#Gate股票畅聊季 Why Even an Earnings Beat Can't Save Nvidia
In the early hours of November 18 Beijing time, Nvidia delivered a quarterly report that was almost impossible to fault. Revenue was $115.8 billion, double year over year and nearly $6.5 billion above market expectations; next-quarter guidance was $130 billion, $4.7 billion above expectations; even its revenue growth outlook for fiscal 2028 was raised from 70% to 75%. This was the 17th consecutive quarter it had beaten Wall Street expectations. Its shares rose about 4% after hours—a very limited recovery.
It is unusual for the gap between “better-than-expected results” and “a stock that doesn’t care” to widen this much. The market isn’t rejecting this earnings report. It’s saying: I’m not testing you on this question anymore.
01 The Market Has Changed the Exam
Start with November 5. That day, the Nasdaq fell 5.52%, the Philadelphia Semiconductor Index fell 12.25%, and Nvidia dropped 9.26% in a single session. No bad earnings report triggered the sell-off; it was the midterm election results. Democrats won both chambers of Congress, and among 41 local data-center-related ballot measures across 10 states, 40 leaned toward restricting development. Texas’s newly elected governor even pledged during the campaign to pause new approvals until dedicated legislation was enacted.
Together, these two developments shook the two external prerequisites for expanding AI computing capacity: permits and power. Morgan Stanley’s October report estimated that U.S. data-center electricity demand would reach 97 GW by 2028. Even accounting for natural gas generation, fuel cells, and nuclear power, the net shortfall would still be 33 GW. You can buy GPUs, but you can’t deliver the power—that’s a physical constraint, not a valuation issue. Soochow Securities’ post-earnings assessment was on point: the central question has shifted from “Can demand keep growing?” to “Can high capital spending, financing costs, and investment returns be brought into alignment?”
For the past three years, Nvidia only had to prove that demand was still there; now the market wants it to prove that customers can afford that demand and that the investment will pay off. It answered the first question for 17 quarters. It hasn’t answered the second even once.
02 The Most Important Line Isn’t Revenue
The number I’m watching isn’t $115.8 billion; it’s gross margin.
Third-quarter GAAP gross margin was 74.8%, down 0.2 percentage points from 75.0% the previous quarter. The company guided to fourth-quarter gross margin of 71% to 72%, said this would be the low point, and forecast a recovery to 72% to 73% in fiscal 2028. Put the three figures side by side: 74.8%, 71.5% (the midpoint of guidance), and 72.5% (the midpoint of the fiscal 2028 commitment). By the company’s own account, gross margin has already peaked and won’t return to 74.8%.
The math isn’t hard. At the midpoint of fourth-quarter revenue guidance, $130 billion, gross margin falling from 74.8% to 71.5%—a drop of 3.3 percentage points—means $4.3 billion less gross profit in a single quarter. At that revenue scale, that’s roughly $17.2 billion a year. Nvidia’s net income for fiscal 2026 was $120.07B, so a 3.3-point drop in gross margin would mean a year’s lost profit equivalent to 14% of its net income for the previous full fiscal year.
Kress also said something else on the call: under multiyear agreements, the re-contracting price of Hopper GPUs has reached as much as twice the original price, while B300 prices have risen more than 60% since launch.
Prices are rising, yet gross margin is still expected to fall by more than 3 percentage points. Together, those two statements point to one thing: costs are rising faster than prices. HBM is getting more expensive, advanced packaging is getting more expensive, and power and data centers are getting more expensive. Nvidia’s greatest feat over the past three years wasn’t computing power; it was increasing volume and prices while also expanding gross margin. Gross margin was 71.07% in fiscal 2026 and reached 74.8% in the third quarter this year. Now that trend has turned.
This is the “pick-and-shovel seller” admitting that it can’t keep selling shovels at higher prices.
03 Profits Doubled, Cash Flow Halved
The cash-flow problem had already surfaced in the previous fiscal quarter. In the quarter ended July 26, Nvidia posted revenue of $96.22B, up 106% year over year, and net income of $59.69B. In that same quarter, accounts receivable rose from $40.7 billion to $63.06B, days sales outstanding lengthened from 45 to 60 days, and operating cash flow fell from $50.34B to $24.08B.
Management explained that some investment-grade customers had signed large multiquarter contracts, extending payment cycles. That explanation holds up, but it also reveals something else: customers are using Nvidia’s money. The 10-Q for the period ended July 26 shows that five unnamed direct customers accounted for 70% of accounts receivable, with the largest accounting for 22%. Eisman, the inspiration for the film The Big Short, seized on this figure on a podcast in October: one of the world’s largest companies by market capitalization has 70% of its accounts receivable tied up in five accounts. Extending payment terms from 45 to 60 days means Nvidia is fronting customers an extra half month of funding. At its core, that gap is a credit issue: who is financing whom.
04 The Buyers of the Shovels Can’t Make the Numbers Work
Why are customers paying late? Their books make it clear. The four largest cloud providers spent about $200 billion combined on capital expenditures in the most recent quarter, a record for a single quarter; over the same period, their combined free cash flow was negative $26.4 billion, down $33.1 billion quarter over quarter. Only Microsoft remained in positive territory; Amazon, Alphabet, and Meta all turned negative. Amazon’s capital spending that quarter was 1.8 times its operating cash flow, while Microsoft’s was just 0.7 times. GPU purchases are no longer being funded entirely by money they earned themselves. This investment ultimately has to be recouped through AI revenue, and revenue recognition is much slower than depreciation. Alphabet itself has acknowledged that depreciation and third-party computing expenses will weigh on cloud margins in the short term. Some of this investment has to be covered by external financing. Access to financing depends on whether capital markets are willing to keep providing funds. The significant widening of CoreWeave’s credit default swap spreads in October was a warning.
Three months ago, Nvidia’s cloud business had a $2 trillion backlog; now it’s $2.5 trillion. That extra $500 billion isn’t cash—it’s a promise.
How much of it is fulfilled depends not on Nvidia’s capacity, but on customers’ access to financing.
05 40% of Net Income Went to Buy Its Own Stock
There’s another piece of the math the market has barely discussed. In the third quarter, Nvidia returned $28.1 billion to shareholders through buybacks and dividends, 43% of its quarterly GAAP net income of $65.3 billion. In September, the board approved an additional $150 billion in buyback authorization, leaving $227.3 billion in authorization at quarter-end. A company whose revenue is still doubling year over year is using more than 40% of its net income to buy its own stock. That’s what mature companies do, not growth companies.
There are several ways to read this move: the company thinks its stock is cheap; it can’t find a better use for its money than buybacks; or it’s using buybacks to offset valuation shocks from executive turnover. Whichever it is, they all point to the same thing: in the eyes of insiders, Nvidia’s “growth phase” may already be over.
The governance issue can’t be ignored either. On November 12, amid personal controversy, Jensen Huang resigned as president, CEO, and director. Kress became interim CEO, and the board explicitly said it had no intention of reinstating him.
The market priced this in decisively: Bernstein cut its price target from $400 to $200, Baird from $500 to $200, while longtime bull Ives estimated “$300 if he stays, $225 with a leadership change.”
Nvidia’s business model hinges on the credibility of its roadmap; chip performance is just the price of entry. Rubin Ultra launches next year, and Feynman in 2028. Customers dare to sign multiyear agreements because they believe the schedule will be met. And for 33 years, the only person deciding “where the next technological turning point lies” has been Jensen Huang. He has no COO, no successor, a 3.5% stake, and no dual-class shares. When he’s there, this structure is called efficient; when he’s gone, it’s a vacuum. The phrase “the roadmap is on track” is now coming from an interim CEO with a finance background.
06 Chinese Production Lines Need to Adjust Their Schedules Too
If Nvidia’s fourth-quarter gross margin really lands at 71% to 72%, it will push its suppliers for lower prices. Optical modules, PCBs, liquid cooling, and power supplies already operate on thin margins, and they’ll be the first squeezed when system costs rise. This isn’t a demand problem; it’s about how profits are distributed. The AI pie is still growing, but the way it’s sliced is changing. Another issue is the delivery schedule. A $2.5 trillion backlog, annualized using fourth-quarter guidance, is equivalent to more than four and a half years of revenue. If delivery is delayed by even one quarter, plans to expand supporting capacity will have to be rescheduled. Domestic production lines supplying Nvidia lock in their production schedules quarterly; a one-quarter shift throws the entire year’s schedule off.
07 Watch Three Numbers Next
First, where fourth-quarter gross margin lands within the 71% to 72% range. Landing at the low end would mean cost pressures are still spreading; landing at the high end would mean Vera Rubin is ramping more smoothly than expected. This line matters far more than revenue.
Second, what share of the newly added backlog comes from non-investment-grade customers. A rising share would mean the credit quality of the order book is deteriorating.
Third, whether the permanent CEO is an insider or an outsider. An insider would be read by the market as “business as usual”: the roadmap stays intact, but the upside potential is discounted. An outsider would mean the roadmap needs to be reassessed, bringing greater short-term volatility.
What Nvidia needs to prove now isn’t whether AI demand exists—that no longer needs proving. It needs to prove that, with a new CEO and rising electricity prices and financing costs, this business can still sustain the kind of gross-margin structure it had over the past three years.
The earnings report proved the former; the market is waiting for the latter. What lies between the two isn’t a quarter, but a cycle.$NVDA
In the early hours of November 18 Beijing time, Nvidia delivered a quarterly report that was almost impossible to fault. Revenue was $115.8 billion, double year over year and nearly $6.5 billion above market expectations; next-quarter guidance was $130 billion, $4.7 billion above expectations; even its revenue growth outlook for fiscal 2028 was raised from 70% to 75%. This was the 17th consecutive quarter it had beaten Wall Street expectations. Its shares rose about 4% after hours—a very limited recovery.
It is unusual for the gap between “better-than-expected results” and “a stock that doesn’t care” to widen this much. The market isn’t rejecting this earnings report. It’s saying: I’m not testing you on this question anymore.
01 The Market Has Changed the Exam
Start with November 5. That day, the Nasdaq fell 5.52%, the Philadelphia Semiconductor Index fell 12.25%, and Nvidia dropped 9.26% in a single session. No bad earnings report triggered the sell-off; it was the midterm election results. Democrats won both chambers of Congress, and among 41 local data-center-related ballot measures across 10 states, 40 leaned toward restricting development. Texas’s newly elected governor even pledged during the campaign to pause new approvals until dedicated legislation was enacted.
Together, these two developments shook the two external prerequisites for expanding AI computing capacity: permits and power. Morgan Stanley’s October report estimated that U.S. data-center electricity demand would reach 97 GW by 2028. Even accounting for natural gas generation, fuel cells, and nuclear power, the net shortfall would still be 33 GW. You can buy GPUs, but you can’t deliver the power—that’s a physical constraint, not a valuation issue. Soochow Securities’ post-earnings assessment was on point: the central question has shifted from “Can demand keep growing?” to “Can high capital spending, financing costs, and investment returns be brought into alignment?”
For the past three years, Nvidia only had to prove that demand was still there; now the market wants it to prove that customers can afford that demand and that the investment will pay off. It answered the first question for 17 quarters. It hasn’t answered the second even once.
02 The Most Important Line Isn’t Revenue
The number I’m watching isn’t $115.8 billion; it’s gross margin.
Third-quarter GAAP gross margin was 74.8%, down 0.2 percentage points from 75.0% the previous quarter. The company guided to fourth-quarter gross margin of 71% to 72%, said this would be the low point, and forecast a recovery to 72% to 73% in fiscal 2028. Put the three figures side by side: 74.8%, 71.5% (the midpoint of guidance), and 72.5% (the midpoint of the fiscal 2028 commitment). By the company’s own account, gross margin has already peaked and won’t return to 74.8%.
The math isn’t hard. At the midpoint of fourth-quarter revenue guidance, $130 billion, gross margin falling from 74.8% to 71.5%—a drop of 3.3 percentage points—means $4.3 billion less gross profit in a single quarter. At that revenue scale, that’s roughly $17.2 billion a year. Nvidia’s net income for fiscal 2026 was $120.07B, so a 3.3-point drop in gross margin would mean a year’s lost profit equivalent to 14% of its net income for the previous full fiscal year.
Kress also said something else on the call: under multiyear agreements, the re-contracting price of Hopper GPUs has reached as much as twice the original price, while B300 prices have risen more than 60% since launch.
Prices are rising, yet gross margin is still expected to fall by more than 3 percentage points. Together, those two statements point to one thing: costs are rising faster than prices. HBM is getting more expensive, advanced packaging is getting more expensive, and power and data centers are getting more expensive. Nvidia’s greatest feat over the past three years wasn’t computing power; it was increasing volume and prices while also expanding gross margin. Gross margin was 71.07% in fiscal 2026 and reached 74.8% in the third quarter this year. Now that trend has turned.
This is the “pick-and-shovel seller” admitting that it can’t keep selling shovels at higher prices.
03 Profits Doubled, Cash Flow Halved
The cash-flow problem had already surfaced in the previous fiscal quarter. In the quarter ended July 26, Nvidia posted revenue of $96.22B, up 106% year over year, and net income of $59.69B. In that same quarter, accounts receivable rose from $40.7 billion to $63.06B, days sales outstanding lengthened from 45 to 60 days, and operating cash flow fell from $50.34B to $24.08B.
Management explained that some investment-grade customers had signed large multiquarter contracts, extending payment cycles. That explanation holds up, but it also reveals something else: customers are using Nvidia’s money. The 10-Q for the period ended July 26 shows that five unnamed direct customers accounted for 70% of accounts receivable, with the largest accounting for 22%. Eisman, the inspiration for the film The Big Short, seized on this figure on a podcast in October: one of the world’s largest companies by market capitalization has 70% of its accounts receivable tied up in five accounts. Extending payment terms from 45 to 60 days means Nvidia is fronting customers an extra half month of funding. At its core, that gap is a credit issue: who is financing whom.
04 The Buyers of the Shovels Can’t Make the Numbers Work
Why are customers paying late? Their books make it clear. The four largest cloud providers spent about $200 billion combined on capital expenditures in the most recent quarter, a record for a single quarter; over the same period, their combined free cash flow was negative $26.4 billion, down $33.1 billion quarter over quarter. Only Microsoft remained in positive territory; Amazon, Alphabet, and Meta all turned negative. Amazon’s capital spending that quarter was 1.8 times its operating cash flow, while Microsoft’s was just 0.7 times. GPU purchases are no longer being funded entirely by money they earned themselves. This investment ultimately has to be recouped through AI revenue, and revenue recognition is much slower than depreciation. Alphabet itself has acknowledged that depreciation and third-party computing expenses will weigh on cloud margins in the short term. Some of this investment has to be covered by external financing. Access to financing depends on whether capital markets are willing to keep providing funds. The significant widening of CoreWeave’s credit default swap spreads in October was a warning.
Three months ago, Nvidia’s cloud business had a $2 trillion backlog; now it’s $2.5 trillion. That extra $500 billion isn’t cash—it’s a promise.
How much of it is fulfilled depends not on Nvidia’s capacity, but on customers’ access to financing.
05 40% of Net Income Went to Buy Its Own Stock
There’s another piece of the math the market has barely discussed. In the third quarter, Nvidia returned $28.1 billion to shareholders through buybacks and dividends, 43% of its quarterly GAAP net income of $65.3 billion. In September, the board approved an additional $150 billion in buyback authorization, leaving $227.3 billion in authorization at quarter-end. A company whose revenue is still doubling year over year is using more than 40% of its net income to buy its own stock. That’s what mature companies do, not growth companies.
There are several ways to read this move: the company thinks its stock is cheap; it can’t find a better use for its money than buybacks; or it’s using buybacks to offset valuation shocks from executive turnover. Whichever it is, they all point to the same thing: in the eyes of insiders, Nvidia’s “growth phase” may already be over.
The governance issue can’t be ignored either. On November 12, amid personal controversy, Jensen Huang resigned as president, CEO, and director. Kress became interim CEO, and the board explicitly said it had no intention of reinstating him.
The market priced this in decisively: Bernstein cut its price target from $400 to $200, Baird from $500 to $200, while longtime bull Ives estimated “$300 if he stays, $225 with a leadership change.”
Nvidia’s business model hinges on the credibility of its roadmap; chip performance is just the price of entry. Rubin Ultra launches next year, and Feynman in 2028. Customers dare to sign multiyear agreements because they believe the schedule will be met. And for 33 years, the only person deciding “where the next technological turning point lies” has been Jensen Huang. He has no COO, no successor, a 3.5% stake, and no dual-class shares. When he’s there, this structure is called efficient; when he’s gone, it’s a vacuum. The phrase “the roadmap is on track” is now coming from an interim CEO with a finance background.
06 Chinese Production Lines Need to Adjust Their Schedules Too
If Nvidia’s fourth-quarter gross margin really lands at 71% to 72%, it will push its suppliers for lower prices. Optical modules, PCBs, liquid cooling, and power supplies already operate on thin margins, and they’ll be the first squeezed when system costs rise. This isn’t a demand problem; it’s about how profits are distributed. The AI pie is still growing, but the way it’s sliced is changing. Another issue is the delivery schedule. A $2.5 trillion backlog, annualized using fourth-quarter guidance, is equivalent to more than four and a half years of revenue. If delivery is delayed by even one quarter, plans to expand supporting capacity will have to be rescheduled. Domestic production lines supplying Nvidia lock in their production schedules quarterly; a one-quarter shift throws the entire year’s schedule off.
07 Watch Three Numbers Next
First, where fourth-quarter gross margin lands within the 71% to 72% range. Landing at the low end would mean cost pressures are still spreading; landing at the high end would mean Vera Rubin is ramping more smoothly than expected. This line matters far more than revenue.
Second, what share of the newly added backlog comes from non-investment-grade customers. A rising share would mean the credit quality of the order book is deteriorating.
Third, whether the permanent CEO is an insider or an outsider. An insider would be read by the market as “business as usual”: the roadmap stays intact, but the upside potential is discounted. An outsider would mean the roadmap needs to be reassessed, bringing greater short-term volatility.
What Nvidia needs to prove now isn’t whether AI demand exists—that no longer needs proving. It needs to prove that, with a new CEO and rising electricity prices and financing costs, this business can still sustain the kind of gross-margin structure it had over the past three years.
The earnings report proved the former; the market is waiting for the latter. What lies between the two isn’t a quarter, but a cycle.$NVDA
















