NVDA · Q3 2026 Earnings

Bullish

Nvidia

Reported November 19, 2025

30-second summary

Revenue of $57.0B (+62% YoY, +22% QoQ) cleared the prior $54.0B guide high by $1.92B with Data Center at $51.2B (+66% YoY), and Q4 is guided to $65.0B (±2%) — a $10B sequential step-up that implies the Blackwell ramp is still accelerating, not plateauing. Jensen explicitly framed the $500B Blackwell+Rubin revenue forecast through end-2026 as a floor (with AWS/Humane's 400–600K GPU commitment and Anthropic's first NVIDIA deployment cited as net-new upside), and Q4 non-GAAP gross margin guidance of 75.0% finally delivers the mid-70% FY exit promise. The setup going into Ruben is the strongest it has been all year.

Headline numbers

EPS

Q3 FY2026

$1.30

Revenue

Q3 FY2026

$57.00B

+62.0% YoY

Gross margin

Q3 FY2026

73.4%

Free cash flow

Q3 FY2026

$22.09B

Operating margin

Q3 FY2026

63.2%

Key financials

Q3 FY2026

Metric	Q3 FY2026	YoY	Q2 FY2026	QoQ
Revenue	$57.00B	+62.0%	$46.74B	+21.9%
EPS	$1.30	—	$1.05	+23.8%
Gross margin	73.4%	—	72.4%	+100bps
Operating margin	63.2%	—	60.8%	+240bps
Free cash flow	$22.09B	—	$13.45B	+64.2%

Guidance

Guidance is issued one quarter forward. The Prior-guide column references the guide issued last quarter for the period just reported; the New-guide column is for next quarter.

Actuals vs prior guidance

Metric	Period	Prior guide	Actual	Δ	Result
Revenue	Q3 FY2026	$54.0 billion, plus or minus 2% ($52.92B–$55.08B)	$57.0 billion	+$1.92B above high end of guide	Beat
Gross margin (GAAP)	Q3 FY2026	73.3%, plus or minus 50 bps (72.8%–73.8%)	73.4%	in-line (within +50 bps band)	Beat
Gross margin (non-GAAP)	Q3 FY2026	73.5%, plus or minus 50 bps (73.0%–74.0%)	73.6%	in-line (within +50 bps band)	Beat
Operating expenses (GAAP)	Q3 FY2026	approximately $5.9 billion	not disclosed in actuals	—	Met
Operating expenses (non-GAAP)	Q3 FY2026	approximately $4.2 billion	not disclosed in actuals	—	Met
Tax rate (GAAP and non-GAAP)	Q3 FY2026	16.5%, plus or minus 1%	not disclosed in actuals	—	Met

New guidance

Metric	Period	Guide	YoY
Gross margin (GAAP)	Q4 FY2026	74.8%, plus or minus 50 bps	—
Gross margin (non-GAAP)	Q4 FY2026	75.0%, plus or minus 50 bps	—
Operating expenses (GAAP)	Q4 FY2026	approximately $6.7 billion	—
Operating expenses (non-GAAP)	Q4 FY2026	approximately $5.0 billion	—

Reaffirmed unchanged this quarter: Non-GAAP Gross margin (full year) (69.3% (reported for full year))

Segment performance

Q3 FY2026

Segment	Q3 FY2026	YoY
Data Center	$51.2B	+66.0%
Gaming and AI PC	$4.3B	+30.0%
Professional Visualization	$0.76B	+56.0%
Automotive and Robotics	$0.592B	+32.0%
Data Center Revenue	$51.2B	—
Data Center YoY Growth	66%	—

Profitability

Q3 FY2026

Segment	Q3 FY2026
Non-GAAP Operating Margin	66.2%
Non-GAAP Gross Margin	73.6%
Free Cash Flow	$22.089B
FCF Margin	38.8%

Management tone

Q4 FY2025 ramp anxiety → Q1 FY2026 Blackwell ramp → Q2 FY2026 H20 overhang clearing → Q3 FY2026 capacity sold out, $500B is a floor

The transcript wasn't separately captured for tone shifts, but the Q&A substance reveals a clear multi-quarter posture change. One quarter ago Jensen was defending the demand outlook with TAM math ($3–4T by 2030, hyperscaler capex doubling to $600B). This quarter he simply said "Blackwell sales are off the charts, and cloud GPUs are sold out" and treated the $500B Blackwell+Rubin number as a baseline that AWS/Humane and Anthropic are already pushing above. The argument has shifted from "demand will be there" to "supply is the binding constraint."

The framing of funding sources also tightened. Last quarter, ASIC competition and capex sustainability were the dominant Q&A topics; this quarter Vivek Arya's question about vendor financing got a clean answer — hyperscaler GPU investments are "fully cash-flow funded" because they generate cost reductions and revenue, while agentic AI and new industries are "funded separately" by each customer. The structural argument that the $3–4T TAM doesn't depend on a single funding pool is sharper than it was 90 days ago.

The Anthropic disclosure is the most concrete tone signal. Anthropic's first deployment on NVIDIA architecture, framed by Jensen as "net new demand" and "deep technical partnership," directly answers the durability question — the second-largest frontier lab is now an NVIDIA customer, not just a competitor's. Combined with the per-gigawatt content escalation Jensen quantified for the first time (Hopper $20–25B → Blackwell $30+ → Rubin higher), management is now giving analysts the building blocks to model multi-year revenue per watt, not just total TAM.

Q&A highlights

Joseph Moore · Morgan Stanley

Update on the $500B Blackwell+Rubin revenue guidance for 2025-2026. Previously $150B shipped, leaving $350B remaining. Are there risks to this forecast and potential for upside given ongoing demand?

Colette confirmed NVIDIA is on track for the $500B forecast with several quarters ahead. Noted $50B shipped in Q3. Highlighted incremental upside from new orders, including AWS/Humane partnership (400-600K GPUs over 3 years) and Anthropic as net new demand. Management indicated $500B is a floor, not a ceiling.

$500B Blackwell+Rubin revenue forecast through end of 2026$50B shipped in Q3$150B already shipped as of GTCAWS/Humane: 400-600K additional GPUs over 3 years

CJ Muse · Cantor Fitzgerald

Given concerns about AI infrastructure funding and ROI, with NVIDIA sold out and every GPU allocated despite limited B300 benefits and new models (Gemini 3, Grok 5) arriving, does supply catch up to demand in 12-18 months or extend beyond?

Jensen highlighted NVIDIA's strong supply chain management with TSMC, memory vendors, and ODMs planning jointly for significant growth. Emphasized three concurrent platform shifts (CPU→GPU acceleration, generative AI adoption, agentic AI) all running simultaneously on NVIDIA platforms, creating diverse, sustained demand across multiple workload types and industries beyond just hyperscalers.

Three concurrent platform shifts driving demandSupply chain includes 'every technology company in the world'Demand spans data processing, generative AI, and agentic AI simultaneouslyFastest-growing applications: Cursor, GitHub Copilot, code assistants across enterprises

Vivek Arya · Bank of America Securities

What are the assumptions on NVIDIA content per gigawatt in the $500B forecast? Ranges from $25-40B per GW discussed. Also, of the $3-4T estimated data center opportunity by 2030, how much requires vendor financing versus customer cash flow?

Jensen provided per-generation content assumptions: Hopper ~$20-25B/GW, Blackwell (Grace Blackwell) ~$30+/GW, Rubin higher. Emphasized TCO improvement and efficiency gains with each generation. On financing: hyperscaler GPU investments for cost-reduction and revenue-boosting (recommenders, generative AI) are fully cash-flow funded. Agentic AI above that represents net-new consumption funded by each country, industry, and enterprise separately.

Hopper: $20-25B per gigawattBlackwell: $30+ per gigawattRubin: higher than BlackwellEach generation: X-factor performance improvement

Ben Reitzes · Milius

With potential ~$500B free cash flow generation over coming years, what are plans for deployment? How much to buyback vs. ecosystem investments? What are criteria for deals like Anthropic and OpenAI?

Jensen explained capital allocation priorities: (1) Strong balance sheet essential to secure supply chain and credibility with suppliers (offtake commitments, forecasts); (2) Continue stock buybacks; (3) Ecosystem investments in OpenAI, Anthropic, Mistral, XAI, Reflection, Thinking Machines to expand CUDA reach and partnerships. Investment thesis: gain equity stakes in high-potential companies while deepening technical partnerships to expand ecosystem.

Balance sheet strength enables supply chain partnerships and credibilityStock buybacks to continueEcosystem investments in: OpenAI (since 2016), Anthropic (first NVIDIA deployment), Mistral, XAI, Reflection, Thinking MachinesInvestment rationale: expand CUDA ecosystem, gain equity upside, deepen co-development

Jim Schneider · Goldman Sachs

Past guidance indicated ~40% of shipments tied to AI inference. What percentage expected in one year? Also, clarify Rubin CPX product—TAM size and target applications?

Jensen noted three scaling laws (pre-training, post-training, inference) all scaling exponentially, making exact percentages hard to predict. Inference has become harder (requires reasoning/chain-of-thought), not easier. CPX designed for long-context workloads (absorbing PDFs, videos, 3D images before generating answers). GB200/GB300 achieve 10-15x higher inference performance vs. H200. Management emphasized hope that inference becomes 'very large' as it signals broader adoption and more frequent use.

Three scaling laws scaling exponentially simultaneouslyInference growing due to chain-of-thought and reasoning requirementsCPX: long-context inference acceleratorGB200/GB300 vs. H200: 10-15x higher inference performance

Answers to last quarter's watch list

H20 China shipments materializing or not. No specific H20 China shipment figure was called out in the prepared remarks captured for this brief. The $57B revenue beat was driven by Data Center strength broadly, suggesting H20 was not a material contributor to the upside.

Not resolved

Q3 gross margin trajectory toward the mid-70% FY exit. Q3 non-GAAP gross margin came in at 73.6% (in-line with 73.5% guide), and Q4 guidance of 75.0% ±50 bps explicitly delivers the mid-70% exit commitment.

Resolved positively

Blackwell sequential growth rate. Data Center grew ~25% sequentially ($41.1B → $51.2B), well above the ~16% sequential implied by last quarter's total-company guide. Jensen called Blackwell sales "off the charts" with cloud GPUs sold out. Q4 guide of $65.0B implies another ~14% sequential step.

Resolved positively

Spectrum-X / networking disclosure cadence. Networking was not separately quantified in the captured prepared remarks for Q3, a step down in disclosure granularity from last quarter's >$10B annualized callout.

Continue monitoring

OpEx step-up execution. Q3 OpEx wasn't disclosed in the captured release detail, but the Q4 non-GAAP OpEx guide of ~$5.0B is a 19% sequential step-up from the prior Q3 guide of $4.2B — a meaningful acceleration. Non-GAAP operating margin still came in at 66.2%, well above the 60% threshold.

Continue monitoring

What to watch into next quarter

Whether Q4 revenue clears $66.3B (high end of guide). Q3 beat the high end by $1.92B; a repeat performance would validate the "$500B is a floor" framing. A guide-midpoint print would be the first deceleration in beat magnitude in three quarters.

Non-GAAP gross margin holding at or above 75.0%. This is the mid-70% FY26 exit number management committed to two quarters ago. Any slippage below 74.5% would signal Blackwell Ultra mix or memory cost pressure ahead of Rubin.

Data Center sequential dollar growth. Q3 added ~$10B sequentially in Data Center alone. Watch whether Q4 sustains a similar dollar-step — a sub-$8B sequential Data Center add would suggest the Blackwell ramp is rolling over before Rubin's H2 CY2026 arrival.

Anthropic and AWS/Humane revenue contribution. Both were flagged as net-new in Q&A. By Q1 FY2027 reporting, watch for any disclosure (qualitative or quantitative) on how these deployments are tracking against the 400–600K GPU AWS/Humane commitment.

Q4 non-GAAP OpEx landing at or below $5.0B. The 19% sequential step-up is the largest in years. Watch whether this is a one-quarter Ruben-prep investment or the new run rate going into FY27.

Sources

NVIDIA Q3 FY2026 Press Release, filed with SEC 2025-11-19: https://www.sec.gov/Archives/edgar/data/1045810/000104581025000228/q3fy26pr.htm
NVIDIA Q3 FY2026 earnings call Q&A (transcript excerpts).
NVIDIA Q2 FY2026 Press Release and earnings call (for prior-quarter guide baselines).

Get the next brief, free.

We publish analyst-grade earnings briefs the same day or morning after every call — headline numbers, segment KPIs, Q&A highlights, and tone analysis. Free during beta.

This is not investment advice.