Nvidia just posted the biggest quarter in its history — $115.8 billion in revenue, up 103% from a year ago — and it did it days after its co-founder and chief executive of more than three decades walked out the door. Jensen Huang resigned on November 11, 2026. His chief financial officer, Colette Kress, presented the results as interim CEO. The board says it has no intention of bringing Huang back. And the demand that built those numbers hasn’t slowed down at all.
That’s the strange shape of this earnings report: a company in the middle of a leadership crisis that is, by every financial measure, doing better than ever. The question for the rest of us — whether you hold the stock, build PCs, or just want a graphics card that doesn’t cost as much as a used car — is what happens next. The numbers answer part of that. The rest is still open.
The quarter in numbers
Nvidia reported third-quarter fiscal 2027 results on Tuesday, November 17, 2026, for the three months ended October 25. Our first results rundown covered the headline beat — this piece is about what the numbers mean now that the dust has settled, and what they cost you if you’re buying hardware. Here’s the headline table:
| Measure | Q3 FY2027 | Q2 FY2027 | Q3 FY2026 (a year ago) |
|---|---|---|---|
| Total revenue | $115.8 billion | $96.2 billion | $57.0 billion |
| Data Center revenue | $107.6 billion | $89.0 billion | $51.2 billion |
| Adjusted EPS | $2.68 | — | — |
| GAAP EPS | $2.70 | — | — |
| Net income | $65.3 billion | — | — |
A few things in that table deserve a second look. First, the growth: revenue more than doubled year over year (+103%) and rose 20% from the previous quarter. Data Center — the division that sells the chips powering AI training and inference — grew even faster at 110% year over year and now accounts for 93% of everything Nvidia sells. Gaming, professional visualization, and automotive combined make up the remaining 7%. Nvidia is, for practical purposes, a data center company that also sells graphics cards.
Second, the beat: adjusted earnings per share of $2.68 came in ahead of the roughly $2.47 Wall Street expected, and revenue beat a consensus of about $109.1 billion as well as Nvidia’s own guidance of $108.0 billion. The company didn’t just clear the bar — it cleared it by nearly $7 billion.
Third, the margin story is starting to turn. Nvidia guided fourth-quarter gross margin to 71% to 72%, a step down from recent levels. That’s still an extraordinary margin by any normal standard — most hardware companies would kill for 40% — but the direction matters. The new Vera Rubin systems now shipping are expensive to build, and the mix is shifting toward massive rack-scale systems that carry lower margins than standalone chips.
What happened to Jensen Huang
You can’t write about this quarter without the resignation, because it is the first earnings report in nearly 28 years — since Nvidia’s IPO — without Huang at the helm. Here’s what is confirmed and what is reported, kept separate because the distinction matters.
Confirmed: Huang resigned as president, CEO, and director, with the resignation announced on November 12, 2026. We covered the resignation itself the day it was announced. Nvidia’s board unanimously accepted it. Colette Kress, Nvidia’s chief financial officer for 13 years, is serving as interim CEO while also keeping the CFO role. The board has hired executive search firm Heidrick & Struggles to find a permanent successor, looking inside and outside the company. On the earnings call, board chair Harvey Jones said the board “does not intend to reinstate” Huang as CEO — and said it hours after Huang offered to return. The earnings release itself carried no mention of Huang at all.
Reported, not confirmed by the company: Multiple outlets reported that Huang admitted to an extramarital affair and that his departure followed the disclosure of personal conduct that violated the company’s code of conduct. Some reports have gone further, alleging connections to a person described as a Chinese national who may have funneled restricted chips to China. The board has hired law firm Wachtell, Lipton to investigate potential national security issues, and Huang is scheduled to be interviewed behind closed doors by House committees on December 1. Treat the most serious allegations as allegations until an investigation or a public record says otherwise — they are reported, not established.
What is established is the board’s stance: it is not considering his return, and it is running a full search for a successor. For a company built so visibly around one founder — Huang’s leather jacket and kitchen-table origin story were part of Nvidia’s brand — that is a genuine inflection point, and the market’s calm reaction to it is itself a story. Shares rose in late trading after the report. Investors, at least for now, are betting the machine runs without its founder.
The guidance: $130 billion, with a catch
Nvidia guided fourth-quarter revenue to about $130 billion, plus or minus 3% — a range of $127.4 billion to $132.6 billion — above the roughly $125.3 billion Wall Street expected. It also raised its fiscal 2028 revenue growth outlook to about 75%, up from the 70% it had forecast earlier.
But read the fine print before you celebrate. The fourth quarter has 14 weeks, not 13. That extra week changes the math in a way most headlines skip:
| Period | Revenue | Weeks | Revenue per week |
|---|---|---|---|
| Q3 FY2027 | $115.8 billion | 13 | ~$8.91 billion |
| Q4 FY2027 (guided) | ~$130 billion | 14 | ~$9.29 billion |
On a per-week basis, the guided growth is about 4%, not the 12% the headline comparison suggests. That’s still growth — and still a record — but it’s the difference between a company accelerating and a company cruising at an already absurd speed. If you’re trying to judge whether AI demand is still speeding up or has found its cruising altitude, the per-week number is the honest one.
Two more details in the guidance matter. First, it assumes zero Data Center compute revenue from China. Nvidia has stopped Hopper chip shipments to China, and the forecast doesn’t count on that market coming back. Second, the gross margin guide of 71% to 72% confirms the margin step-down is real and planned, not a one-quarter blip. Vera Rubin NVL72 rack systems have begun shipping, and rack-scale systems simply cost more to build than the chips alone.
The leadership question the numbers don’t answer
Here’s what I keep coming back to: this quarter proves Nvidia’s demand engine runs on contracts, roadmaps, and installed infrastructure — not on one person’s presence in the room. Data Center revenue of $107.6 billion doesn’t happen because a charismatic CEO gave a good keynote. It happens because hyperscalers have multi-year build-outs, because the Vera Rubin ramp was planned years ago, and because customers can’t get these chips anywhere else at scale.
But there are three things the numbers can’t tell us yet, and they’re the things that will decide whether this quarter is a high-water mark or a waypoint:
1. Who runs the company next, and how fast? Kress is a respected operator — 13 years as CFO means she knows where every dollar goes — but she is interim, and the search is just starting. A permanent CEO who doesn’t understand the technology deeply, or who the engineering ranks don’t trust, is a real risk at a company where product decisions are the strategy. The board’s refusal to consider Huang’s return closes one door firmly; it doesn’t open another.
2. Does the China investigation widen? The Wachtell investigation and the December 1 House committee interview are early-stage. If they produce findings that restrict Nvidia’s ability to sell into more markets, or that implicate business practices rather than one person’s conduct, the zero-China assumption in guidance could look optimistic. Right now the market is pricing in a contained problem. Contained is an assumption, not a fact.
3. Can supply keep up with the guidance? Nvidia is guiding to $130 billion while telling you, in the same breath, that margins are compressing and that Vera Rubin systems are ramping. The company has to actually build and ship those systems. Its supply chain — TSMC’s advanced packaging, HBM memory, the power and cooling infrastructure for rack-scale deployments — is the constraint on the whole AI industry right now, and it’s the same constraint squeezing consumer GPUs. Which brings us to the part that affects you directly.
What this means if you’re buying a GPU
This is where the earnings report stops being a Wall Street story and becomes a personal finance story. The same AI demand that produced $107.6 billion in Data Center revenue is what’s making graphics cards unaffordable — and the earnings call confirmed the connection.
The flagship number: Newegg sold the ASUS ROG Astral RTX 5090 at $6,249.99 in its November 11–12 restock — 213% over the card’s $1,999 launch price — and it sold out. By November 15, Newegg listed 21 RTX 5090 models from $6,199.99 to $7,349.99, all more than three times MSRP, with a one-per-customer limit. TweakTown’s restock coverage has the full listing spread. The RTX 5080, launched at $999, is sitting near $2,199.99 at major retailers.
Why? The memory. A 5090 carries 32GB of GDDR7, and AI data centers now take roughly 70% of global memory output, up from 20% to 30% in 2022, according to IDC figures cited in recent coverage. Board partners report they cannot currently order GB202 GPUs and GDDR7 memory as a bundle. Nvidia says the card is not discontinued and shipments continue, and its own US store restocked the 5090 and 5080 on November 10 after roughly a month in notification-only status — but the cheapest 5090 there was $6,759.99, more than Newegg’s price. Stock exists. Affordable stock does not.
So what should you actually do? Here’s my honest read, as someone who has watched this market all year:
If you need a card now, buy the whole PC. Complete RTX 5090 desktops from HP, Alienware, and MSI were discounted to roughly the card’s standalone price through much of 2026, because OEMs buy GPUs at contract prices. That gap has narrowed sharply since October as OEM prices repriced, so check the math on the specific system — but a prebuilt is still usually the cheapest way to get a 5090 into your house right now, and you get a warranty on the whole machine.
If you can wait, wait — but don’t expect $1,999. Memory makers and analysts don’t expect real relief before 2027 or 2028, and some say 2030. The 5090 at triple MSRP is the market clearing price while AI labs compete with gamers for the same memory. A return to launch pricing would require either a memory supply surge or an AI demand collapse, and neither is on the near-term horizon. Waiting might save you a few hundred dollars as restocks continue; it won’t save you thousands.
If you run local AI, the 5090 is still the cheapest 32GB card. That’s the other side of the shortage: small labs, researchers, and cloud marketplaces buy 5090s because 32GB of very fast memory on one card has no cheaper substitute. You’re not just competing with gamers for this card. You’re competing with everyone who wants to run inference locally without paying data center prices. That demand doesn’t go away when the holiday season ends.
Don’t buy from a marketplace scalper at $9,400. Marketplace floors reached roughly $8,000 to $9,400 during the October drought. Retail restocks at $6,200 to $6,600 are still painful, but they’re real listings from real retailers with return policies. The gap between retail and marketplace prices has narrowed enough that the scalper premium buys you nothing but risk.
And one more connection worth making: the same shortage pushing DDR5 RAM kits to $459 for 32GB — up from $251 in January — is the same shortage in this earnings report. Nvidia’s Data Center business is the demand side of your RAM price. When Kress says AI demand is strong enough to raise next year’s growth forecast to 75%, she’s also telling you your next PC build will cost more than your last one. Read our DDR5 price breakdown for the full build math.
The bottom line
Nvidia’s first quarter without Jensen Huang is, financially, its best quarter ever. Revenue doubled, guidance rose, the Vera Rubin ramp is shipping, and the board has drawn a hard line on the past. The numbers say the company doesn’t need its founder to keep growing. The open questions — who leads it next, what the investigation finds, and whether supply can keep up with $130 billion quarters — are real, but none of them showed up in this quarter’s results.
For the rest of us, the earnings report is a reminder that the AI boom has a consumer price tag. Every billion in Data Center revenue is memory and silicon that isn’t going into a graphics card or a RAM kit. The 5090 at $6,249.99 isn’t a glitch in the market. It’s the market, working exactly as the earnings report describes it.
This article is not investment advice. Figures are from Nvidia’s Q3 FY2027 earnings release and contemporaneous reporting — see the full earnings rundown — and prices cited are dated retail readings that change frequently.
Frequently Asked Questions
How much did Nvidia earn in Q3 FY2027?
Nvidia reported revenue of $115.8 billion for the third quarter of fiscal 2027 (ended October 25, 2026), up 103% year over year and 20% sequentially. Data Center revenue was $107.6 billion, up 110% year over year. Adjusted earnings per share were $2.68, beating the roughly $2.47 analysts expected, and net income reached $65.3 billion.
Who is Nvidia’s CEO after Jensen Huang resigned?
Colette Kress, Nvidia’s chief financial officer for 13 years, is serving as interim CEO after Jensen Huang resigned on November 11, 2026 (announced November 12). The board has hired executive search firm Heidrick & Struggles to find a permanent successor and has said it does not intend to reinstate Huang.
Why did Jensen Huang resign from Nvidia?
Huang resigned following the disclosure of personal conduct that violated the company’s code of conduct, according to the board’s statement as reported. Multiple outlets reported he admitted to an extramarital affair, and some reports allege connections to restricted chip shipments to China — those more serious allegations are reported, not established. The board has hired law firm Wachtell, Lipton to investigate, and Huang is scheduled for a closed-door House committee interview on December 1.
What did Nvidia guide for Q4 FY2027?
Nvidia guided fourth-quarter revenue to about $130 billion, plus or minus 3%, above the roughly $125.3 billion consensus, with gross margin of 71% to 72%. The quarter has 14 weeks instead of 13, so per-week revenue growth is about 4%, not 12%. The guidance assumes zero Data Center compute revenue from China, and the company raised its fiscal 2028 revenue growth outlook to about 75%.
Why is the RTX 5090 so expensive in November 2026?
The RTX 5090 launched at $1,999 but sells for $6,199 to $7,349 at Newegg because of a memory shortage: AI data centers now take roughly 70% of global memory output, and the card’s 32GB of GDDR7 is scarce. Board partners reportedly cannot order the GB202 GPU and GDDR7 memory as a bundle, and local-AI users and small labs compete with gamers for the same cards. Nvidia says the card is not discontinued.
Should I buy an RTX 5090 now or wait?
If you need one now, a complete prebuilt PC is usually the cheapest route, since OEMs buy GPUs at contract prices — but check the math, because that gap narrowed sharply in October. If you can wait, restocks may shave a few hundred dollars off, but a return to the $1,999 launch price isn’t expected before 2027 or 2028. Don’t pay marketplace prices near $9,400 when retailers list the same cards at $6,200 to $6,600.
