Is Nvidia stock a buy without Jensen Huang? At $155.46 we think it is, bought in stages.
Nvidia trades at about half the S&P 500's multiple of expected earnings
AMD28.5x
Marvell20.2x
S&P 50017.7x
Broadcom17.6x
TSMC15.4x
Nvidia9.8x
Forward P/E: the November 17, 2026 closing price divided by the average analyst estimate of earnings per share for the next fiscal year. For Nvidia and Marvell that year ends in January 2028, for Broadcom in October 2027, and for AMD and TSMC in December 2027. Estimates were read on November 17 and had not yet changed for Nvidia's results.
Source: Yield Theory calculation from Yahoo Finance closes and consensus estimates; FactSet estimate for the S&P 500
Nvidia closed on November 17 at 9.8 times the profit analysts expect it to make next fiscal year. That ratio is the forward price-to-earnings ratio, or P/E: the share price divided by expected earnings per share. The S&P 500 trades at 17.7 times. Broadcom, which designs custom AI chips for the big cloud companies, trades at 17.6 times and AMD at 28.5.
Six weeks ago the same shares cost 15.1 times those earnings. Since then Nvidia has lost about $2 trillion of market value and the founder who ran it for 33 years.
What happened. On November 8 Jensen Huang, who had run Nvidia since 1993, told employees in an email that he had been unfaithful to his wife. The same day the board said it was investigating "recent reporting that founder and CEO Jensen Huang has been the target of a foreign intelligence effort," and that it had "no reason to believe that either national security or NVIDIA secrets were compromised." On November 12 Nvidia said he had resigned. It gave no reason that day.
Press reports, starting with The New York Times on October 29, allege a relationship with a former executive of Megaspeed, a buyer of Nvidia chips, whom US officials are reported to suspect of ties to Chinese intelligence. China's government has denied that, and no charge against Huang or Nvidia has been made public. On November 17 Huang said he had "made serious mistakes in my personal life" and would be "honored" to return as CEO if the board asked. Hours later Nvidia reported the largest quarter in its history. On the earnings call its chair, Harvey Jones, said Huang had resigned "after disclosure of personal behavior inconsistent with the company's code of conduct, standards, and cultural norms," and that the board "does not intend to reinstate Jensen as CEO."
This issue asks what the stock is worth if he never comes back. Whether he will is a separate question, covered on our free page about Huang's statement and the board's position. After Jones spoke, traders on Kalshi, a prediction market, put the chance of Huang being Nvidia's next permanent CEO at about 7%, down from 12% an hour earlier. We wouldn't buy the stock for that chance, and our numbers don't rely on it.
Our findings:
As much as $36 of the fall is specific to Nvidia. That is the gap between Tuesday's close and where the stock would be if it had moved with other chip stocks since Huang's first apology. Across all the shares, it comes to $873 billion.
The price needs only $8.81 of earnings. If Nvidia deserved the same multiple as the S&P 500, $155.46 would be justified by profit of $8.81 a share. Nvidia is on course to make about $9.60 this fiscal year, and analysts expect $15.86 next year.
On profits already earned, the stock is the cheapest it has been since 2013. It came close to this level twice, in 2015 and 2018. We look at what happened next.
Past exits like this one hurt for about a year. In five earlier cases, the stock trailed the S&P 500 by a median of 18.5 percentage points over 12 months. Nvidia has trailed it by 16.2 points in seven trading days.
We put a value and a probability on three scenarios. The price sits below two of them.
We end with three calls: on Nvidia, on the rivals that gained $319 billion of market value while it fell, and on the smaller companies that depend on Nvidia's balance sheet.
Why is Nvidia stock down 35%?
Three different things have hit the stock since its record close of $239.24 on October 6. Only the last one is about Huang.
Of the $84 a share Nvidia has lost, $31 has come since Huang's apology
Record close, Oct 6$239.24
AI sell-off, Oct 6 to Nov 3−$32.55= $206.69
After the midterms, Nov 3 to Nov 6−$20.05= $186.64
Since the apology, Nov 6 to Nov 17−$31.18= $155.46
Close, Nov 17$155.46
Closing prices per share. The legs are split at election day (November 3) and at the last close before Huang's apology email (November 6). The labels name the main news in each period. Other news moved the stock too, including the first press report about Huang on October 29.Source: Yield Theory calculation from Yahoo Finance closes
Period
Nvidia
Chip stock index
S&P 500
October 6 to November 3
−13.6%
−19.8%
−3.2%
November 3 to November 6
−9.7%
−11.4%
−2.5%
November 6 to November 17
−16.7%
+2.7%
−0.5%
Whole period
−35.0%
−27.0%
−6.1%
The chip stock index is the PHLX Semiconductor index, which tracks 30 large US-listed chip companies.
October 6 to November 3: the whole AI trade sold off.Reuters put the fall in chip stocks down to "rising bets of an interest-rate hike and concerns over the cost of the AI buildout." Rates were the subject of last week's issue. Nvidia fell 13.6%, less than the 19.8% drop in the index.
November 3 to November 6: the midterms. Two days after the elections, on November 5, the chip index dropped 12.3% in a day. Reuters said chipmakers sank "on fears of tougher AI regulation" after Democratic wins in states where opposition to new data centers has grown. Issue 07 explains those votes. Over the three days Nvidia lost 9.7% and the index 11.4%.
November 6 to November 17: Huang. The apology email went out on Sunday, November 8, and the resignation followed on November 12. From the Friday close before the email to Tuesday's close, Nvidia fell 16.7%. The chip index rose 2.7%.
So until the apology, Nvidia was the safer chip stock. On November 6 it was 22.0% below its record, against 29.0% for the index. Everything it has lost relative to other chip stocks, it has lost in the seven trading days since.
How big is the Huang discount?
The stock has fallen $31.18 since November 6, the last close before the apology. Chip stocks rose 2.7% over those days. If Nvidia had moved with them, it would have closed on Tuesday at $191.66. It closed at $155.46. The gap is $36.20 a share.
That benchmark is generous to the gap. The chip index includes Nvidia's rivals, and some of their gains came from Nvidia's trouble. Measured against the S&P 500 instead, which slipped 0.5%, the gap is about $30 a share, or roughly $730 billion. We use the chip index because it strips out the chip-sector moves of those days, and we treat $36 as the top of the range.
With 24.1 billion shares outstanding, that is $873 billion, or 23% of what the company is worth today. We call it the Huang discount, though it bundles three worries: a company without its founder, the legal and political fallout from the reports about him, and new export rules published on November 16. We take each in turn below.
One way to judge whether $873 billion is too much is to ask what profit it stands for. At 12 times earnings, a multiple between Nvidia's own and the market's, $873 billion is worth about $73 billion of profit a year. Analysts expect Nvidia to make about $380 billion next year ($15.86 a share on roughly 24 billion shares). So the market is pricing the events since November 8 as if they will cost Nvidia about a fifth of its profit, for good.
That is a large number to put on a company that has already ordered the parts for next year's products, as the results show.
Is Nvidia stock cheap?
On next year's expected earnings it is the cheapest large chip designer by a wide margin, and it got that way in seven trading days.
Prices are Yahoo Finance closes. Estimates are Yahoo Finance's analyst consensus, read on November 17 before Nvidia reported, and the same estimates are used for both dates. The S&P 500 figure is FactSet's estimate of the index's earnings over the next 12 months.
The comparison is not exact. Nvidia's next fiscal year ends in January 2028, a month or more after the others', and the S&P 500 figure looks only 12 months ahead. But the gap survives a stricter test. On the roughly $9.60 a share Nvidia is on course to earn in the fiscal year that ends this January, it trades at 16.2 times, still below the market's 17.7.
Micron shows how the market treats profits it thinks are at a peak. Micron makes memory chips, whose prices swing hard. Its profits have jumped with AI demand, and investors pay only 3.4 times next year's expected earnings. In our reading, that is the market saying it doesn't believe the profits will last. At 9.8, Nvidia's multiple now sits between Micron's and Broadcom's.
Broadcom and AMD got more expensive as Nvidia got cheaper. Their forward P/E ratios rose by about 2.5 and 3 points in the same seven trading days. Investors didn't leave AI chips. They moved to the two companies they expect to gain most if Nvidia stumbles.
What profit does $155.46 assume?
Turn the P/E around. Pick the multiple you think Nvidia deserves, and the share price tells you how much profit it is counting on.
If Nvidia deserves the multiple of
P/E
Earnings per share the price needs
Compared with the $15.86 analysts expect
The S&P 500
17.65
$8.81
44% lower
Broadcom
17.60
$8.83
44% lower
TSMC
15.38
$10.11
36% lower
Nvidia itself on October 6
15.08
$10.31
35% lower
A discounted 12 times
12.00
$12.96
18% lower
Nvidia reported adjusted earnings of $6.77 a share for the first nine months of this fiscal year. Its own forecast for the current quarter works out to about $2.82 more, on our arithmetic: $130.0 billion of revenue, a 71.5% gross margin, $10.7 billion of operating costs, a 17% tax rate and about 24.2 billion shares. That makes roughly $9.60 for the year that ends in January 2027.
So at the market's multiple, today's price assumes Nvidia earns less next year than this year. Even at a discounted 12 times, it assumes earnings come in 18% below forecast. Analysts expect them to rise about 65%. Those forecasts had not yet moved for Tuesday's results when we read them, and some may not yet allow for Huang's exit, so treat $15.86 as a starting point. Our own cases, further down, run from $12.10 to $17.30.
How low is this against Nvidia's own history?
On earnings already reported, Nvidia made $9.31 a share over the last four quarters under standard accounting rules (GAAP). At $155.46 that is a trailing P/E of 16.7.
Nvidia's P/E on reported earnings is the lowest in 13 years
020406080100x20142016201720182020202220232024NowAfter a 56% fall16.716.7xTrailing P/E201419.7x201530.5x201655.9x201748.0x201817.8x201960.2x202085.5x202190.6x202262.1x202365.3x202452.9x202546.3xNow16.7x
Share price at each year end, and on November 17, 2026 for now, divided by GAAP diluted earnings per share for the four most recently reported quarters, adjusted for share splits.Source: Yield Theory calculation from SEC filings and Yahoo Finance closes
We ran that ratio for every trading day back to 2013, switching to new earnings on the day each quarterly report was filed. The last time it was below 16.7 was November 7, 2013. In the 13 years since, it has come close twice.
Date
Trailing P/E
What was happening
Stock 12 months later
Earnings 12 months later
July 27, 2015
17.2
Nvidia was a PC graphics company and the PC market was shrinking
+193%
About flat
December 24, 2018
17.0
Stock down 56% in 12 weeks after crypto miners stopped buying graphics cards
+88%
Down 48%
November 17, 2026
16.7
Founder resigned; investigation open
?
Analysts expect about +65%
The 2018 row is the useful one. The low P/E was right about profits: Nvidia's earnings nearly halved over the following year as it worked through unsold chips. And the stock still rose 88%, because by the time the P/E reached 17 the price had already allowed for the fall.
Two rows don't make a rule, and Nvidia is almost 50 times larger now than it was at that low. But the pattern argues against waiting for proof that earnings will hold. In 2018 the proof came a year after the bottom.
One caution on the trailing number. GAAP earnings include gains on Nvidia's stakes in other companies, which came to $25.5 billion before tax in the first nine months of this fiscal year. On Nvidia's adjusted measure, which leaves those gains out and counts stock pay as a cost, the last four quarters add up to $8.36 and the trailing P/E is 18.6.
What did Nvidia's first results without Huang show?
They showed a business that was still accelerating on October 25, when the quarter ended. That was two weeks before Huang's apology, so the numbers can't show what his exit costs. That will show up in the deals signed over the next two or three quarters. What the results do show is how much of next year is already committed.
The headline figures are on our free page: revenue of $115.8 billion, up 103% in a year, and a forecast of $130.0 billion for the current quarter, about $5 billion above what analysts expected. The table below has the lines further down the results release and the finance chief's written commentary that matter most for value.
Line
Latest quarter
Quarter before
Why it matters
Sales to the largest cloud companies
$60.1 billion
$48.7 billion
Up 23% in three months
Sales to other data center customers
$47.5 billion
$40.3 billion
Nearly three times the level of a year ago
Adjusted gross margin
74.9%
75.0%
Forecast to fall to 71% to 72% this quarter
Parts and factory capacity ordered ahead
$390 billion
$279 billion
$203 billion of it falls due next fiscal year
Money owed by customers
$81.0 billion, or 64 days of sales
60 days of sales
Some large customers were given longer to pay
Unsold stock
$39.7 billion
$31.6 billion
Built up for the new Rubin chips
Free cash flow
$43.8 billion
$21.3 billion
Helped by the timing of tax payments
1. Next year's chips are already being bought
Nvidia has promised its suppliers $390 billion for parts and factory capacity, $111 billion more than three months ago. It says most of the increase is for memory chips, bought years ahead. Of the total, $203 billion falls due in the fiscal year that ends in January 2028.
Nvidia has already ordered $203 billion of parts for next fiscal year
50100150200$250B0
$69B
$203B
$92B
$10B
$8B
$8B
Rest of fiscal 2027Fiscal 2028Fiscal 2029Fiscal 2030Fiscal 2031Later
Commitments to suppliers as of October 25, 2026, by the fiscal year in which they fall due. Fiscal 2028 ends in January 2028. The total is $390 billion.Source: Nvidia CFO commentary, third quarter of fiscal 2027
That number tells you what Nvidia's own planners expect. Nvidia keeps about 72 cents of each revenue dollar as gross profit, so parts cost about 28 cents. Analysts' revenue forecast for next year, $688.7 billion, needs about $193 billion of parts on that rough rule. Nvidia has already ordered more than that. The orders were placed by a management team that included Huang, but they are contracts, and they don't leave with him.
On the same rough rule, $203 billion of parts would support about $725 billion of revenue, 71% more than this year. That is close to what Nvidia now says it expects. On Tuesday's call it raised its outlook for next year's growth to about 75%, from the 70% it outlined in August.
The orders cut both ways. If demand holds, the supply is locked in. If it falls short, Nvidia still has to pay. In April 2025, when export rules stopped it selling one chip to China, it took a $4.5 billion charge for stock and orders it could no longer use. A shortfall against $203 billion would be far bigger.
2. Profit margins step down, as flagged
Gross margin, the share of revenue left after the cost of making the chips, was 74.9% on Nvidia's adjusted measure. The forecast for the current quarter is 71% to 72%. Colette Kress, the finance chief who is now also interim CEO, told investors in August that margins would hit their low point this quarter, Fortune reported. So this drop was in the plan before any of the news about Huang.
It still matters for value. On next year's expected revenue, each point of gross margin is worth about $6.9 billion of profit before tax, or roughly 24 cents a share. The move from 75% to 72% takes about 70 cents a share out of next year's earnings compared with holding margins flat. Analysts' $15.86 already allows for it, on our arithmetic.
3. Customers are taking longer to pay
Customers owed Nvidia $81.0 billion at the end of the quarter. That equals 64 days of sales, up from 60. Nvidia gives two reasons: more chips shipped late in the quarter, and "extended payment terms on large, multi-quarter agreements with certain investment-grade customers."
Four days is small, and investment-grade customers are the safest kind. But longer payment terms are a form of lending, and this line would show stress first if buyers were stretching to fund their orders. Issue 05 covers how Nvidia has used its balance sheet to help customers buy.
4. The cash is real, and a lot of it is coming back
Nvidia produced $113.7 billion of free cash flow in nine months. Free cash flow is the cash a business generates after paying for equipment. Over a year, that pace equals about 4% of Nvidia's market value.
It returned $28.1 billion to shareholders in the quarter through buybacks and dividends. It has $227.3 billion of buyback authority left, which it says it plans to use "through fiscal 2028." At Tuesday's close that would retire about 1.46 billion shares, 6% of the total. The lower the share price, the more shares that money retires.
5. The detailed quarterly report hasn't been filed
In each of the last six quarters, Nvidia filed its full quarterly or annual report with the SEC on the same day as its results. This time it didn't. The finance chief's commentary says: "We plan to file our 10-Q in the coming weeks."
The 10-Q is the detailed quarterly filing. It carries items the results release leaves out: revenue by country, how much comes from the biggest customers, legal proceedings and updated risk warnings. Those are the pages investors most want to read right now. The deadline for the quarter that ended on October 25 is December 4.
A filing that comes a little later than usual, during a board investigation and a change of CEO, is not alarming by itself. A missed deadline would be.
6. On the call: no return for Huang, and a higher outlook
The call opened with the chair, Harvey Jones, instead of the finance chief. He said the board "unanimously accepted his resignation and does not intend to reinstate Jensen as CEO," according to the transcript of his remarks published by The Verge. He said the board had hired the search firm Heidrick & Struggles to find a permanent CEO, and that it would take no questions on the resignation or "the ongoing investigation."
A higher outlook. She said Nvidia now expects revenue to grow about 75% in the fiscal year to January 2028, up from the 70% it outlined in August. On this year's likely revenue of about $424 billion, that is roughly $741 billion, about $53 billion more than analysts were forecasting before the call.
A message of continuity. "From a business perspective, it's business as usual."
A tighter line on the neoclouds. These are the smaller cloud companies that rent out Nvidia chips. She said Nvidia would keep supporting them but would be "disciplined and selective," and that its funding guarantees start only once a neocloud has raised the capital for its own build-out.
The shares traded at $162.32 at 6:00 p.m. Eastern time, up 4.4% from the close, on Yahoo Finance's after-hours quote. At that price the forward P/E is 10.2, on estimates that had not yet moved. If analysts raise them toward the new outlook, the multiple at $162 would be nearer 9.4.
Two cautions. The 75% is a spoken outlook from an interim team in its first week, not formal guidance, which covers only the current quarter. And we had not seen a full transcript of the call when this issue closed, so the quotes above rely on the two outlets named.
What happens to a stock after a dominant CEO leaves?
We looked at five US-listed companies whose chief executive or founder left abruptly after questions about personal conduct, and at five whose former boss came back. For each, we measured the stock against the S&P 500 from the last close before the news. We chose the cases, and five is a small number, so read the tables as a guide.
When the boss leaves after questions about conduct
Company and event
Stock the next day
3 months, vs S&P 500
6 months, vs S&P 500
12 months, vs S&P 500
Intel, CEO resigns, June 2018
−2.4%
−18.8
−5.4
−18.5
Hewlett-Packard, CEO resigns, August 2010
−8.0%
−14.9
−14.4
−36.5
McDonald's, CEO fired, November 2019
−2.7%
+5.2
+1.9
+3.2
Wynn Resorts, first report on founder, January 2018
−10.1%
−2.1
−17.6
−34.8
Papa John's, founder quits as chair, July 2018
−4.8%
+5.0
−6.1
−16.6
Median
−4.8%
−2.1
−6.1
−18.5
Nvidia, since November 6
−4.3%
−16.2 after 7 trading days
The three columns on the right are in percentage points: the stock's price change minus the S&P 500's over the same period. A reading of −18.5 means the stock did 18.5 points worse than the index.
Nvidia has lagged the market in seven trading days by about as much as the typical case did in a year
McDonald's, 12 months3.2 pts
Nvidia, 7 trading days so far−16.2 pts
Papa John's, 12 months−16.6 pts
Intel, 12 months−18.5 pts
Wynn Resorts, 12 months−34.8 pts
Hewlett-Packard, 12 months−36.5 pts
Share price change minus the S&P 500's change, in percentage points, measured from the last close before the news. Nvidia is measured from November 6 to November 17, 2026.
Source: Yield Theory calculation from Yahoo Finance closes
The first day understates it. The median stock fell 4.8% the next day and ended the year 18.5 points behind the market. Four of the five lagged over 12 months.
The company with a ready successor did fine. McDonald's named a new CEO in the same announcement and beat the market over the year. Hewlett-Packard hired an outsider in 55 days, replaced him within a year and lagged by 36.5 points. Intel left its finance chief in charge for seven months before giving him the job, and lagged by 18.5.
Nvidia has moved faster. It has trailed the S&P 500 by 16.2 points in seven trading days, and other chip stocks by 19.4. The median case took 12 months to lose that much.
That doesn't mean the damage is over. It means the market has already priced Nvidia as a typical case. Nvidia's problem is larger than typical in one way, because government investigations touch the story. It is smaller in another, because its supply is ordered further ahead than a restaurant chain's or a casino's sales.
When the old boss comes back
Company and event
Stock the next day
12 months, vs S&P 500
Stock vs S&P 500 in the year before
Apple, Steve Jobs named interim CEO, September 1997
+2.0%
+64.8
Behind by 33 points
Charles Schwab, founder returns, July 2004
+6.6%
+54.1
Behind by 39 points
Starbucks, Howard Schultz returns, January 2008
+8.1%
−10.4
Behind by 48 points
Starbucks, Schultz returns again, March 2022
+5.2%
+28.2
Behind by 31 points
Disney, Bob Iger returns, November 2022
+6.3%
−11.3
Behind by 25 points
Median
+6.3%
+28.2
Every one of these stocks rose on the news, and three of the five beat the market over the next year. But in every case the old boss returned to a company whose shares had been lagging badly, and none had left over conduct.
Larger studies are less kind to comebacks. Researchers writing in MIT Sloan Management Review in 2020 looked at 167 returning CEOs at companies in the S&P 1500 between 1992 and 2017. On average, their companies' annual stock performance was 10.1% lower than that of first-time CEOs.
What would a Huang return be worth?
If the whole $36.20 discount vanished on a return, a 7% chance of that would be worth about $2.50 a share today. It wouldn't all vanish, because the investigations and the export rules would still be there. We put the value of the comeback option at $1 to $2 a share. That is too small to buy the stock for.
Tuesday evening gave a rough test of that. In the hour after the board said it does not intend to reinstate him, Kalshi's odds fell from 12% to 7%. Over the same hour Nvidia reported its results and raised its outlook, and the shares rose about 4%. The comeback was never what the price was resting on.
What changes at Nvidia without Huang?
Next year's products don't change. They are designed, and the parts are ordered. What changes is who decides four things: how Nvidia spends its cash, how it deals with Washington, how it handles the legal fallout, and what becomes of the largest personal stake in the company.
Who signs the next deal?
Under Huang, Nvidia became a financier of its customers as well as their supplier. The results commentary lists what that has built up.
Commitment
Amount
What it is
Supply and capacity
$390 billion
Parts and factory space ordered from suppliers
Data center leases not yet started
$74 billion
Leases of up to 20 years, beginning between now and fiscal 2033
Cloud service agreements
$28 billion
Computing power Nvidia rents for its own research
Equity investments
$20 billion
Promised future investments in AI labs and financing firms
Capital spending
$6 billion
Equipment and buildings
Total commitments
$518 billion
Neocloud agreements
$32 billion
Capacity Nvidia has agreed to pay for at smaller cloud companies, once they secure financing and it is running
Guarantees, maximum exposure
$118 billion
Promises to pay if partners default, including $105 billion for an Ohio campus leased to OpenAI
A neocloud is a smaller cloud company built to rent out Nvidia chips. The Ohio guarantees start only as each building is finished, and the first is expected in fiscal 2029. Nvidia says it has made no payments under any guarantee so far.
On top of this, Nvidia spent $58.1 billion buying stakes in other companies in the first nine months of the fiscal year. It now holds $44.2 billion of listed shares and $65.5 billion of stakes in private companies. In the latest quarter it wrote $1.8 billion off the private ones. A deal to buy the AI software platform Hugging Face for about $11.9 billion, signed in September, has not closed.
In our view this is the part of Nvidia that depended most on Huang's judgment. We doubt an interim CEO who is also the finance chief will sign the next $100 billion guarantee, and a new outside CEO would probably review the ones already signed. For Nvidia's shareholders that is not obviously bad: money not spent on stakes in customers can go to buybacks at 9.8 times earnings. For the companies on the other side of these deals it is a problem, and we come back to them below.
How exposed is Nvidia to new export rules?
On November 16 the Commerce Department issued a rule that does two things. It requires a license to send advanced chips to overseas data centers that will serve companies headquartered in China and a group of other restricted countries. And it requires a license for advanced chips going to Indonesia, Malaysia, Singapore, Thailand or Brazil, with an exemption for buyers headquartered in the US who report what they do. It takes effect when it is published in the Federal Register.
A license requirement is not a ban. But it puts a government decision in front of each sale, and Nvidia's standing in Washington is weaker than it was a month ago.
Here is what the filings say about the exposure.
China. Customers headquartered in China, including Hong Kong, made up 9.1% of revenue last fiscal year and 8.2% ($7.9 billion) in the quarter to July, by the quarterly report for that period. In the latest quarter Nvidia shipped $2.4 billion of older data center chips to China. Its $130.0 billion forecast for the current quarter assumes no data center chip revenue from China at all. So the forecast already has China at zero.
Southeast Asia. This is harder to see. Until last year Nvidia reported revenue by where the bill was sent, and Singapore was 18% of revenue in fiscal 2025 on that basis, though Nvidia said under 2% of its products actually shipped there. It now reports by where the customer is headquartered, and the Singapore line is gone. Nvidia disclosed the change. Critics, including the investor Michael Burry, have argued in posts reported by Benzinga that it makes China-linked sales harder to track. No evidence of a misstatement has been published.
The delayed quarterly report matters here, because the country table is in it.
We can't put a precise number on the new rule. In our bear case we assume export limits and slower licensing cost Nvidia a large slice of its sales outside the US, and we cut next year's revenue by about $130 billion against forecasts.
How big could the legal bill be?
Four processes are running: the board's own investigation, a shareholder class action in federal court in California, a criminal investigation of the former customer Megaspeed led by the Justice Department, and, by The Wall Street Journal's reporting, a civil investigation by the SEC. Reuters said it could not confirm the last. The government's statement on Megaspeed does not accuse Nvidia of anything.
As of November 17, no charge or enforcement action against Nvidia or Huang had been made public.
Fines alone are unlikely to matter at this scale. The largest stand-alone penalty the Commerce Department's export enforcement arm has ever imposed is $300 million, on Seagate in 2023, for selling hard drives to Huawei. Nvidia earns that in about ten hours. The legal outcomes that would matter are restrictions: tighter conditions on export licenses, limits on financing customers, or findings that force changes in senior management beyond Huang. None of these has been alleged by any authority. They are the tail risk, and they are why our bear case exists.
What about Huang's 870.6 million shares?
Huang owned 870,604,104 shares in March, 3.58% of the company, by the proxy statement. At Tuesday's close they were worth $135 billion. As a former officer with less than 10%, he generally no longer has to report his trades.
Huang has given no sign of selling. His five SEC filings this year show no open-market sale, and his statement on November 17 says he wants to come back.
If he did sell, the market could absorb it. About 152 million shares changed hands on an average day over the last three months, so his whole stake equals under six days of trading. Nvidia's remaining buyback authority, $227.3 billion, is 1.7 times the value of everything he owns. The precedent investors will have in mind is Steve Wynn, who sold his entire stake in Wynn Resorts within weeks of resigning in 2018, CNBC reported at the time. We treat the stake as a risk to sentiment more than to the price.
And the people?
Nvidia's own annual report warns that "we are highly dependent on the services of our longstanding executive team." It had about 42,000 employees at the end of its last fiscal year, and much of their pay is in shares that are worth 35% less than six weeks ago. Nothing in the filings measures departures yet. The speed of the CEO search is the best early sign, which is why it sits at the top of our watch list.
Which stocks are affected by Nvidia's fall?
Since November 6, Nvidia has lost $752 billion of market value. Broadcom has gained $229 billion and AMD $90 billion.
Since Huang's apology, rivals have gained and companies that lean on Nvidia have lost
Broadcom (rival)16.3%
Celestica (supplier)14.4%
AMD (rival)12.1%
Astera Labs (supplier)11.0%
Arista Networks (supplier)9.2%
CoreWeave (neocloud)7.5%
Marvell (rival)5.3%
Micron (supplier)3.2%
Chip stock index2.7%
Samsung Electronics (supplier)0.8%
Oracle (customer)−0.9%
Dell (server maker)−1.7%
SK Hynix (supplier)−3.0%
TSMC (supplier)−4.2%
Microsoft (customer)−4.6%
Super Micro (server maker)−16.1%
Nvidia−16.7%
Nebius (neocloud)−18.1%
Share price change from the close on November 6, 2026 to the close on November 17, 2026. TSMC is its US-listed shares; Samsung and SK Hynix are their Seoul-listed shares. The chip stock index is the PHLX Semiconductor index.
Source: Yield Theory calculation from Yahoo Finance closes
Every stock in the chart is still below its October 6 level. The rivals have recovered some of that loss since November 6. Broadcom is now 9.1% below its October 6 close and AMD 22.5% below. Nvidia is 35.0% below, Super Micro 44.5% and Nebius 57.6%.
The big suppliers say chip volumes are fine. TSMC makes Nvidia's chips. SK Hynix, Samsung and Micron sell it memory. If investors thought Nvidia would sell far fewer chips next year, these stocks would have fallen with it. They have moved between −4% and +3% since November 6, close to the index. So the market's worry, as we read it, is not demand for AI chips. It is how much of the profit Nvidia keeps, and what the investigations cost.
Rivals are priced for share gains that take years. The case for Broadcom and AMD is that big customers will lean harder on alternatives while Nvidia has no permanent leader. That is reasonable. But a custom chip takes years to design and bring into production, as our page on custom AI chips versus GPUs explains, and the chips for 2027 are already chosen. AMD now costs 28.5 times next year's expected earnings. A buyer today is paying nearly three times as much for a dollar of AMD's expected profit as for a dollar of Nvidia's.
The companies that lean on Nvidia fell hardest once he resigned. Super Micro builds servers around Nvidia chips. Nebius and CoreWeave are neoclouds. The chart starts on November 6, and chip stocks bounced in the first days of that stretch. Measured from November 11, the last close before the resignation, the picture is clearer.
Stock
Since Nov 6
Since Nov 11
Nebius
−18.1%
−31.5%
Super Micro
−16.1%
−20.7%
CoreWeave
+7.5%
−17.9%
Dell
−1.7%
−13.6%
Nvidia
−16.7%
−10.5%
Chip stock index
+2.7%
−8.7%
All four have fallen further than Nvidia itself since the resignation. These businesses need early supply of Nvidia's newest chips, and the neoclouds also need help with financing. Nvidia held $4.1 billion of CoreWeave shares at the end of September, by its latest holdings report, and its filings list $32 billion of agreements with neoclouds.
Nvidia stock forecast: three scenarios
We don't forecast a share price. We work out what the stock would be worth under three sets of assumptions for the fiscal year that ends in January 2028, and how likely each looks to us.
Each case needs two numbers: earnings per share, and the multiple investors pay for them. Earnings come from a short model: revenue, times gross margin, less about $47 billion of operating costs, less 17% tax, divided by the share count after buybacks, which we put at 23.3 to 23.7 billion. With analysts' revenue forecast of $688.7 billion and a 72.3% margin, a little above the low Nvidia has guided to for this quarter, the model gives $15.86, the same as the consensus.
Bear
Base
Bull
What happens
No permanent CEO for a year, or sanctions; export licenses slow; big customers shift orders
A credible CEO by mid-2027; investigations end without sanctions on the company; some orders slip
A strong CEO is named early, or the board changes course on Huang; revenue grows at the 75% Nvidia outlined on Tuesday's call
Revenue, year to January 2028
$560 billion
$690 billion
$741 billion
Growth on this year
32%
63%
75%
Gross margin
69%
72%
73%
Earnings per share
$12.10
$15.80
$17.30
Multiple
9 times
12 times
14.5 times
Value per share
$109
$190
$251
Change from $155.46
−30%
+22%
+61%
Our probability
25%
50%
25%
Weighted by those probabilities, the three values average $185, which is 19% above Tuesday's close and 14% above the after-hours price of $162.32.
What Nvidia is worth at different earnings and multiples
Pick next year's earnings and the multiple investors pay
Value per share$109$145$175$214$142$190$229$280$156$208$251$306
−29.9%−6.7%+12.6%+37.7%−8.7%+22.2%+47.3%+80.1%+0.3%+33.8%+61.5%+96.8%below Close on November 17below Close on November 17above Close on November 17above Close on November 17below Close on November 17above Close on November 17above Close on November 17above Close on November 17above Close on November 17above Close on November 17above Close on November 17above Close on November 17
Close on November 17$155.46
$109$306
Earnings per share times the multiple. Our bear, base and bull cases pair $12.10 with 9 times, $15.80 with 12 times and $17.30 with 14.5 times. Illustrative, not a forecast.
The base case is below Nvidia's own outlook on purpose. It uses $690 billion of revenue, about what analysts forecast before the call and $51 billion less than the 75% growth Kress outlined. And it pays 12 times, far less than the 15.1 times the stock fetched at its record or the 17.7 times the S&P 500 fetches now. That multiple is our allowance for a company with no permanent CEO and four open legal processes. The value still comes out 22% above the close.
The bear case is not a collapse. It has revenue growing 32% and margins three points lower, at a multiple below today's. It gets to $109. A deeper fall in earnings is possible if the AI build-out itself turns down, and that risk existed before Huang left. It is the subject of our capex tracker, and it would hurt every stock in the table above.
The bull case doesn't need Huang. It needs the revenue Nvidia outlined on Tuesday, which its supply orders roughly support, and a multiple still below Broadcom's.
The grid shows what the price implies. Today's $155.46 sits between the bear and base cells. Even bear-case earnings at the base-case multiple give $145, only 7% below the close.
Should I buy, hold or sell Nvidia stock?
These are our views, with the reasoning, not personal advice. Your position size, tax situation and time horizon matter more than our numbers.
Call 1: Nvidia is a buy at $155 to $165, in stages, with a 12-month value of about $185
The price assumes that earnings stop growing. Nvidia's own outlook, raised on Tuesday's call, is for revenue to grow about 75%. Our base case sits between the two and is worth $190. The three cases together average $185. We call that a 12-month value because by November 2027 the fiscal year these earnings belong to will be almost over, and the market will be pricing them as close to known.
The shares were about $162 after the call. At that price the base case is 17% higher and the bear case 33% lower, so the call holds, with a little less room.
We would buy in stages because the bad news isn't finished. Three dated events could each move the stock sharply: the detailed quarterly report due by December 4, Huang's interview with House committees on December 1, and the first findings of the board's investigation, which has no date. A buyer who puts a third in now, a third after the quarterly report and a third once a CEO is named gives up some upside in exchange for seeing those cards.
For an investor who already holds the stock, we would not sell here. Nvidia has already lost nearly as much ground against the market as the median earlier case lost in a full year. Selling at 9.8 times earnings to avoid a further fall means being right about the bear case, to which we give one chance in four.
Size it so that $109 doesn't hurt. If a fall of 30% or more in this position would force you to sell, the position is too big.
Call 2: Don't swap Nvidia for AMD or Broadcom at these prices
The swap has already been made by others, and it shows in the prices. AMD's multiple has gone from 25.5 to 28.5 in seven trading days and Broadcom's from 15.1 to 17.6. Between them they have added $319 billion of market value.
For that to be right, Nvidia's loss has to become their gain within a year or two. The supply orders, the chips already in production and the time it takes to design an alternative all argue against that timetable. Broadcom's long-run case in custom chips stands on its own, and we would hold it. We wouldn't add to either stock because of Huang.
Call 3: Avoid the neoclouds until Nvidia's next CEO sets out a capital policy
Nebius and CoreWeave have fallen 31.5% and 17.9% since the last close before the resignation. They may look like the cheap way to bet on a recovery. We think they carry more of the Huang risk than Nvidia does.
Nvidia's value rests on chips it has already ordered parts for. A neocloud's rests on getting the next chips early and on financing them. The $32 billion of neocloud agreements in Nvidia's filings were signed under Huang, as were its financing guarantees, and we expect a new chief executive, or a cautious interim one, to review them first. Kress's words on Tuesday's call, that Nvidia would be "disciplined and selective" with these companies, read to us like the start of that review. Until its outcome is public, these stocks are a bet on a decision nobody has made yet.
Super Micro, down 20.7% since the resignation, sits in the same group: its edge has been early access to Nvidia's newest chips.
Risks to these calls
The investigation finds more. Our scenarios assume the facts are roughly as reported. If the board's lawyers or prosecutors find that company secrets or controlled chips reached people they shouldn't have, the bear case becomes the base case.
Earnings are at a peak. Micron's 3.4 multiple is what the market does to profits it thinks are cyclical. If the big cloud companies cut spending in 2027, Nvidia's $203 billion of orders for next year turns from an asset into a liability.
A long vacancy. Every month without a permanent CEO raises the chance that senior engineers leave and that customers hedge. Hewlett-Packard is the warning: two successors in 14 months and a stock 36.5 points behind the market after the first 12.
Our comparison group is small. Five departures and five returns, chosen by us, in different industries. No company of Nvidia's size has lost a founder this way, so the tables are a guide and not a base rate.
Estimates are stale. The consensus figures here were read on Tuesday evening and had not yet moved for the results or the higher outlook. We expect them to rise, though some analysts may cut for the leadership change. The forward P/E will move with them.
What would change our view
The quarterly report, due by December 4. Filed on time with no new legal disclosures: supports the base case. A notice of late filing: we would stop buying.
The CEO search. A respected chip executive named before the fourth-quarter results in late February would move us toward the bull case. No name by next June's annual meeting would move us toward the bear case.
Huang's House interview on December 1, and whatever the committees say afterwards. A statement that clears the company would remove a large part of the discount.
Fiscal 2028 guidance. On Tuesday Nvidia raised its outlook for next year's revenue growth to about 75%. Formal guidance comes with fourth-quarter results in late February. If the 75% is still standing then, our base case is too low. If it is cut or withdrawn, the base case is too high.
Gross margin below 71% in the current quarter would say the memory costs locked into the supply orders are biting harder than planned.
Days of sales owed by customers above 70 would say Nvidia is lending more to keep orders flowing.
New export limits. The November 16 rule adds licenses for five countries. The commerce secretary testifies to the Senate on December 10. A rule that reaches sales to US-based cloud companies abroad would move us toward the bear case.
How we built this
Share prices are Yahoo Finance closing prices through November 17, 2026, without dividends. Market values use 24.1 billion shares, the basic share count in Nvidia's third-quarter results. The chip index is the PHLX Semiconductor index.
Results, forecasts, commitments, guarantees and cash flow figures are from Nvidia's results release and the finance chief's commentary, filed with the SEC on November 17, 2026. Ownership is from the proxy statement filed on May 12, 2026. Country revenue is from the quarterly report for the period that ended July 26, 2026 and the annual report for fiscal 2026.
Forward P/E ratios use Yahoo Finance's consensus earnings estimates, read on November 17 before Nvidia reported and again at 6:00 p.m. Eastern time, when they had not changed; the S&P 500's uses FactSet's estimate of $415.82 for the next 12 months. Nvidia's estimate for this fiscal year, about $9.60, is our own: nine months as reported plus the fourth quarter implied by company guidance.
The trailing P/E history divides each day's closing price by GAAP diluted earnings per share for the four most recent quarters on file with the SEC that day, adjusted for the share splits of 2021 and 2024.
The CEO tables measure price changes from the last close before each announcement to the close 91, 182 and 365 days later, against the S&P 500 over the same days. Announcement dates are from company statements.
Remarks from the November 17 earnings call are from The Verge's transcript of the chair's statement and Yahoo Finance's live coverage. The after-hours price is Yahoo Finance's quote at 6:00 p.m. Eastern time, and the Kalshi prices were read at 5:00 p.m. and 6:00 p.m.
Scenario earnings come from the model described above, with 23.3 to 23.7 billion shares depending on the price at which buybacks are made. Multiples and probabilities are our judgment.
Reports about investigations are attributed to the outlets named. Allegations about the reasons for Huang's departure have not been tested in court, and no charge against Huang or Nvidia had been made public as of November 17, 2026.
A detailed photonics investment study covering five companies, optical architectures, manufacturing yields, reliability, cash conversion and valuation. Includes worked models and the evidence that would strengthen or break the thesis.
Homebuilders fall about five times as much as the S&P 500 when Treasury yields rise, and banks don't gain. Our 2026 rate screen, valuations and three calls.
The thesis, the numbers behind it, and what would break it. Full access is $39 a month.