tapebrief
NVDA · Q4 2026

NVDA · Q1 2027 Earnings

Bullish

Nvidia

Reported May 20, 2026

30-second summary

Q1 revenue of $81.6B (+85% YoY, +20% QoQ) cleared the prior $79.6B guide high by $2.06B, with Data Center at $75.2B (+92% YoY) — accelerating again from +75% last quarter — and non-GAAP gross margin landing at exactly the 75.0% guide midpoint. Q2 FY2027 is guided to $91.0B (±2%), implying approximately +91% to +99% YoY, a $9.4B sequential dollar step on top of this quarter's $13.5B step. The bigger story is structural: management blew up the old reporting framework, splitting Data Center into Hyperscale ($38B) and ACIE ($37B, +31% QoQ vs Hyperscale's +12%), disclosed ~$20B in standalone Vera CPU revenue visibility for the year against a new $200B TAM, and reframed the trillion-dollar Blackwell+Rubin forecast as a baseline against $3–4T annual AI infrastructure spend by decade-end. Capital return stepped up materially: $80B incremental buyback authorization and a 25x dividend hike from $0.01 to $0.25 per share.

Headline numbers

EPS

Q1 FY2027

$1.87

Revenue

Q1 FY2027

$81.61B

+85.0% YoY

Gross margin

Q1 FY2027

74.9%

Free cash flow

Q1 FY2027

$48.55B

Operating margin

Q1 FY2027

65.6%

Key financials

Q1 FY2027
MetricQ1 FY2027YoYQ4 FY2026QoQ
Revenue$81.61B+85.0%$68.13B+19.8%
EPS$1.87—$1.62+15.4%
Gross margin74.9%—75.2%-30bps
Operating margin65.6%—67.7%-210bps
Free cash flow$48.55B—$34.90B+39.1%

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

MetricPeriodPrior guideActualΔResult
RevenueQ1 FY2027$78.0B +/- 2% ($76.44B - $79.56B)$81.615B+$2.06B above high end of guideBeat
GAAP Gross MarginQ1 FY202774.9% +/- 50 bps74.9%in-lineMet
Non-GAAP Gross MarginQ1 FY202775.0% +/- 50 bps75.0%in-lineMet
GAAP Operating ExpensesQ1 FY2027approximately $7.7BNot explicitly reported—Met
Non-GAAP Operating ExpensesQ1 FY2027approximately $7.5BNot explicitly reported—Met

New guidance

MetricPeriodGuideYoY
Operating Expense Growth YoYFY 2027upper 40s—
RevenueQ2 FY2027$91.0B +/- 2% ($89.18B - $92.82B)+94.9% to +98.5% YoY
GAAP Gross MarginQ2 FY202774.9% +/- 50 bps—
Non-GAAP Gross MarginQ2 FY202775.0% +/- 50 bps—
GAAP Operating ExpensesQ2 FY2027——

Changes to prior guidance

MetricPeriodPrior guideNew guideΔResult
GAAP and Non-GAAP Tax Rate
FY 2027
17.0% - 19.0%16.0% - 18.0%-100 bps at both low and high endLowered

Segment performance

Q1 FY2027
SegmentQ1 FY2027YoY
Data Center$75.2B+92.0%
Edge Computing$6.4B+29.0%
Data Center Compute$60.4B+77.0%
Data Center Networking$14.8B+199.0%
Data Center YoY Growth92%—
Data Center Compute YoY Growth77%—
Data Center Networking YoY Growth199%—
Edge Computing YoY Growth29%—

Profitability

Q1 FY2027
SegmentQ1 FY2027
Non-GAAP Gross Margin75.0%
Non-GAAP Operating Margin65.9%
Free Cash Flow$48.6B
FCF Margin59.5%

Management tone

Q2 FY2026 H20 overhang clearing → Q3 FY2026 capacity sold out, $500B is a floor → Q4 FY2026 agentic AI inflection arrives → Q1 FY2027 AI factory economics replace GPU unit pricing

Two quarters ago management was defending demand durability with TAM math; this quarter Jensen reframed the entire competitive narrative around AI factory unit economics. The verbatim anchor: "Customers do not buy GPUs. They build AI factories. And the right economic metric is not the purchase price of the GPU. It is the lifetime cost of an AI factory producing intelligence, token per watt, tokens per dollar, uptime, utilization, time to production, software durability, and asset life." This is the cleanest articulation yet of the moat argument — pricing power is no longer about GPU specs but about total-cost-of-ownership over an asset's life, which structurally insulates against ASIC commoditization in a way last quarter's full-stack-co-design defense did not.

The segmentation overhaul is the second multi-quarter shift. Three quarters ago Data Center was a single monolithic line item, with hyperscalers as the obvious dominant customer. This quarter management decomposed it into Hyperscale ($38B, ~5–7 customers) and ACIE ($37B, 250,000+ companies), and Jensen used the Ben Reitzes exchange to explicitly reconcile the implied growth gap: the hyperscaler CapEx CAGR that has worried investors does not constrain NVIDIA because the second category is far larger long-term and growing faster near-term. This shift signals management's confidence that the bull case no longer depends on hyperscaler capex sustainability — a structurally different durability argument than the one being defended three quarters ago.

Third, the Vera CPU disclosure is a discrete TAM expansion event, not an incremental product update. Jensen's verbatim framing in the Vivek Arya exchange: "$20 billion is for standalone CPU" — incremental to GPU revenue, not cannibalistic, because agentic workloads require both CPU orchestration and GPU inference. Two quarters ago CPU was not part of the NVIDIA growth story; this quarter it opens "a brand new 200 billion TAM for NVIDIA, a market we have never addressed before," with ~$20B of standalone CPU revenue visibility this year. The forward-leaning language is paired with concrete spec claims (1.5x perf per core, 2x perf per watt) rather than left as aspirational.

Fourth, supply confidence has hardened. Three quarters ago supply constraints were framed as an ongoing risk; this quarter Colette disclosed $145B of total supply, inventory, and purchase commitments, with Jensen citing this as the basis for "full confidence in the 1 trillion in Blackwell and Rubin revenue we foresee from 2025 through calendar 2027." The trillion-dollar number, which was new disclosure last quarter as an aspirational forecast, is now anchored to a hard committed-supply figure.

Finally, the China posture has fully decoupled from the growth narrative. Q2 of last year management was advocating for Blackwell China licenses; this quarter the guide assumes zero China Data Center compute revenue with no defensive framing, and even H200 approval was mentioned only to note "we are uncertain whether any imports will be allowed into the country." China is now optionality, not a base-case assumption — the cleanest version of this posture yet.

Recurring themes management leaned on this quarter:

Agentic AI inflection driving new workload categories and TAMsACIE (non-hyperscale AI infrastructure) as faster-growing segment than hyperscaleVera CPU entering 200B TAM with 1.5x performance per core, 2x perf per wattBlackwell ubiquity across all major model makers and hyperscalers with fastest ramp in company historyAI factory economics (token cost, ROI, asset life) replacing GPU unit pricing as value narrativeRubin production ramp commencing Q3 with 35x inference throughput vs Blackwell

Risks management surfaced:

U.S. government China export licenses uncertain—H-200 approved but no revenue generated and uncertain import allowanceSupply chain vulnerabilities despite $145B commitments and partnershipsConsumer edge demand softness due to higher memory and system pricesPower and capital constraints limiting AI factory buildout velocityTax environment changes potentially impacting effective rate assumptions

Q&A highlights

Joseph Moore · Morgan Stanley

What drove the change in segmentation? What's the philosophy behind the new reporting structure? Can you discuss competitive differences between segments and explain the surprising CPU growth trajectory?

Management explained the three-segment structure (hyperscale clouds, AI natives/enterprise/sovereign AI, and robotic edge) as reflecting AI's diversity across applications, deployment locations, and governance models. NVIDIA's unique vertically integrated but open platform approach serves each segment differently. The segmentation simplifies understanding of a now-massive, complex business spanning 250,000+ companies globally.

Three new segments: hyperscale clouds, AI natives/enterprise/sovereign AI, robotic edgeHyperscale represents ~5-7 companies; second segment represents 250,000+ companies globallyEach segment has different operating systems, go-to-market strategies, and technology stacksNVIDIA provides full-stack integrated solutions across all segments

Vivek Arya · Bank of America Securities

Is the CPU growth for agentic applications incremental or cannibalizing GPU demand? Is the $20 billion CPU number standalone or included in Vera Rubin bundles?

The $20 billion is standalone CPU revenue. CPUs are incremental, not cannibalizing—they power agent orchestration, tool use, and I/O, while GPUs handle inference/thinking. Agents require both: GPU for model inference and sub-agent spawning, CPU for orchestration and tool execution. With billions of future agents using tools, CPU demand is incremental and massive.

$20 billion figure is for standalone Vera CPU onlyVera designed specifically for agentic CPU workloads, not traditional cloud multi-core rentalsEconomics shifting from 'dollars per core' to 'tokens per dollar'Vera optimized for low latency and fast work completion, not core rentability

Ben Reitzes · Melius Research

Should NVIDIA be growing faster than hyperscaler CapEx growth? With hyperscaler CapEx forecast at 90-100% growth and data center guidance at 3-4% CAGR, how does this reconcile?

NVIDIA should and will grow faster than hyperscaler CapEx because its data center revenue comes from two large sources: (1) hyperscale CapEx (~$1T growing toward $3-4T by decade-end), and (2) the second category—AI natives, enterprise, sovereign clouds, industrial on-premises—which is far larger long-term (representing $50-100T of global economy) and growing faster. The second category is poorly understood but represents majority future growth.

Hyperscale CapEx currently ~$1T, expected to grow to $3-4T by end of decadeAI compute is now tied to revenue and profit generationSecond category (AI natives/enterprise/sovereign) growing at 'incredible pace'Second category represents hundreds of thousands of future companies vs. 5-7 hyperscalers

Timothy Arcuri · UBS

What traction are you getting with custom merchant solutions like CPX and LPX? If fast inference is 20% of the market, shouldn't LPX be gaining significant traction?

LPX is niche by design—optimized for low latency/high token rate but limited throughput, model capacity, and context processing. Use cases are narrow: providers with large premium token portfolios but small customer counts. LPX is intentionally SRAM-based and will remain niche 'for some time to come.' Grace Blackwell and Vera Rubin support the full AI lifecycle and are the primary platforms. LPX is a tack-on supplement for specific high-token-rate services, not a primary growth driver.

LPX throughput is intentionally lowLPX model size capacity is limitedLPX context processing ability is lowerLPX use cases are narrow and premium-focused

Stacy Raskin · Bernstein Research

Where do AI native clouds fit in the segmentation? Are they in hyperscale or AI cloud segment? Will the second segment (AI natives/enterprise/sovereign) grow faster than hyperscale long-term?

AI native clouds are in the second category (AI natives/enterprise/sovereign). They don't build/design chips, need integrated full-stack solutions, and benefit from NVIDIA's most rentable architecture. Hyperscale developed AI first (better CS talent, data center capability, consumer focus tolerance). Long-term, the second category will be much larger because it represents $50-100T of industrial/enterprise economy. Near-term, both grow incredibly fast, but second category expected to grow faster. Within 5 years, physical AI/robotic segment expected to accelerate dramatically.

AI native clouds classified in second segment (AI natives/enterprise/sovereign)AI natives lack patience for assembly of unrelated parts; need integrated architectureNVIDIA's architecture is 'most rentable of any computing platform in the world'Hyperscale developed AI first due to CS talent and consumer application tolerance

Answers to last quarter's watch list

Whether Q1 FY2027 revenue clears $79.6B (high end of guide). Q1 revenue of $81.6B beat the high end by $2.06B (+2.6% above the high), a smaller proportional beat than Q4's 2.8% and Q3's 3.6% — the first beat-magnitude deceleration in four quarters. Still a clean beat, but the trend bears watching as comps stiffen.
Resolved positively
Any disclosure of China Data Center compute revenue. Management explicitly disclosed that no China Data Center compute revenue was generated in Q1 and none is assumed in the Q2 outlook. H200 was approved but the company stated it is "uncertain whether any imports will be allowed into the country." China remains pure optionality.
Resolved negatively
Underlying ex-SBC non-GAAP OpEx trajectory. Non-GAAP OpEx of $7.45B in Q1 was up 12% sequentially. The Q2 non-GAAP OpEx guide of ~$8.3B is +11% QoQ vs the Q1 guide of $7.5B, and the new FY27 "upper 40s" YoY OpEx growth disclosure confirms a meaningful investment step-up. Operating margin still expanded ~60 bps QoQ to 65.9% as revenue growth outpaced OpEx growth.
Continue monitoring
Rubin ramp quantification. Management confirmed Vera Rubin commences production shipments in H2 CY2026, starting in Q3, with 35x inference throughput versus Blackwell. No specific Q3 unit or dollar figure was disclosed, but the $1T Blackwell + Rubin revenue forecast through CY2027 is now backed by the $145B supply commitment.
Continue monitoring
Networking disclosure becoming a reported segment. Resolved decisively — Data Center Networking was disclosed as a discrete reporting line at $14.8B in Q1, +199% YoY. This is the transparency step the prior brief flagged and validates Jensen's "largest Ethernet networking company in the world" claim from last quarter.
Resolved positively
Automotive and Robotics re-acceleration as the Physical AI signal. The old Automotive and Robotics segment was folded into a new Edge Computing line at $6.4B, +29% YoY — making direct YoY comparison to Q4's +6% impossible. Edge Computing is significantly larger than the old Automotive line, but the segmentation change obscures whether the underlying Automotive/Robotics business specifically re-accelerated.
Not resolved

What to watch into next quarter

Whether Q2 FY2027 revenue clears $92.82B (high end of guide). A repeat ~2.6% high-end beat implies ~$95B and continued acceleration; a guide-midpoint print would be the second consecutive beat-magnitude deceleration and the first real signal that comps are catching up.

ACIE vs Hyperscale growth gap sustaining. ACIE grew +31% QoQ in Q1 versus Hyperscale's +12%. If ACIE growth converges toward Hyperscale in Q2, the structural reframing of the bull case loses force; if the gap widens, Jensen's "second segment will be larger" thesis gains material support.

Non-GAAP OpEx landing at or below $8.3B with operating margin holding above 65%. The "upper 40s" FY27 OpEx growth guide is the most aggressive investment commitment in years; watch whether Rubin-prep spending is contained to the guided range and whether the 65.9% operating margin holds or drifts.

Any first-time Rubin revenue contribution disclosure. Production shipments commence in Q3 (calendar H2 2026). The Q2 print and call will be the last reporting period before that begins; any quantitative signal on Rubin's revenue contribution to Q3 guidance would be a material disclosure.

China Data Center compute revenue actualization vs zero baseline. Both Q1 actual and Q2 guide assume zero. Any licensed shipments through the H200 path remain pure upside to the $91B Q2 guide.

Vera CPU concrete revenue or design-win disclosure. ~$20B of standalone CPU revenue visibility for the year was disclosed against the new $200B TAM. First quarterly revenue anchor or named hyperscaler design wins would validate the trajectory.

Sources

  1. NVIDIA Q1 FY2027 Press Release, filed with SEC 2026-05-20: https://www.sec.gov/Archives/edgar/data/1045810/000104581026000051/q1fy27pr.htm
  2. NVIDIA Q1 FY2027 earnings call (transcript excerpts via tone and Q&A analysis).
  3. NVIDIA Q4 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.