Revenue up 106% year-over-year, stock down 28.6%: Nvidia is pricing a problem that isn't decided by its financials
Nvidia's second-quarter revenue was $96.2 billion, up 106% year-over-year, but its stock fell from $243 on October 6 to a closing price of $173.73 on November 11, a decline of 28.6%, with a price-to-sales ratio of 13.85 times, the lowest in five years. The author believes this decline stems mainly from a governance discount rather than the financials, and lists drags such as a downward-sloping gross margin guidance and $530 billion in commitments and guarantees. He gives the conditions for a wait signal: the board's conclusions land, third-quarter gross margin holds at 73.5%, and the stock price climbs back above 185 on shrinking volume.
Why read this
AI score 72
The article breaks down Nvidia's financial data, valuation multiples and governance events together, letting readers see the mismatch between the falling stock price and accelerating fundamentals.
Topics
Body · AI translation
Revenue doubled, stock down nearly 30%, valuation lowest in five years
Last quarter's revenue was $96.2 billion, double year over year, with growth accelerating for a fourth consecutive quarter. The stock fell from $243 on October 6 to a close of $173.73 on November 11, a drop of 28.6%, and fell to around $160 in premarket trading on November 12. There is nothing wrong with the earnings; what's falling is the person. My judgment: the business is still accelerating, the valuation has fallen to a five-year low, but the price is now pricing in a problem that earnings reports cannot control, and I won't catch a falling knife before a signal appears.
A $96.2 billion business per quarter, 90% from data centers
NVIDIA sells AI compute, not individual chips. The GPUs, CPUs, switches, network cards, and software in a rack are packaged into an AI factory solution sold to customers, with money collected for the entire system. This business can be explained clearly in five minutes.
Second fiscal quarter (ended July 26) revenue was $96.2 billion, up 106% year over year and 18% quarter over quarter. Data centers brought in $89 billion, 92.5% of the total, up 117% year over year.
Data centers split into two parts. Hyperscale cloud providers accounted for $48.7 billion, up 13% quarter over quarter and double year over year. AI cloud, industrial, and enterprise customers accounted for $40.3 billion, up 138% year over year and 25% quarter over quarter. The latter is growing twice as fast as the former, showing that buyers are spreading from five cloud giants to new cloud providers, sovereign funds, and enterprises.
Edge computing (mainly workstation graphics cards) was $7.2 billion, up 27% year over year, accounting for less than 8% and dragging down the total, because memory price increases suppressed consumer PC sales.
There are three cards not yet played. The Vera Rubin platform began mass production in August, which the company says is the fastest-ramping product in its history, accounting for about 20% of data center revenue in the third quarter. Vera CPU revenue is set to more than double next year. China data center compute revenue is zero in guidance, which amounts to a free option.
The moat, quantified: for every gigawatt of data center, NVIDIA could get $18 billion in the Hopper era, $25 billion with Blackwell, and $40 billion with Vera Rubin. The same kilowatt-hour of power can sell for more money, which is the fundamental reason its gross margin can stay at 75%.
The books: 75% gross margin, 66% operating margin, but cash is starting to lag profit
Gross margin was 75.0%, up 2.6 percentage points year over year. Operating margin was 66.2%, with operating profit of $63.7 billion. This number is abnormal for a hardware company, showing that what it sells now is not hardware but scarce capacity.
Revenue scale has long ceased to be a problem. The past twelve months totaled $303 billion, with $96.2 billion in a single quarter.
Growth is no problem either. 62%, 73%, 85%, 106% — accelerating for four consecutive quarters.
Cash flow has a problem. Second fiscal quarter operating cash flow was $24.1 billion, half of the first fiscal quarter's $50.3 billion. Free cash flow was $21.4 billion, versus $48.6 billion the previous quarter.
The reason is accounts receivable. Quarter-end receivables were $63.1 billion, up 55% quarter over quarter, while revenue grew only 18%. Days sales outstanding stretched from 45 days to 60 days. Inventory also rose to $31.6 billion, up 22% quarter over quarter. These two items plus other working assets locked profit in customers and warehouses, without turning into cash.
GAAP net income of $59.7 billion includes $7.8 billion of investment gains; non-GAAP net income was $54 billion. Use the latter to assess profit quality.
The balance sheet is still clean. Cash plus short-term investments were $62.5 billion, long-term investments $51.2 billion, total debt $38.9 billion, and net cash $23.6 billion. The company issued $24.9 billion of debt this quarter while repurchasing $22.1 billion and paying $6 billion in dividends, reducing shares outstanding from 24.3 billion to 24.1 billion.
The real risk is off the balance sheet. The company disclosed commitments and guarantees totaling more than $530 billion, including $108.5 billion of credit guarantees, of which $105 billion is for OpenAI's Ohio project, plus $36 billion of cloud service purchase commitments. One professor estimates total obligations at about $400 billion. These are not today's debt, but they are a structure of "if the customer can't afford it, I'll help them buy it."
The market pays only $13.8 for every dollar of revenue, the lowest in five years
Market cap is $4.195 trillion. Price-to-sales is 13.85 times, price-to-earnings 21.97 times, price-to-book 18.3 times, and free cash flow multiple 33 times. Based on the premarket $160, price-to-sales is about 12.8 times and price-to-earnings about 20 times.
Historically, this is the lowest in five years. The price-to-sales range over the past five years was 12 to 35 times, and from 2024 to 2025 it was persistently 24 to 30 times. The price-to-earnings was still 26 times at the end of last quarter.
Compared with peers, the two main competitors have price-to-sales of 18 to 20 times, putting it in the cheapest tier.
What expectations is this multiple pricing in? The market's EPS expectation for the next fiscal year (FY2028) is $15.79, corresponding to a forward P/E of 11 times. The company itself says FY2028 revenue growth will be about 70%; based on roughly 406 billion in FY2027, FY2028 revenue would be about 690 billion, corresponding to a forward P/S of about 6 times. In other words, the price the market is paying now implies the assumption that "70% growth cannot be delivered."
The average analyst consensus target price is $311.51, the median 306.55, the low 180, and the high 515. But this average is already stale. Bernstein cut its target from 400 to 200 and downgraded, Baird cut from 500 to 200 and downgraded, Stifel cut from 315 to 205, Oppenheimer withdrew its target outright, and Morgan Stanley removed it from its top picks. The pre-crisis average cannot serve as today's anchor.
What fell was not the earnings report, but one person
On November 8, CEO Jensen Huang admitted to an extramarital affair. The first was Alice Huang, an executive at cloud company Megaspeed, whom U.S. intelligence suspects of espionage for China. The second was Kimberly Powell, NVIDIA's own vice president of the medical business, who reports directly to Jensen Huang.
After the close on November 11, the Wall Street Journal and Bloomberg reported that the board had hired an outside firm to search for a successor and would meet that week to decide his fate. Members of Congress called for his resignation, and the FBI has opened an investigation.
The stock's reaction: down 9.3% on November 5, down 4.3% on November 9, down 5.5% on November 10, a cumulative 18.2% over three trading days, and down another roughly 7.6% premarket on November 12 to 160.5. Measured from the October 6 close, about $1.6 trillion in market value has evaporated.
This is a pure governance discount. Over the same period, AMD and Broadcom rose after hours, showing that the market is not repricing AI demand, only repricing NVIDIA as a company.
The second drag is the downhill gross margin. Memory prices have risen beyond the company's own expectations; third-quarter gross margin guidance is 74%, it bottoms at 71% to 72% in the fourth quarter, and only returns to 72% to 73% in FY2028. Revenue is still rising while margins are falling, and when the two directions diverge, the market believes margins first.
The third is doubts about circular financing. 530 billion in commitments, 108.5 billion in guarantees, and credit support for customers have its bond credit default swap spreads trading like those of a BBB-rated company, while its rating is AA. Morgan Stanley's credit analysts call this approach "balance sheet as a service."
The fourth is macro. The November 10 inflation data came in hot, and the market began betting on a December rate hike.
Now is not a buying point, but a position to wait for signals
First, what I will not do. I will not build a position before the November 17 earnings report, I will not build a position before the board's conclusion comes out, and I will not average down during the decline just because "the company is good."
The trend points to 4 o'clock. Fundamentals are accelerating, yet the stock has broken below all of the 20-day, 50-day, 120-day, 200-day, and 250-day moving averages, and every moving average sits above the price. The 20-day line is 203.8, the 50-day 217.1, the 200-day 204.5, and the 250-day 200.3. This is the market rejecting the stock, not the market hesitating.
Key price bands.
164.27 is the March low and also the 52-week low. The premarket 160.5 on November 12 has already broken through it; this line is now broken.
Below that is 150 to 155. This is the July 2025 platform and the next high-volume zone after the August low of 168.8 was lost.
Above is 175 to 185. This band holds 37% of the past twelve months' volume, but to be clear, that volume was mainly dumped in the last five trading days. Average daily volume over those five days was 340 million shares, 2.6 times the 50-day average. This is a cost basis formed by panic turnover, not a long-term consensus, and cannot be treated as reliable support. The November 9 gap from 179.9 to 186.6 is also here, and a rebound into this area will meet trapped sellers.
Further up is 200 to 205, where the 200-day and 250-day lines sit. The strongest resistance is at 210 to 215; 62% of the past three months' volume is piled in these two bands, the cost center of the past six months. For the price to return there would mean every trapped holder gets back to even.
Three things to have. Expectations: the board's conclusion lands, plus third-quarter gross margin of no less than 73.5%—only when both conditions are met can we even talk about "governance risk cleared." Bottom line: if after these two conditions are met I test a position between 165 and 175, I set a stop at 150 and exit on a break below the July 2025 platform. Plan: execute on a break, and do not change my mind because "the CUDA moat is still there."
November 17 decides two things in one day
After the U.S. market close on November 17, which is the morning of November 18 Sydney time, comes the fiscal third-quarter earnings report. Consensus revenue is 108.8 billion, non-GAAP EPS $2.47. Company guidance is 108 billion plus or minus 2%, with gross margin of 74%. What matters is not this quarter but the fourth-quarter guidance and gross margin. The threshold circulating in the market is fourth-quarter revenue above 124 billion and gross margin no lower than 73.5%; below that combination, the decline will continue.
The board's decision on whether Jensen Huang stays or goes has no date and could come at any time. This is a binary event and the biggest variable in the current price. Some analysts estimate that if he leaves, the stock would fall another 20% to 30%.
The December Fed meeting. Inflation is hot and the market is betting on a rate hike, which determines the overall discount on high-valuation growth stocks.
Next February's fourth-quarter earnings report is the verification point for gross margin bottoming at 71% to 72%. If the actual figure is lower, it means memory costs are out of control, a longer-term problem than the CEO scandal.
Conclusion: two signals; if they don't come, don't move
A good business at a bad time. Fundamentals are the best in five years and valuation is the lowest in five years, but the price is being set by a problem that the earnings report does not decide.
Wait for two signals to appear at the same time: the board's conclusion is clear, and third-quarter gross margin holds at 73.5%, while the stock reclaims above 185 on shrinking volume. If neither arrives, don't move. If only the first arrives, watch. If only the second arrives, don't chase.
I am not a financial advisor; the above merely lays out the data and the logic.
Data gaps: third-quarter actual figures are not yet out, so consensus and company guidance are used; the average analyst target price includes stale pre-crisis values, and the new post-cut target prices are marked separately; the ultimate loss on the 530 billion in commitments and guarantees cannot be quantified; the 160.5 on November 12 is a premarket price, not a closing price; the technical picture is based on daily closes and volume distribution and does not include options positioning.
The above does not constitute investment advice.