvidia’s share vary by definition. Forbes says it generated about 85% of global AI-chip revenue, with share expected to fall to about 75% this year [62]; Al Jazeera cites about 90% of the market for advanced AI chips [63]; the book put it at more than 70% in 2023 (p.22). The direction is down, not the level. One industry analysis projects Nvidia’s inference share falling from about 90% to 20–30% by 2028, but notes the key custom chips – Trainium, TPU, Maia, MTIA – are captive to their builders and cannot be rented outside them [64]. That qualification matters: it limits how much of Nvidia’s business the hyperscalers’ chips can take directly.
10.2 The first-day market read
On the day of the resignation AMD and Broadcom rose about 4.7% and Qualcomm about 2.7%, while Intel slipped 0.8%, all by 11 a.m. ET [18]. Elsewhere, Marvell fell 2.6%, Micron 1.6% and Dell 0.7%, and Supermicro dropped about 11% on a separate DOJ report [36,60]. The Seoul Economic Daily lists Samsung, AMD, Broadcom and Google as potential beneficiaries of an unwinding of Nvidia-centred circular deals [65]. Analysts quoted by Mint expect AMD’s Instinct accelerators to compete head-on with Nvidia and Broadcom to gain from big tech’s preference for custom accelerators [66].
10.3 Google: the TPU is the credible alternative
Google’s case is the strongest because it owns both the chip and the model. In April, Alphabet reported Google Cloud revenue up 63% to $20 billion with backlog nearly doubled to $462 billion; the eighth-generation TPU offers three times the processing power of Ironwood; and Sundar Pichai said Google will deliver TPUs to a select group of customers in their own data centres this year, with most revenue from 2027 [67,68]. Morgan Stanley estimated that selling 500,000 TPUs could add about $13 billion to Google’s 2027 revenue [68]. On 20 October Google released TorchTPU as open source, letting PyTorch applications move to TPUs with a configuration change [69].
This is the scenario the book flagged in 2024. Page 29 identified Chris Lattner’s Modular as a threat to Nvidia’s software moat, and page 42 argued Google’s TPUs could deliver 40–70% cost savings on some workloads. TorchTPU and Amazon’s TorchNeuron are the industrial version of that argument: reduce the switching cost that made CUDA a mousetrap.
10.4 Amazon: Trainium sold to Nvidia’s own customers
Amazon’s chips business topped a $20 billion annual revenue run-rate, growing at triple digits, with Trainium commitments above $225 billion; Trainium2 is about 30% cheaper per unit of performance than comparable GPUs, and Anthropic, OpenAI and Meta are named customers. AWS grew 28%, its fastest in 15 quarters [67]. Andy Jassy said there is a good chance Amazon will offer full racks of Trainium beyond its own cloud over the next couple of years [68]. Amazon released TorchNeuron on 19 October and is submitting code to decouple PyTorch features from Nvidia hardware [69].
10.5 OpenAI and Anthropic: more dependent on Nvidia, less constrained by Huang
These are the labs most exposed to the financing loop and also those with the most to gain from the political reset. Business Insider’s reading of Anthropic is the clearest: Dario Amodei has battled with Huang for influence in Washington over chip exports, Huang used his perch to support open-source models that compete with Anthropic’s, and Anthropic’s diversified chip strategy now looks wise. The risk is that its race for compute relies in part on Nvidia-backed neoclouds [19]. OpenAI reportedly told lawmakers it supports the Chip Security, AI Overwatch and MATCH acts, a pivot from a company with deepening financial ties to Nvidia, according to a trade-press summary of reporting by Politico and Bloomberg Law that this post has not independently confirmed [20]. It also has up to $105 billion of Nvidia credit support attached to its Ohio campus [58].
10.6 The hyperscalers: dependent customers who are also competitors
Forbes notes the exit “generated great uncertainty for hyperscalers and neoclouds” that rely on Nvidia GPUs [62]. Hyperscaler bond yields rose slightly on Thursday on the persistent concern about AI debt [8]. But the hyperscalers are in the best position of any group: they buy Nvidia chips, build their own, and increasingly sell the alternatives. Their bargaining position with Nvidia improves with every month Nvidia lacks a permanent CEO, because the person who set prices, allocations and financing terms is gone.
10.7 Chinese models: a nuanced winner
A common thesis is that Huang’s loss is good for Chinese models. The reality is more complicated. Narratively, yes: every report about Nvidia’s weakness strengthens the case for Huawei-based stacks and for open-weight models trained on domestic chips, and Huang was the most effective lobbyist for access to the Chinese market. Practically, no: the immediate legislative effect of his exit is tighter controls on exactly the compute Chinese labs rent offshore, and that is bad for Moonshot and its peers in the short run. The panel’s argument is that the substitution toward domestic silicon is happening “with or without the remote access laws being passed”, so the controls arrive late and mostly raise the cost of the transition.
One closing observation from the panel is worth keeping for the book. One panelist’s summary of the industry impact was brief: “this is a bad day for AI progress.” The case for it is that US, Chinese and Taiwanese engineers collaborate across Nvidia, AMD and the labs, and an escalation breaks those links [2]. The case against it is that the technology is now deeper than any one person. Both can be true.
11. China versus the US: the chip gap closes, the model gap is contested
11.1 The panel’s framing, tested
The panel’s clearest empirical claim is that the model gap is wider than it looks but the chip gap is “as close as it’s ever been”. Reporting a few months old supports both halves, with caveats.
- Chips. Huawei’s Ascend line sits at the centre of China’s stack. DeepSeek’s V4, released on 24 April, was adapted for Ascend, and Alibaba says its in-house T-Head GPUs have reached scaled mass production [71,72]. The panel’s own list of domestic accelerators supported by GLM (Moore Threads, Hygon, Cambricon, Kunlunxin, MetaX, Enflame) is about breadth.
- Models. Moonshot’s Kimi K3, released on 16 July, is a 2.8-trillion-parameter model with a one-million-token context window, which the company says benchmarks neck-and-neck with the top proprietary US systems; benchmarks are lab-reported and not independently reproduced [73]. DeepSeek V4-Pro has 1.6 trillion total parameters (49 billion active), an MIT licence and a $0.44 per million input-token price [26].
- Adoption. On Vercel’s enterprise developer gateway, Chinese models rose from 11% to 55% of token traffic between January and August 2026, driven mainly by DeepSeek V4 and Kimi K3. That is a single platform, but the direction is unmistakable [26].
(book p.14)
V4-Pro
Max
The US response is not standing still: Thinking Machines’ Inkling and Meta’s Apache-licensed models are US open-weight answers, according to a November overview [74], and Beijing was drafting its own restrictions on exports of advanced AI models on 16 October, with already-released open weights exempt [26]. The race has become two-sided, which is why the panel’s “flip the table” logic is about equipment and value-chain position rather than any single chip.
11.2 The H200 story: licences granted, deliveries throttled
Washington authorised H200 sales to China in December 2025, with 25% of revenue going to the US government [75]. By mid-May about ten Chinese companies, including Alibaba, Tencent, ByteDance and JD.com, were licensed to buy up to 75,000 each, yet no chips had been delivered [76]. The reason was Beijing, not Washington: Chinese authorities told companies to limit their use of Nvidia chips to overseas operations and support domestic suppliers, while US rules require H200s to be used in China [71]. In August, ByteDance and Tencent each received about 10,000 units, around 13% of their quota, and every further purchase needs individual approval from China’s National Development and Reform Commission, which also forces buyers to justify why they cannot use domestic alternatives [70].
(book p.29)
(CNBC)
(AI News, Invezz)
China Hopper, % of Data Center
China DC compute
The numbers explain why the panel thinks Nvidia’s China business is largely gone for the purposes of valuation. The book (p.29) quoted China at around 21% of revenue; reports now put it at roughly 5% in recent quarters, China Hopper shipments were under 1% of Data Center revenue in Q2 FY27, and Nvidia’s Q3 outlook assumes no China Data Center compute revenue at all [21,58,71]. What remains is indirect exposure – through offshore rental, through Southeast Asia and through whatever the diversion estimates imply.
11.3 Policy matrix: where the controls bite
Dylan’s argument can be drawn as a matrix. The horizontal axis is the control point: restricting who may use compute, or who may build it. The vertical axis is blowback to US firms and supply chains. His preference is to apply pressure where the US already has leverage and the blowback is lowest.
11.4 Shenzhen next week
The politics converge on 18–19 November, when Trump is expected to meet Xi around the APEC summit in Shenzhen with trade, rare earths and advanced chips on the table. Outlook India notes that Washington has allowed case-by-case H200 licensing to approved Chinese customers while Congress pushes a law to regulate remote access to US-controlled compute, and that the Huang scandal has added urgency [77]. The previous Trump–Xi summit, in Beijing in May, took Huang along at the last minute and produced no chip deliveries [71]. This one will take place without him and with a Congress that has more leverage.
12. What to watch next
| Date | Event | Why it matters |
|---|---|---|
| 17 Nov | Nvidia Q3 FY27 results (~1:20pm PT) and call (2pm PT); guidance of $108B ±2% against consensus near $109B [44] | First report under Kress; guidance and tone on supply, financing and succession |
| Early the following week | Anthropic IPO targeted at ~$1.4T [19] | Tests Nvidia’s planned $10B participation |
| 18 Nov | House GOP chairs’ requested closed-door briefing with Huang [17] | First formal congressional exposure |
| 18–19 Nov | Trump–Xi meeting around APEC, Shenzhen [77] | H200, remote access and rare earths |
| Before Thanksgiving | Huang reportedly agreed to testify [29] | Date and format to be confirmed |
| 1 Dec | Closed-door transcribed interview requested by House Foreign Affairs, Intelligence and China committees [12] | Subpoena threatened if he declines |
| 10 Dec | Lutnick to testify to Senate Banking Committee [14] | Entity List delay; his future as Commerce Secretary |
| Rolling | Senate–House NDAA negotiations on the Chip Security, AI Overwatch and MATCH acts [34] | Whether the bills reach law this year |
| Unscheduled | Wachtell investigation findings; DOJ probes of Megaspeed and Supermicro; permanent CEO announcement [1,4] | Determines whether any Huang return is even discussable |
Five predictions, labelled as judgements
- Congress passes at least the Chip Security Act and the AI Overwatch Act in some form this year; MATCH is the one most likely to be diluted or delayed. Thune’s “high probability” and Kennedy’s “better than even” support this, but no deal exists yet [34].
- Nvidia’s Q3 numbers will be strong, because the quarter closed before the scandal, and the stock will react to the guidance, supply commentary and any financing-commitment language rather than to the print [44].
- Nvidia names a permanent CEO later rather than sooner. A board that has just created a chair and promised “internal and external” candidates is not looking to rush, and an interim CEO who is also CFO is a stabilising rather than strategic choice [1].
- Lutnick’s position is the most fragile in the cabinet; a departure before or after 10 December is plausible, given Bessent’s public blame and reports of internal discussions [15,29].
- Hyperscaler and neocloud financing terms tighten at the margin; the first visible sign will be in credit spreads and the next multi-party deal that fails to close rather than in any single announcement [8].
13. Takeaways
- The scandal is the trigger; the structure is the cause. A 60-report flat organisation, no chair, one insider and no succession plan turned a personal crisis into an industry event [3].
- Export policy has changed direction. The reaction is bipartisan and aimed at enforcement. Expect the bills the panel debated to move, and expect Commerce to be on the defensive [14,34].
- Economic war is the correct frame. China is substituting domestic chips and open models faster than controls can act, and the legal rental loophole matters as much as illegal smuggling [2,26].
- Circularity is the amplifier. Nvidia’s role as financier makes its governance a systemic question, which is why a CEO resignation moved credit spreads [8].
- A comeback is possible in theory and improbable in practice. Jobs and Altman returned because of leverage and time; Huang is in a different category [78].
- The challengers are real but constrained. Google and Amazon have the chips and the software bridges; AMD and Broadcom have the narrative; none can take over Nvidia’s role in two years [64].
“The graveyards are filled with indispensable men.”— Lloyd Blankfein, quoted by CNBC [78]
Numbers in square brackets in the text refer to this list, in order of first appearance. Book page references (“book p.X”) refer to the author's book Gen AI – The New Reality.
- NVIDIA, “NVIDIA Announces CEO Transition”, 12 Nov 2026 — link
- ChinaTalk podcast, emergency episode on Jensen Huang’s resignation (transcript supplied by the author; speakers are not labelled in the transcript) — link
- ABC News (Australia), “Nvidia without succession plan as CEO Jensen Huang resigns”, 12 Nov 2026 — link
- Fortune via Yahoo Finance, “Before Megaspeed, there was Supermicro”, 12 Nov 2026 — link
- WIRED, “Jensen Huang Resigns as Nvidia CEO”, 12 Nov 2026 — link
- CBS News, “Nvidia CEO Jensen Huang resigns in wake of affair with woman suspected of spying for China”, 12 Nov 2026 — link
- CNBC, “Nvidia’s leadership crisis tests the chipmaker’s role as the ‘central bank of AI’”, 11 Nov 2026 — link
- CNBC, “What’s ahead for AI financing after Nvidia CEO Jensen Huang departs”, 13 Nov 2026 — link
- CNBC, “Nvidia stock closes down over 4% after CEO Jensen Huang announces resignation”, 12 Nov 2026 — link
- The Motley Fool, “Nvidia Stock Has Wiped Out $1.7 Trillion in Value Since October”, 10 Nov 2026 — link
- RoboForex, “Nvidia Earnings Date: 17 November Scenarios With an Interim CEO”, 13 Nov 2026 — link
- Euronews, “Ex-Nvidia CEO Huang and US Commerce Secretary Lutnick summoned to testify before Congress”, 13 Nov 2026 — link
- Forbes, “Jensen Huang Resigns From Nvidia—Trump Says Tim Cook Should Replace Him”, 12 Nov 2026 — link
- CBS News, “Lutnick agrees to testify as Bessent blames him for not banning Nvidia chip exports to China-linked firm”, 13 Nov 2026 — link
- CNN via Yahoo News, “Bessent points finger at Commerce Department in Nvidia scandal as Trump advisers discuss possible Lutnick ouster”, 13 Nov 2026 — link
- Fox Business, “Democratic senators demand NVIDIA records from Trump administration over alleged Chinese spy ties”, 12 Nov 2026 — link
- Seoul Economic Daily, “‘Huang Scandal’ Widens as Republicans Join Call to Halt AI Chip Exports to China”, 10 Nov 2026 — link
- Forbes (Fiona Riley), “Nvidia Sinks 11% At Open After Jensen Huang Resigns”, 12 Nov 2026 — link
- Business Insider, “How Jensen Huang’s dramatic ouster from Nvidia affects Anthropic’s $1.4 trillion IPO”, 13 Nov 2026 — link
- In Plain English, “OpenAI Backs Chip Security Act, AI Overwatch, and MATCH Act”, 13 Nov 2026 — link
- NVIDIA, “NVIDIA Announces Financial Results for Second Quarter Fiscal 2027”, 26 Aug 2026 — link
- Reuters, “Jensen Huang steps down: the hands-on kingmaker behind the AI boom”, 12 Nov 2026 — link
- Economic Times (Reuters), “Nvidia CEO Jensen Huang resigns, Colette Kress appointed interim CEO”, 12 Nov 2026 — link
- AP News, “3 men are charged with conspiring to smuggle US artificial intelligence to China”, 19 Mar 2026 — link
- Tom’s Hardware via Yahoo, “Mass smuggling of Nvidia AI GPUs could be the underlying cause of Jensen Huang’s departure, analysts claim”, 13 Nov 2026 — link
- Latin Times, “DeepSeek and Kimi K3: The Open-Source Strategy Turning Chinese AI Into a Geopolitical Force”, 22 Oct 2026 — link
- SAYS, “Jensen Huang Resigns As NVIDIA CEO As Report Of 3rd Alleged Affair Emerges”, 13 Nov 2026 — link
- The Guardian, “Jensen Huang resigns as Nvidia CEO after alleged affairs”, 12 Nov 2026 — link
- Reuters, “Nvidia CEO Jensen Huang resigns in blow to AI boom he helped foster”, 12 Nov 2026 — link
- Quartz, “Nvidia market cap faces biggest one-day drop ever after Jensen Huang resigns”, 12 Nov 2026 — link
- Korea Daily, “Nvidia Stock — US Retail Investors Pour $110M Following Crash”, 13 Nov 2026 — link
- CNBC, “Nvidia stock sinks 4% in fallout from Jensen Huang resignation”, 13 Nov 2026 — link
- Forbes (Peter Cohan), “Jensen Huang Is Out. Can Nvidia Protect $3.3 Trillion In Gains?”, 12 Nov 2026 — link
- Semafor, “‘A high probability’: Congress pushes tighter export controls following Nvidia scandal”, 12 Nov 2026 — link
- Yahoo Finance, “Super Micro Contractor Pleads Guilty in $2.5 Billion AI Server Smuggling Case”, 10 Oct 2026 — link
- Stocks Down Under, “Why Is Super Micro Stock Crashing 11%? DOJ Probes China Chip Smuggling Allegations”, 13 Nov 2026 — link
- Crowell & Moring, “House Passes Remote Access Security Act”, 14 Jan 2026 — link
- Institute for AI Policy and Strategy, “Remote Access Security Act (RASA)”, 15 Jul 2026 — link
- NewsNYou, “The US Banned These Chips From China. China Rented Them Instead.”, 2 Sep 2026 — link
- Reuters via Deccan Chronicle, “Nvidia Spy Allegations Likely To Spur Laws To Tighten China’s Access To AI Chips”, 10 Nov 2026 — link
- Sen. Jim Banks, “AI & Tech Brief: Nvidia’s Capitol Hill fallout”, 11 Nov 2026 — link
- Forbes (Emma Waldman), “What Jensen Huang’s Exit Means For Nvidia And Investors”, 12 Nov 2026 — link
- CNBC-TV18, “Jensen Huang’s resignation: From Elon Musk to Donald Trump, what top leaders said”, 13 Nov 2026 — link
- Invezz, “Nvidia earnings preview: what to expect as Kress faces first test without Huang”, 13 Nov 2026 — link
- Business Insider, “Elon Musk says he would hire Jensen Huang if Nvidia’s board removes him”, 11 Nov 2026 — link
- International Business Times (Australia), “Will Jensen Huang Join Elon Musk?”, 13 Nov 2026 — link
- CBS News, “Lutnick should have prevented the diversion of sensitive Nvidia AI chips to China, critics charge”, 13 Nov 2026 — link
- Australian Financial Review, “Lutnick fights for his job in Nvidia spy scandal fallout”, 13 Nov 2026 — link
- US Senate Banking Committee (minority), “Senator Warren Statement on NVIDIA CEO Jensen Huang’s Resignation”, 12 Nov 2026 — link
- Sen. Jim Banks, “Banks Secures AI OVERWATCH Act in Senate NDAA”, 14 Jul 2026 — link
- International Policy Digest, “Congress Could Hand China the AI Race by Trying to Win It”, 8 Nov 2026 — link
- Sens. Banks and Warren, joint statement on NVIDIA, 9 Nov 2026 — link
- The Hill, “Elon Musk floated hiring Jensen Huang ahead of Nvidia resignation”, 12 Nov 2026 — link
- Telegraph India, “Is Elon Musk hiring Jensen Huang? Buzz in ‘Air’ after richest man posts clip”, 13 Nov 2026 — link
- International Business Times (Australia), “SpaceX Stock Defies Tech Selloff, Rising 3% as Nvidia Plunges”, 13 Nov 2026 — link
- 24/7 Wall St. via Yahoo Finance, “Elon Musk and Jensen Huang’s New Partnership Could Create the Next Era of Technology”, 16 Aug 2026 — link
- One America News Network, “Musk: Hiring fmr Nvidia CEO Jensen Huang ‘would be an honor’”, 12 Nov 2026 — link
- Jon Peddie Research, “Nvidia financial results for Q2’26–FY’27”, 27 Aug 2026 — link
- CNBC, “Jensen Huang’s Nvidia exit casts shadow over Wall Street’s $500 billion AI financing bet”, 12 Nov 2026 — link
- Business Insider, “Stocks drop as Jensen Huang’s resignation drags Nvidia and rattles AI bulls”, 12 Nov 2026 — link
- The Guardian, “Jensen Huang built an AI kingdom. Now he’s gone, will Nvidia’s ramparts be breached?”, 12 Nov 2026 — link
- Forbes (Mike Patton), “Nvidia In Turmoil: What Jensen Huang’s Departure Means For A.I.”, 12 Nov 2026 — link
- Al Jazeera, “Why did Nvidia CEO Jensen Huang resign, and who is Colette Kress?”, 12 Nov 2026 — link
- Spheron, “Hyperscaler Custom AI Chips in 2026”, updated 19 Oct 2026 — link
- Seoul Economic Daily, “Nvidia’s Huang Forced Out, Shaking AI Circular-Deal Ecosystem”, 13 Nov 2026 — link
- Livemint, “NVIDIA share price crash: Broadcom, AMD to Micron”, 13 Nov 2026 — link
- 24/7 Wall St., “The Race to Beat Nvidia: Does Google or Amazon Have the Better In-House Silicon”, 17 Jul 2026 — link
- Business Insider, “Nvidia’s $4.9 trillion chip empire has a new problem: its biggest customers”, 30 Apr 2026 — link
- TechCrunch, “Google has joined Amazon in attacking Nvidia’s stronghold with AI app developers”, 20 Oct 2026 — link
- Tech Times, “Nvidia H200 Chips Enter China at 13% of Quota; Beijing, Not Washington, Controls Rest”, 20 Aug 2026 — link
- AI News, “The Nvidia H200 China deal survived the Trump-Xi summit–just not in the way anyone expected”, 19 May 2026 — link
- Safeguard, “Chinese AI labs in 2026: DeepSeek, Qwen and Kimi releases and direction”, 19 Oct 2026 — link
- VentureBeat, “China’s Moonshot AI releases Kimi K3, the largest open-source model ever”, 16 Jul 2026 — link
- GeoToolbox, “Best Chinese AI Models (2026): DeepSeek, Qwen, GLM, Kimi”, updated 9 Nov 2026 — link
- Council on Foreign Relations, “The Consequences of Exporting Nvidia’s H200 Chips to China”, 9 Dec 2025 — link
- CNBC, “U.S. clears H200 chip sales to 10 China firms as Nvidia CEO looks for breakthrough”, 14 May 2026 — link
- Outlook India, “A New AI Dialogue, An Old Chip War: What Trump And Xi Face In Shenzhen”, 13 Nov 2026 — link
- CNBC, “Nvidia after Jensen Huang: What founder transitions show”, 13 Nov 2026 — link
