Date: 2026-02-22 Quarter: Q3 FY26 (Oct-25), reported 2025-11-19 Atlas baseline: NVDA_earnings-review_Q3_FY26.md — read and incorporated
THAT IS SUCH A PREPOSTEROUS RESULT! Really, really, crazy numbers!
$57 billion in a single quarter. Up 62.5% year over year, with a sequential jump of $10.3 billion — the largest quarterly revenue add in NVIDIA's history. And then they guided $65 billion for next quarter, implying yet another acceleration to 65% YoY. The bear thesis — that NVIDIA was decelerating into irrelevance — was just eviscerated by the numbers.
This is not a company I want to sell. This is a company doing things I've never seen before in my investing career.
Thesis: Strengthening. Position: Hold/Add on weakness.
Let me be absolutely clear about what just happened with this revenue trajectory. The bears had a story: deceleration. And they were right for a while! Look at this:
| | Q2 FY24 | Q3 FY24 | Q4 FY24 | Q1 FY25 | Q2 FY25 | Q3 FY25 | Q4 FY25 | Q1 FY26 | Q2 FY26 | Q3 FY26 | | | Jul-23 | Oct-23 | Jan-24 | Apr-24 | Jul-24 | Oct-24 | Jan-25 | Apr-25 | Jul-25 | Oct-25 | |---|---|---|---|---|---|---|---|---|---|---| | Revenue ($m) | 13,507 | 18,120 | 22,103 | 26,044 | 30,040 | 35,082 | 39,331 | 44,062 | 46,743 | **57,006** | | YoY % | 101.5% | 205.5% | 265.3% | 262.1% | 122.4% | 93.6% | 77.9% | 69.2% | 55.6% | **62.5%** | | QoQ % | 87.8% | 34.2% | 22.0% | 17.8% | 15.3% | 16.8% | 12.1% | 12.0% | 6.1% | **22.0%** | | GM [NG] | — | — | — | — | — | 75.0% | — | — | 72.7% | **73.6%** | | FCF ($m) | 6,048 | 7,042 | 11,246 | 14,936 | 13,483 | 16,787 | 15,519 | 26,135 | 13,450 | 22,089 |
See what happened? YoY growth went from 265% at the peak down to 55.6% in Q2 FY26. The bears said: "This is over. Deceleration." But what does it mean when a company drops from 265% to 56%? It means they were lapping the most extraordinary ramp in semiconductor history. And now? Growth has reaccelerated. Two consecutive quarters of YoY acceleration. The bear narrative is wrong.
The QoQ number is even more striking. Q2 FY26 was +6.1% QoQ — the slowest sequential growth of the AI era. Bears said demand was plateauing. Then Q3 comes in at +22.0% QoQ, the largest absolute dollar sequential add in the company's history. $10.3 billion in one quarter! Tell me again the bears were right!
$51.2 billion in Data Center revenue. Up 66% YoY. Up 25% QoQ. This is 89.8% of total revenue.
The mix breakdown is extraordinary:
That networking number deserves its own analysis. A year ago, networking was $3.1B. Now it's $8.2B. NVLink scale-up plus Spectrum X Ethernet plus Quantum X InfiniBand. NVIDIA didn't just build a GPU company — they're building the entire AI factory infrastructure. The majority of AI deployments now include NVIDIA switches. This is a second growth engine that didn't exist 18 months ago!
GB300 has crossed over GB200 — now roughly two-thirds of Blackwell revenue. The transition has been seamless. This is management execution at the highest level.
Colette Kress said this on the call: they have visibility to a half-trillion dollars in Blackwell and Rubin revenue from the start of 2025 through end of calendar 2026. HALF A TRILLION DOLLARS!
I follow the money. $500 billion in revenue visibility is unprecedented for a hardware company. This is the durability of revenue that I care about — whether it's SaaS subscription revenue or two years of locked-in AI infrastructure commitments. What matters is: is this growth durable? With these leading indicators, the answer is yes:
These are not lagging indicators. These are forward commitments. The clouds are sold out. The GPU installed base is fully utilized — Blackwell, Hopper, AND Ampere. This is not a company running out of customers.
This was my biggest concern earlier in the year. Q1 FY26 had a 60.5% GAAP gross margin — a shocking drop from 78.4% in Q1 FY25. Management blamed the Blackwell transition. I was watching carefully.
They delivered:
Operating margin of 66.2% [Non-GAAP]. Net margin of 56% [GAAP]. FCF of $22.1 billion in a single quarter. The profitability of this business at this scale is something I've never seen in my investing career.
I always read these transcripts myself. Every word. Jensen Huang's remarks were extraordinary.
Three simultaneous platform shifts:
And NVIDIA is the singular platform for all three. That's the moat. Not just the chips — CUDA is the software ecosystem that runs every major AI model. Jensen made this explicit: if you want to run OpenAI, Anthropic, xAI, Gemini, Thinking Machines — you run on NVIDIA.
The ASIC question: I've been watching this closely. Management's answer is sophisticated. ASICs face five challenges that NVIDIA doesn't: (1) they must accelerate every phase of the transition, not just inference; (2) they must handle every phase of AI — pre-training, post-training, inference; (3) CUDA runs every AI model architecture; (4) NVIDIA is in every form factor; (5) NVIDIA's ecosystem creates customer demand. As models evolve, custom ASICs become obsolete. This is a real risk, but the 2027-2028 timeline for meaningful impact still seems right.
The biggest competitive moat signal? Jensen said: "More customers coming to us after exploring others — not decreasing." They're winning the trials.
Atlas gave a 4/5 conviction and called this "objectively reasonable" on valuation. I agree on the verdict but want to be more direct about two things:
On the positives: That networking number (+162% YoY, $8.2B) is being underappreciated. NVIDIA is becoming the AI equivalent of Cisco in the internet era — the picks-and-shovels provider for the entire AI factory, not just the GPU. When networking becomes as large as Gaming, the TAM story gets much bigger. This is a company that could be worth more in five years than anyone currently models.
On the risks: I take the ASIC risk more seriously than Atlas's 2027-2028 framing might suggest. OpenAI + Broadcom. Meta MTIA T1. Google TPUs handling 75%+ of Gemini inference. The roster of custom chip developers is expanding fast. This is not going to hit revenues in 2026 — but I want to watch the inference mix on every call. If NVIDIA's share of inference workloads declines while training holds, that's an early warning signal.
On valuation: I don't anchor to price targets. What I know is: at 25x forward P/E for a company growing 65% YoY with 73% gross margins and $22B per quarter in free cash flow, the math works. This is not expensive for what you're getting. My framework is simple: if the business keeps executing, the stock follows. NVIDIA keeps executing.
For Q4 FY26 (reporting ~Feb 26, 2026, i.e., next few days!):
Q1 FY27 guidance. Street expects ~$75B. Anything above signals continued acceleration. This is the most important number on the call.
Gross margin. 75% Non-GAAP guided for Q4. If they beat this and guide Q1 FY27 margins at 75%+, the FY27 margin concern evaporates.
Jensen's comments on ASIC competition. OpenAI-Broadcom is new since Q3. The answer needs to evolve.
Networking trajectory. $8.2B in Q3. Does it continue accelerating or plateau? This is my leading indicator for NVIDIA's TAM expansion story.
Rubin timeline. H2 CY2026 ramp confirmed with silicon back from fabs. Any pull-forward would be hugely positive.
| Promise | Made | Status |
|---|---|---|
| Blackwell ramp in H2 FY26 | Q2 FY26 call | ✅ Delivered — beyond expectations |
| GM recovery ≥73% post-Blackwell transition | Q1 FY26 | ✅ Delivered — 73.6% Q3, 75% guided Q4 |
| $500B Blackwell+Rubin visibility through CY2026 | Q3 FY26 | 🔲 On track |
| Rubin ramp H2 CY2026 | Q3 FY26 | 🔲 On track — silicon back from fabs |
| Mid-seventies GM in FY27 | Q3 FY26 | 🔲 Watch — "input costs rising" caveat |
Management has been excellent at delivering on promises. No miss-and-lower patterns. No excuses on macro. Jensen and Kress have earned trust over multiple cycles.
I've been following NVIDIA for years. I held it at nearly 18% of my portfolio in 2024. This Q3 FY26 report is one of the most extraordinary earnings reports I have ever read. A company doing $57 billion in a quarter, growing 62.5% year over year, with 73.4% gross margins and $22 billion in free cash flow — and guiding to $65 billion next quarter with further acceleration?
The numbers speak for themselves. The growth is durable — $500 billion in visibility says so. The margins are expanding. The second growth engine in networking (+162% YoY!) validates the platform story. And Rubin is coming in 2026.
Should we run away from this company??? Absolutely not!
This is the kind of company I want to own at a significant weight. The bear narrative of ASIC displacement and CapEx cyclicality is real — I take those risks seriously — but the timeline is 2027-2028, not now. Right now, the business is exceptional.
Thesis: Strengthening. Action: Hold/Add on weakness.
Saul Rosenthal | 2026-02-22 | Q3 FY26 Earnings Review
Best, Saul