I started a stock analysis on Nvidia when all hell broke loose. Jensen Huang, the CEO who co-founded Nvidia in 1993 and turned it into the world's largest company over 33 years, resigned abruptly. On 6 October 2026 Nvidia was worth $5.77 trillion, 19% more than Apple's $4.85 trillion that day. On 12 November it closed at $166.19, 30.5% below its record close of $239.24, and at the intraday low of $148.80 it was down 37.8%. Apple, at $5.16 trillion, is the largest company in the world again, now 29% larger than Nvidia.[1]
This article follows the argument of the video: the crash and the strange way it unfolded, what I call the drip conspiracy and who might be behind it, why Huang had too little control to resist, the parallels with Travis Kalanick and Elon Musk, what part of the crash was the market and what part was Nvidia, then the business itself, what is going well, the signs of trouble, the buildout, the competition, the strategy under Huang and after him, and finally the outlook. Market figures are anchored to the 12 November 2026 close unless another date is given. Nvidia's fiscal year ends in late January, so fiscal 2026 ran from February 2025 to January 2026, and the latest reported quarter is the second quarter of fiscal 2027, which ended on 26 July 2026.
The Nvidia Files
The fall is deep but not unprecedented. On closing prices it is the fourth-deepest since 2020, after the 2021 to 2022 crash (-66.4%, more than twice as deep), the 2020 Covid crash (-37.6%) and the 2025 tariff crash (-36.9%). On intraday prices the 2025 tariff low was -42.0%, deeper than this month's -37.8%. Every earlier fall ended at a new high; this one is still open.[1]
The valuation is now lower than at the bottom of the 2022 crash. At $166.19 Nvidia trades at 21 times the earnings of the last four quarters ($7.91 a share), down from 23.6 times on 6 November. On analysts' consensus of $15.68 for next fiscal year it trades at 10.6 times, down from 11.9. At the October 2022 low, by our math, the trailing P/E was 36.5, and at the November 2021 peak it was 103.[1][2]
Normally I first go into the crash and its cause, then describe the business strategy. In this case the situation was so strange that it looked like a conspiracy. The relationship between Jensen Huang and Megaspeed's Alice Huang, the woman US intelligence later called a Chinese asset, had been public for 5.5 months already. On 13 May 2026 the short seller Culper Research flagged an "uncomfortably close relationship" between the two. The stock rose 2.3% that day and 4.4% the next: the market ignored it. The spy allegation came only on 29 October, in the New York Times.[3][4]
The entire news cycle, with slow drips of new information, felt like a political playbook. If you run for office and you have dirt on your opponent, you release it slowly for maximum impact, to stack the pressure and force them to resign. The board also moved very quickly to accept his exit, as if it had no intention of keeping him: 13 days passed from the first spy report to the board accepting his resignation on 11 November, and only 3 days after Huang admitted the affair in a memo to staff.[5]
The Drip Conspiracy
Here is my conspiracy. The past few weeks saw an almost coordinated chain of events that all seemed to be anti-AI. On 29 October Reuters reported debt-financing concerns at the neoclouds CoreWeave and Firmus. On 3 November came the midterm elections, with data-center moratoriums in campaign platforms from the Texas governor-elect to Bernie Sanders. On 5 November Nvidia fell 9.26% and lost about $447 billion of market value in one day by our math. AMD fell even harder, 17.26%, and Micron 17.46%: that day hit all of AI, not just Nvidia.[6][1]
Then the drip, day by day. On 29 October the New York Times reported that US intelligence had concluded Alice Huang, a Chinese citizen and former television reporter who was Megaspeed's managing director in its first months, had been recruited by China's Ministry of State Security and cultivated Jensen Huang. Megaspeed is a Singapore cloud partner of Nvidia whose Malaysian sister company bought at least $4.8 billion of Nvidia hardware from June 2024 to February 2026. The sources "could not explain what information" she might have passed, and no one has been charged. Since then, some intelligence officials have called the spy label "not entirely accurate and over-torqued", according to CNN on 12 November, while the Justice Department says it has not reached a conclusion.[4][7]
Four more New York Times follow-ups came between 5 and 8 November, among them the spring memo on Alice Huang and the Commerce Department's approval of a Chinese firm, New Matrix, to assemble chips for Nvidia in Beijing. On 8 November Huang admitted the affair in a memo to staff, and the board opened an independent investigation. On 10 November The Information reported that a network around Alice Huang allegedly moved about 240,000 chips worth $7.5 billion into China; Nvidia fell 5.47% that day. On 11 November the New York Times reported a second affair, with Kimberly Powell, a vice president who reported to Huang, and the board accepted his resignation the same day. On 12 November Reuters reported a third relationship, with another, unnamed employee.[4][8][5]
He worked with a crisis PR firm a year earlier, so something may have been coming. According to WIRED, two unnamed sources say Huang had worked with the crisis firm HELLER since at least October 2025, with a reported scope covering his personal life and family philanthropy. Who hired the firm, and why, is not known.[9]
I owe my family and all of you an apology. I was unfaithful to my wife. My actions hurt her and our children. There is no excuse for what I did. I am deeply sorry. ... You deserve better from me.
Jensen Huang, memo to Nvidia staff, 8 November 2026When he apologized and announced he was unfaithful, I knew he had made a mistake. He should not have commented on his personal life at all; what he does in private is no one's business. He should have said that the connection of that woman with potential government actors would be investigated, with no comment on the affair. Only once it involved an employee did he need to address it, because a relationship with a direct report is a policy violation.
The Culprit
Why did this seem coordinated, and who benefits? China does not look like the beneficiary. Nvidia's China business had already shrunk: revenue from customers headquartered in China fell from 19.2% of the total in fiscal 2025 to 9.1% in fiscal 2026, 10.1 points in one year. Most of that drop came from the US ban on Nvidia's H20 chip in April 2025, which cost about $10.5 billion of revenue in two quarters, but China also limited purchases from its side: regulators summoned Nvidia over alleged "backdoors", told companies to stop ordering the RTX Pro 6000D and told publicly owned data centers to source more than half of their chips at home. Huang himself put it bluntly: "We went from 95% market share to 0%." China-headquartered revenue did rebound to $7.88 billion in the latest quarter, but Nvidia explains less than 11.3% of it.[2][10]
Taking out Huang would have been irrelevant for China, and China could not have mobilised Nvidia's board anyway. The initial leak and escalation seemed to come from the US government side. Classified US intelligence is the basis of the story, and according to the New York Times, Secretary of State Marco Rubio oversaw a memo on it. It fits a pattern: the crash deepened with a US anti-AI fear after the Democrat sweep and the resulting uncertainty over data centers. Anti-AI and anti-data-center sentiment had been growing for months and has become a core political issue.[4]
The question
Is it a coincidence that the neoclouds faced financing concerns, that the market crashed, and that Jensen Huang, the CEO most responsible for the AI buildout, was then personally targeted? To me this looked like a deliberate effort to use governance and public pressure to slow Nvidia down in financing and enabling the data-center buildout.
The next big question is whether the new CEO will be more conservative about pushing the buildout. I expect the next CEO to take it much slower, pull guarantees and underwriting from the neoclouds, and kick the can down the road. If that happens, this was an effort to slow AI from parts of the US government; no one else would have that reach.
Donald Trump does not seem to be involved. He is strongly pro-Huang, and his posts seem oblivious. After the spy report he said: "He's done an amazing job for our country." After the resignation he posted on Truth Social: "There is only one man to take his place - TIM COOK of Apple." No report shows Cook is a candidate; the board leaves the search open to internal and external candidates.[4][11]
This could be the growing fracture in the Republican Party, or a classic Democrat against Republican fight. The first committee action came from three Republican House chairs, Brian Mast (Foreign Affairs), John Moolenaar (China Select Committee) and Rick Crawford (Intelligence), who asked Huang to a closed-door briefing in a letter dated 4 November. The Democrat Elizabeth Warren had already asked for answers on 29 October, the day of the report. Since the resignation the chairs have asked for a transcribed interview on 1 December, Huang has agreed to testify, and the Senate Banking Committee wants him and Commerce Secretary Howard Lutnick in early December.[4][12]
The Risk of Less Control
When I started researching, I immediately checked how much control Huang had. To my surprise he had fewer shares than I thought and basically no voting control. He owned 3.58% of Nvidia, 870.6 million shares worth $144.7 billion at the 12 November close. BlackRock (7.43%) and Vanguard (7.31%) each own about twice as much. Nvidia has one vote per share, so his 3.58% of the stock was also 3.58% of the votes.[13]
Everything hinged on him. He ran Nvidia for 33 years and no succession plan was ever disclosed; asked how long he would stay, he said "for as long as I deserve it". About 60 managers reported directly to him, by his own count, or 36 in an internal chart from October 2025. The relationship with TSMC, which makes Nvidia's chips, has no contract and rests on trust, according to Huang, although Nvidia's filings show $279 billion of supply and capacity commitments in writing. Now the interim CEO, Colette Kress, holds two jobs: CEO and CFO. There is no way one person can do the CEO job well while also being CFO. The ten remaining directors could accept his exit without any vote he controlled.[13][14][5]
The Travis Kalanick and Elon Musk Parallel
Elon Musk was removed from PayPal. He learned the lesson to always have voting control, which is why he now exercises so much control over his companies and makes sure his stake increases over time. Travis Kalanick was likewise ousted from Uber by its investors and board. He held about a third of Uber's super-voting shares, with 10 votes each, but that did not save him, and the board ended super-voting in the SoftBank deal in October 2017.[15]
We can say what we want about Evan Spiegel at Snap, but he has full control over his business, and that is essential if you want to take risks. Mark Zuckerberg similarly has more than enough control to pivot into the metaverse without resistance or the risk of being removed. Huang made a huge mistake by betting only on his charisma. The quick removal was the exact consequence: he left because he knew he had no leverage to resist.
We lost a great co-founder. Musk came back and founded great companies after being ousted at 29. Kalanick came back after being ousted at 40. Huang is 63 now, 34 years older than Musk and 23 years older than Kalanick were at their ousters. Can he come back too? I return to that at the end.
The Cause of the Crash
This was two crashes in one. From the record on 6 October to 6 November Nvidia fell 22%, but that was an overall market crash: the iShares semiconductor ETF (SOXX) fell 30.5% in the same weeks, AMD 30.9% and Broadcom 21.9%. Before the first spy report, from 6 to 28 October, Nvidia fell 10% against 11.5% for chip stocks. In the week of the spy report, from 28 October to 4 November, Nvidia fell 7.1% while chip stocks fell 13.7%: the market ignored the spy story.[16]
The Huang week, from 6 to 12 November, was all Nvidia: the stock fell 11% while chip stocks rose 13.5%. On 9 November, the first trading day after Huang's memo, AMD closed 13.3% higher and Broadcom 8.8% higher. Over the week Nvidia did 21.5% worse than chip stocks. Had it moved with them, it would have closed at $211.84 instead of $166.19.[16]
Market or Nvidia?
Until 6 November the fall was sector-wide and Nvidia did no worse than chips. The Huang week alone cost it 21.5% against the sector, about $1.1 trillion of market value at 24.1 billion shares by our math.
Nvidia Business
Nvidia's revenue rose 1,195% in five years, from $16.7 billion in fiscal 2021 to $215.9 billion in fiscal 2026. Data Center was 89.7% of it, $193.7 billion, against only 40.2% in fiscal 2021. Networking, the switches, network cards and cables that link the chips into one system, made up 16.2% of Data Center, $31.4 billion, more than ten times its fiscal 2021 level. Gaming was 7.4% ($16 billion, a record), Professional Visualization 1.5% ($3.19 billion) and Automotive 1.1% ($2.35 billion), up 338% since fiscal 2021.[2]
Networking's share of Data Center has moved with each generation: about 25% in fiscal 2021 after the Mellanox purchase, a low of 8.4% in the quarter to January 2025 as Blackwell GPUs ramped, and back to 19.7% in the first quarter of fiscal 2027.[2]
Since this fiscal year Nvidia reports Hyperscale, ACIE (AI clouds, industrial and enterprise) and Edge instead. In the latest quarter revenue was $96.2 billion: 50.6% Hyperscale ($48.7 billion), 41.9% ACIE ($40.3 billion) and 7.5% Edge ($7.2 billion). Nvidia's own "physical AI" figure, over $9 billion in twelve months, is about 3.6% of revenue by our math; it has not published a definition. And in fiscal 2022, 2% of revenue came from dedicated crypto mining chips: $550 million.[2]
The Company
Nvidia became the largest company in the world but acts like a startup. It pivoted into crypto quickly, and out again. It sold dedicated mining chips during the boom, and throttled mining on over 80% of its own Ampere gaming cards so that gamers could buy them. When Ethereum stopped mining, Nvidia left the market.[17]
You might think that with about 42,000 employees it would need many factories and hardware facilities. It has no chip factory at all; TSMC makes its chips. It earned $5.14 million of revenue per employee in fiscal 2026, almost six times the $0.88 million of fiscal 2021, and it has committed $279 billion to suppliers, 129% of fiscal 2026 revenue. In fact, 73.8% of staff work in research and development. The business is not manufacturing chips but designing them, which makes being at the cutting edge its unique selling point. Every year it designs a new chip and scales it for use in data centers. The schedule is intense, which is why it has so many research staff.[18]
One of the most impressive facts about Huang's Nvidia: no major layoffs since 2008, when about 360 jobs, 6.5% of staff, were cut. Staff grew 121%, from 18,975 in fiscal 2021 to about 42,000 in fiscal 2026, and were never cut, even at that size. Turnover was 3.7% in fiscal 2026; in fiscal 2024 it was 2.7%, 6.6 times lower than the 17.7% industry rate Nvidia cited that year. That is rare at this scale, where many big tech companies cut 10% to 20% of staff every few years.[18]
The real test came in 2022, when Gaming revenue fell 56.5% in two quarters, from $3.62 billion to $1.57 billion, as the crypto boom ended. Huang's answer to staff:
Do we have a layoff? No. Instead, we have given raises.
Jensen Huang to Nvidia staff, August 2022These are the words of a good man.
What's Going Well
The business itself has never been stronger. Revenue rose 106% to $96.2 billion in the latest quarter, at a 75% gross margin and a 66.2% operating margin. Nvidia beat its own revenue guidance in 25 of the last 26 quarters, by 5.3% on average; the only miss was the quarter to July 2022. It guides $108 billion for the quarter reported on 17 November and about 70% revenue growth for next fiscal year. Free cash flow was $127 billion over the last four quarters.[2][19]
Signs of Trouble
Memory prices are eating the margin. Nvidia guided its gross margin 3 to 4 points lower for the January quarter, from 75% to 71% to 72%, and each point is about $1.1 billion of gross profit per quarter. The memory makers are now the ones with pricing power: Micron's gross margin reached 86.8%, 11.8 points above Nvidia's.[19][20]
The cash is lagging the profit. Receivables rose 55% in one quarter, to $63.1 billion, while revenue grew only 18%, and five customers owe 70% of the total. Operating cash flow was $24.1 billion in the latest quarter, while net income was $59.7 billion, about 2.5 times more. The $7.88 billion of China revenue in the quarter is less than 11.3% explained by Nvidia. Investment gains made up 13.3% of earnings per share over the last four quarters, and total debt rose 294% in one quarter, from $8.5 billion to $33.4 billion.[2][10]
The Buildout
Huang was all about aggressive expansion and dealmaking. Nvidia put $58.9 billion into other companies' shares in four quarters, 19.4% of revenue. The biggest stakes are in AI labs: $30 billion in OpenAI, up to $10 billion in Anthropic and $2 billion in xAI. It guaranteed up to $105 billion of OpenAI's data-center leases and backstopped up to $6.3 billion of CoreWeave capacity. In total, guarantees and cloud capacity deals reach $164.5 billion, 54% of a year's revenue, and in June 2026 Nvidia borrowed $25 billion in bonds.[21][2]
It looks like related-party transactions, but on paper it is not. Nvidia owns 11.5% of CoreWeave, and all the GPUs CoreWeave uses today are Nvidia GPUs, according to CoreWeave. Yet neither company reports the other as a related party.[21]
The reason was simple: Nvidia's biggest customers needed more chips, and Nvidia knew that the more data centers there are, the more Nvidia chips will be sold. Custom chips such as OpenAI's Jalapeño add supply, but they are not even cheaper than Nvidia. Jalapeño, built with Broadcom, has about the same cost per token as Nvidia's Vera Rubin, according to SemiAnalysis. OpenAI plans 1.3 gigawatts of Jalapeño in 2027, against about 12 gigawatts of Nvidia commitments spread over several years, and over 5 gigawatts of its own chips in 2028, according to Broadcom. Despite the massive buildout, the industry fears compute scarcity, which is why Huang was so aggressive.[22]
Why it made sense
Every new data center is a buyer of Nvidia chips. Customers feared running out of compute, and the stakes and guarantees made sure they could keep building.
Why it is a risk
The commitments grow faster than the revenue: $279 billion to suppliers and $164.5 billion of guarantees and capacity deals only bite if demand falls, but then they bite at once.
Competition
Nvidia is in a league of its own. It holds the top spot in AI chips and gaming. In the latest quarter Nvidia's Data Center had 77.9% of the combined AI revenue of Nvidia, Broadcom and AMD, and 84.1% over the last full year. Morgan Stanley estimates Nvidia has booked about 60% of TSMC's advanced packaging (CoWoS) capacity, and Jon Peddie Research puts it at 94% of add-in graphics boards.[23]
Broadcom is the shadow competitor. It designs custom AI chips for six customers: Google, Anthropic, OpenAI, Meta and two more. Buyers choose between Nvidia and Google's TPU, never between Nvidia and Broadcom, yet the money flows to Broadcom. Its share rose to 14.6% in the latest quarter, 5.9 points up from 8.7% over its last full year, almost all of Nvidia's 6.2-point loss. Broadcom's AI revenue was $16.7 billion in its latest quarter, while Nvidia's Data Center was more than five times larger, but Broadcom guides $230 billion of AI revenue for fiscal 2028. AMD's Data Center made $8.5 billion in its latest quarter; Nvidia's was about ten times larger. On the day Huang's resignation was announced, Broadcom rose 5.5%.[23][1]
What Was the Strategy Under Jensen
Creating the demand was the main strategy. Nvidia expanded into open models and physical AI platforms, and agreed to pay up to $12.9 billion for Hugging Face, the largest hub for open AI models, about 86 times its yearly revenue. CUDA, Nvidia's free software for training and running AI on its chips, grew from 2.2 million developers in fiscal 2021 to 7.5 million in fiscal 2026, up 241%. Nvidia gives away its own AI models to sell compute: Nemotron for language and GR00T for robots, while Omniverse and Isaac simulate the real world to train robots and cars.[24]
Huang did not just want Nvidia to be the chip in data centers. He wanted Nvidia to be synonymous with AI, from language models to physical AI systems: the underlying platform to train and use AI in the world. It even let non-Nvidia systems connect, which improves interoperability and cements Nvidia's place: NVLink Fusion lets customers plug their own chips into Nvidia systems. Nvidia invested $2 billion, about 2.4%, in the chip designer Marvell to build those custom chips. So it invested in its competitor's competitor (Broadcom, the shadow competitor): a hedge against the custom chip trend.[23]
What Is the Strategy Now?
Whoever initiated this slowed the speed of the AI buildout. Nvidia will probably be far less aggressive once Huang's momentum wears off, and the expansion into physical AI and other systems is unlikely to be as pronounced. Colette Kress, CFO since 2013, is interim CEO. Harvey Jones, a director since 1993, is Nvidia's first-ever chair. The board is searching for a permanent CEO inside and outside the company. Kress leads the 17 November earnings call, the first without Huang.[5]
What to watch on 17 November
The $108 billion guide (consensus about $109 billion), the fourth-quarter and fiscal 2028 outlook, receivables and cash flow, and any customer comment on the leadership change. At Nvidia's average beat of 5.3%, revenue would be about $113.7 billion by our math.
Outlook
A comeback for Huang? There is a small chance Nvidia takes him back, but I put it below 15%. Some employees want him back: an internal post asking the board to reconsider drew over 1,600 comments, mostly in favour, according to WIRED as reported by others. That is comments, not employees, and at most 3.8% of about 42,000 staff.[25]
He is 63 but still intense and full of energy. I don't think he would start a business from scratch, although I would love to see it. I also can't see him join another business like SpaceX or Tesla and help build there; he is too much of a leader. I would not be surprised if he went to Taiwan, where he is much more loved than in the US. Morris Chang offered him the TSMC CEO job in 2013, according to Chang's memoir, and he declined. He could form a new company and partner with TSMC on a new business. The trust he had with TSMC was exactly what was missing with Nvidia's board.[14]
He might take the Travis Kalanick route, where each next business is a level up from the last. Uber Eats was one of Kalanick's last additions at Uber; then he bought CloudKitchens, and later formed Atoms, which automates the physical world. Atoms launched as a robotics company for food, mining and transport on 13 March 2026, almost nine years after his ouster from Uber, and reportedly raised $1.7 billion in July 2026, in a round led by Andreessen Horowitz with Uber investing.[26]
Maybe Huang will go where Nvidia couldn't yet: physical AI, the training platform and advanced chips for robotics. In Taiwan, with TSMC, he could start building fast, and he would have no trouble raising money or using his own: his Nvidia shares alone are worth $144.7 billion. Nvidia's executives have no employment agreements, and California voids non-competes, although trade-secret law still applies and Nvidia's platform code stays Nvidia's.[13]
Good
Revenue up 106% at a 66.2% operating margin, 25 of 26 quarters above its own guide, $127 billion of free cash flow, a roadmap and supply locked in through 2028, 7.5 million CUDA developers, and a P/E of 21 trailing and 10.6 forward, lower than at the bottom of 2022.
Bad
The founder gone with no successor, an interim CEO in two jobs, memory squeezing the margin, receivables up 55% in a quarter, $279 billion of supplier commitments, $164.5 billion of guarantees and capacity deals, and Broadcom's custom chips taking share.
Sources and chart methodology
- Yahoo Finance and TradingView daily closes for NVDA, January 2020 to 12 November 2026, split-adjusted and not dividend-adjusted, plus TradingView quotes and market values for Nvidia, Apple, Microsoft, Alphabet, Amazon, Broadcom and AMD on 12 November 2026. Falls are close to close from a running high. Market value changes use about 24.1 billion shares. P/E ratios use Nvidia's diluted EPS over the last four reported quarters and the fiscal 2028 consensus of $15.68 on TradingView.
- Nvidia SEC filings: the FY2021 to FY2026 10-K filings, the Q1 and Q2 FY2027 10-Q filings and the quarterly earnings releases and CFO commentaries. Free cash flow is operating cash flow minus capital expenditure. Shares of revenue, networking shares (Q2 FY2027 estimated), the 2021 and 2022 P/E ratios and the share of China revenue explained by Nvidia are our math.
- Hedge Fund Alpha on Culper Research's short report of 13 May 2026 and Culper's X thread. Culper disclosed a short position.
- The New York Times reports of 29 October and 5 to 8 November 2026, read through CNBC, NPR, ABC, CBS, Implicator and Let's Data Science; the originals were behind a paywall. The intelligence allegations are anonymously sourced and unproven; no one has been charged.
- Nvidia 8-K of 12 November 2026 and press release: resignation accepted on 11 November, Colette Kress interim CEO and CFO, Harvey C. Jones first chair, search for internal and external candidates. No reason is given.
- Reuters on debt-financing concerns at CoreWeave and Firmus, 29 October 2026; Invezz and Gotrade on the 5 November AI sell-off and data-center moratorium platforms. Our close-to-close math gives -$447 billion; Dow Jones Market Data gave about $381 billion at the intraday price.
- CNN, 12 November 2026, on doubts inside US intelligence, and AsiaOne on the Justice Department and Singapore.
- The Information, 10 November 2026, on the alleged network around Alice Huang; the NYT, 11 November, on Kimberly Powell, via Axios; Reuters, 12 November, via US News.
- WIRED, 9 November 2026, via Gadget Review and Everything PR; the WIRED original was not read.
- Nvidia 10-K and 10-Q geographic revenue by customer headquarters; China export rules and charges from Nvidia's 8-K filings and releases; Tom's Hardware on Huang's "95% to 0%".
- Donald Trump on Truth Social, 12 November 2026, as reported by TechCrunch and MacRumors.
- RedState on the House chairs' letter of 4 November; TechCrunch, Daily Caller and CBS News on the interview and the Senate Banking request.
- Nvidia 2026 proxy statement (870.6 million shares, 3.58%, one vote per share, no employment or severance agreements for executive officers); TradingView ownership data; Yahoo Finance on "for as long as I deserve it"; California Business and Professions Code section 16600 on non-competes.
- Fortune on Huang's direct reports and Business Insider's October 2025 chart via TipRanks; Huang on the TSMC relationship on the Lex Fridman podcast via Benzinga; Morris Chang's memoir via PC Gamer.
- On Uber's dual-class shares and the SoftBank deal: Quartz, The Irish Times and AICD.
- TradingView daily closes for NVDA, AMD, AVGO, the iShares Semiconductor ETF (SOXX) and the Nasdaq 100 ETF (QQQ), 6 October to 12 November 2026; our math.
- Nvidia 10-K and earnings releases on CMP mining chips and Lite Hash Rate GeForce cards; SEC settlement over crypto disclosures, May 2022.
- Nvidia 10-K human capital sections, fiscal 2021 to 2026 (headcount, R&D share, turnover and the industry comparison); the 8-K of 18 September 2008 on the last layoff; the August 2022 memo via Tom's Hardware.
- Nvidia's Q2 FY2027 earnings release, CFO commentary and call (guidance, gross-margin outlook, fiscal 2028 growth); beats against the midpoint of guidance are our math from each release.
- Micron's latest quarterly report (GAAP gross margin 86.8%); Broadcom and Marvell latest quarterly reports.
- Nvidia 10-Q and 8-K filings (equity purchases, guarantees, the SB Energy guarantee of 17 August 2026, the CoreWeave backstop, the $25 billion bond issue) and 13F reports; CoreWeave 10-Q on its GPUs and related parties. The $164.5 billion combines $108.5 billion of guarantees, $36 billion of AI cloud agreements and $20 billion of leases to be reassigned.
- SemiAnalysis on Jalapeño's cost per token, reported by The Decoder and Techzine; OpenAI's Nvidia commitments from Nvidia's Q2 FY2027 call; OpenAI's 2027 and 2028 plans from Broadcom's Q3 FY2026 call.
- Broadcom's Q3 FY2026 release and call, AMD's Q3 2026 release, Marvell's filings (Nvidia's $2 billion convertible stake, about 2.4% by our math); Morgan Stanley's CoWoS estimate via SemiMedia; Jon Peddie Research via Club386. AI revenue shares compare Nvidia's Data Center, Broadcom's AI semiconductors and AMD's Data Center; our math.
- Nvidia 10-K filings (CUDA developers) and the Q4 FY2025 call; Nvidia's 8-K of 2 September 2026 on Hugging Face, with revenue multiple from Fortune.
- WIRED on the internal post, via Windows Report and Wccftech; the WIRED article itself was not found.
- Atoms: The Next Web, Bloomberg Law and Forbes Italia on the $1.7 billion round.