To The Moon

$NVIDIA(NVDA)$  

Mathematical Money | November 12, 2026

In the three sessions to 11 November, $AMD$ rose 23.1% and $AVGO$ rose 17.5%. $NVDA$ fell 6.9%, to $173.73. Same customers, same chips, same week.

The difference is one man. On Sunday 8 November Nvidia's board said it was opening an independent investigation, and that it "has confidence in Jensen". By Wednesday evening the Wall Street Journal was reporting that the same board was preparing to remove him. That is a report, the allegations behind it are unproven, and the decision was still pending when I wrote this.

I have no position in Nvidia. What I have is a five-minute check that puts a number on a question like this, and four older cases to run it on. Nothing below depends on which way the board goes.

The check: find a control group

When a stock you follow drops, the first question is whether the market did it or the company did. Most of us answer by feel. You can answer it with two other tickers and a calculator.

Step one. Pick the two or three listed companies that sell to the same customers. For Nvidia that is AMD and Broadcom, who sell AI chips to the same handful of cloud giants.

Step two. Pick the day the trouble started for everybody, and measure all of them from that close. Step three. Find the day the lines split apart, and look up what was announced that day. Step four. Turn the gap into dollars, so you know what the market is charging for the company's own problem.

I call it the control-group check. A drug trial needs people who never took the drug. A selloff needs stocks that never got the news.

A month together, three days alone

Nvidia made its record close on 6 October. From there to the 6 November close it fell 22.0%. Broadcom fell 21.9%. AMD fell 30.9%.

That month held a hot inflation print, a 12% one-day fall in the chip index and the first headlines about the chief executive. Through all of it Nvidia was not being singled out at all. It fell with its rivals, and by less than AMD.

Then the dates line up almost too neatly. Sunday 8 November, the board statement. Monday, Nvidia down 4.3% while AMD jumped 13.3%. Tuesday, the independent directors hired outside lawyers and Nvidia lost another 5.5%. So the whole Nvidia-only discount is three trading days old, and each leg down followed something the board said.

The price on one person

Across the full five weeks, AMD and Broadcom are down 11.5% on average. Had Nvidia done the same it would sit near $212. It closed at $173.73.

That gap is about $38 a share, on roughly 24 billion shares. Call it $900 billion. You can do the same sum on any stock in your own account, and it is worth doing before you form a view.

Now set it against something solid. AMD, the entire company, is valued at about $907 billion. Huang's own stake, 3.58% per the last proxy, is worth about $150 billion. The market is charging six times the founder's whole shareholding for the chance that he goes.

One more comparison. Analysts quoted this week put the cost of a removal at about 20%. The control group says roughly 18% is gone already. If they are right, most of the bad outcome is in the price before anyone has voted.

Six votes, and what nobody outside the room knows

Who decides is not a mystery. Nvidia has eleven directors and ten are independent. Under the bylaws on file, officers serve at the pleasure of the board and can be removed "with or without cause" by a majority of directors in office. That is six votes. There is no chairman, only a lead independent director, and a 3.58% stake cannot block it.

Everything else is a gap, and it is worth listing what the company has not said. The 8 November statement covers reports that Huang "has been the target of a foreign intelligence effort". It says the board has "no reason to believe" national security or company secrets were compromised. It gives no deadline. It does not promise to publish what it finds.

And there is no named successor. The proxy talks about "management development" and names nobody. The names in the press, the finance chief among them, are reporters' sources and analysts' guesses.

So be careful who you believe this week. Prediction markets showed better than 80% odds that he is out by year-end, on thin volume. One analyst has published two price targets, one for each outcome. None of them has seen the investigation. Neither have I.

What happened the last four times

I ran the same check on four older cases where the person most identified with a company had his seat in play. Same rule each time: the stock against its closest peers, measured from the last close before the story.

Wynn Resorts, 2018. A newspaper report on 26 January, and the founder resigned on 6 February. By then the stock sat 15 points behind Las Vegas Sands and MGM. The next session it jumped 8.6%.

Tesla, 2018. The regulator sued to bar Elon Musk from running the company. The stock lost 13.9% in a day. He settled over the weekend and kept the job, and it rose 17.3% on the Monday.

Apple, 1985. Steve Jobs, chairman rather than chief executive, lost his operating role in May and resigned in September. Apple was about 12 points behind IBM and the index by then. It rose 6.6% within two sessions of his going.

Papa John's, 2018. The founder resigned as chairman overnight on 11 July. The stock rose 11.0% the next day.

Four cases, four jumps, and it did not matter whether the founder stayed or went. The market paid for the answer, not for the person. If you have ever waited on a medical result and felt better the moment you had it, even before you knew what to do about it, that is the trade.

A year later is a different story, and an honest one. Apple was 83 points ahead of its peers. The other three were behind, by 19 to 30 points. Four cases is an anecdote, not a base rate, and the jump on the answer is the only thing they share.

After it, each stock went back to being about its business. Wynn's runs on gaming licences. Papa John's carried the founder's name on the box. Where a regulator or a customer cared who the founder was, the answer did not end it.

Which kind is Nvidia?

Customers first. In August the finance chief guided next fiscal year's revenue up about 70%, and said customers were asking for roughly double. If they want 200 and the factories can supply 170, they have to cancel 15% of their requests before one guided dollar is lost. Nobody buys these chips for the name on the box.

The hole in that, stated plainly: those requests are forecasts and not contracts. The last quarterly filing shows only $3.2 billion of contracted revenue running beyond one year.

Regulators are the Wynn problem. Every sale to China needs an export licence, and China was under 1% of data-centre revenue last quarter. Three House committee chairs have asked Huang for a closed-door briefing. The investigation is about foreign intelligence, which is exactly the subject those licences turn on.

The third channel is financing. Nvidia has put nearly $50 billion into AI labs and guaranteed up to $108.5 billion of data-centre debt. The finance chief said the labs it supports will be about a quarter of the business next year. Those are relationships as much as contracts.

What Tuesday can and cannot tell you

Results land on Tuesday 17 November, after the close. The quarter being reported ended on 25 October. The first report about Huang ran on 29 October.

So nothing in Tuesday's numbers can contain the thing the stock is being sold for. Management guided revenue to $108.0 billion, give or take 2%. The same quarter a year ago was $57.0 billion. Street estimates run from about $108.8 billion to $111.4 billion. If you are holding it into the print, a beat on any of that is a statement about October.

For scale, last quarter's revenue was $96.2 billion, more than double the year before. Data-centre revenue alone was $89.0 billion. Broadcom's AI chips did $16.7 billion and AMD's whole data-centre arm did $8.5 billion. The rivals are small and growing faster, and AMD expects revenue up more than 80% in 2027.

The roadmap does not need a signature either. Vera Rubin, the next platform, began shipping in August. Coverage of the March keynote has Rubin Ultra in 2027 and Feynman in 2028. Management says a gigawatt of Rubin brings in about $40 billion, against $25 billion for the current generation, and that is where the operating leverage comes from. The pond is still filling: Nvidia's finance chief put the five biggest cloud spenders at nearly $800 billion this year and $1.3 trillion next.

The other risks, briskly. Five customers owe 70% of the money due to Nvidia. Margins are guided to bottom at 71% to 72% this winter on memory costs. Supply commitments stand at $279 billion, a strength while demand holds and a weight if it does not. And there is the threat of local data-centre moratoria after the midterms.

My own numbers, and why I don't trust them this week

My research note on Nvidia is dated 10 November. The good case is $277, at 25% odds. The middle case is $193, at 50%. The bad case is $112, at 25%.

Weighted from the 11 November close that comes to +11.5% over twelve months. Two days earlier the same note said +8.4%. Nothing improved. The price fell and the scenarios stood still.

An expected return that rises because the price fell on news the model has not read is arithmetic, not a signal. None of my three cases includes a change of chief executive, and the note calls the decline "not a company event". That was true on 6 November, and the check says it stopped being true on the 9th. If you keep a target on anything you own, put the date of the thesis next to the date of the news.

Worked backwards, the price assumes about 28% growth a year for ten years, against the 70% guided for next year. That looks cheap on the old facts. The new fact is not in the sum.

Where I stand

No position, and no hurry. If the board keeps him, the older cases say the gap closes fast and the argument moves to Washington. If the board removes him, they say the stock may well jump anyway, and then the argument moves to Washington. Both roads lead to the same place, which is the export licences.

So that is what would move me. Not the vote. Whether the investigation's findings are published, whether the 70% guide is repeated on Tuesday, and whether anyone in Washington starts talking about the licences.

Here is what I cannot get comfortable with. The market has priced one person at six times his own stake, and the people who decide need only six votes and owe nobody a deadline. I can measure the discount to the dollar and I cannot tell you a single thing about what closes it. If you have followed a board investigation at a company you owned, how long did it take before you were told anything at all?

Prices for Nvidia, AMD and Broadcom from Tiger daily bars to the 11 November close. Share counts and the four older cases from yfinance, adjusted closes. Company figures from Nvidia's Q2 FY2027 release, 10-Q, 26 August call, 2026 proxy and bylaws, AMD's 3 November results and Broadcom's 3 September results. Board statements dated 8 and 10 November; the removal report is the Wall Street Journal's, 11 November, as carried by Reuters and CNBC. The gap, the 15% cushion and the rebased return are my own arithmetic.

$NVDA$ $AMD$ $AVGO$

#NVDA #semiconductors #AI #investing #stockmarket

Stop guessing. Start calculating.

Live to fight another day. 🤙