Semiconductors Equity Research

NVDA

NVIDIA Corporation

Last Updated 2026-05-11
Data Source SEC EDGAR 10-K/10-Q + NVIDIA IR + GTC 2026 Materials

Research Note — This is editorial analysis based on public data. It does not constitute investment advice, a recommendation to buy or sell any security, or an offer to transact. sectally has no positions in NVDA. See full disclaimer.

NVDA · NVIDIA Corporation — AI Compute Infrastructure Dominance

Research Date: May 11, 2026 Market Cap: ~$5,229B Research Type: Phase 2 Formal — Fact-based draft with cross-verified public sources


Data Credibility & Verification Layer

This report is based on cross-verified public data sources:

Data Type Source Confidence
8Q quarterly financials NVIDIA IR press release + SEC EDGAR L1–L2
Segment revenue breakdown NVIDIA Q4 FY2026 earnings PR L2
FY27 Q1 guidance NVIDIA IR (published 2026-02-25) L2
GTC 2026 $1T demand forecast Jensen Huang keynote (CNBC/Fortune coverage) L2–L3
Competitive / custom silicon Multiple sources (CNBC/Motley Fool/tech-insider) L3

Limitations:

  • No FactSet/Bloomberg consensus subscription
  • SEC 10-K MD&A not directly accessed
  • Earnings call transcripts accessed via secondary summaries only
  • FY25 Q4 gross margin is a derived estimate (back-calculated from annual figures)

Key Takeaways

Thesis: NVIDIA is the dominant force in global AI compute infrastructure — Data Center revenue accounts for 91% of total, FY26 full-year revenue reached $215.9B (+65% YoY), TTM net income of $120.1B, and exceptionally strong free cash flow. Backlog visibility extending to 2027 across Blackwell and Vera Rubin platforms exceeds $1T. The CUDA software ecosystem, built over 20 years, constitutes a deep competitive moat that rivals (AMD ROCm / Google TPU / AWS Trainium) cannot replicate in the near term.

Coverage Status: Active · Last Updated May 11, 2026 Data Source: SEC EDGAR 10-K/10-Q + NVIDIA Investor Relations + GTC 2026 Materials

Scenario Analysis (Educational Illustration Only):

  • Bear Case Revenue Assumption: Forward PE 25x — assumes AI CapEx cycle peak + full China export restriction
  • Base Case Revenue Assumption: Forward PE 31x — FY27 consensus realization at $370B revenue / EPS $8.34
  • Bull Case Revenue Assumption: Forward PE 42x — inference inflection exceeds expectations + Vera Rubin early ramp

Note: These are arithmetic scenarios derived from publicly disclosed guidance ranges and consensus estimates, not price forecasts or investment recommendations.

Key Risks:

  1. Customer custom silicon (Google TPU v8 / AWS Trainium 3–4 / Microsoft Maia 200 / Meta MTIA) — inference market share may decline from 90%+ to 20–30% by 2028 (analyst projections)
  2. China export controls: Market share dropped from 95% to ~0%; FY26 Q1 included a $4.5B inventory write-down
  3. Valuation concentration risk: $5.2T market cap = world's largest; PE 44x implies perpetual growth assumptions
  4. Gross margin volatility: FY26 full-year 71.1% (vs FY25 75%), reflecting Blackwell/Rubin generational transition costs

This section is for educational purposes only. See full Disclaimer.


1. Company Fundamentals

Dimension Data Source
Company NVIDIA Corporation Polygon ticker_details
SIC Code 3674 — SEMICONDUCTORS Polygon ticker_details
Employees 42,000 Polygon ticker_details
Primary Exchange NASDAQ (XNAS) Polygon ticker_details
Fiscal Year Ends late January (FY26 = 2026-01-25) NVIDIA IR
Beta vs SPY 1.82 (18-month window) Calculated
Volatility 46% Calculated
Current 10Y Treasury 4.41% (2026-05-07) Polygon treasury_yields

Business Segments (Q4 FY26, ended 2026-01-25)

Segment Q4 FY26 Revenue YoY Share Description
Data Center $62.3B +75% 91% Core AI GPU business
↳ Compute $51.3B +58% 75% GPU + DGX systems
↳ Networking $11.0B 16% InfiniBand / Spectrum-X
Gaming $3.7B +47% 5% GeForce RTX series
Professional Visualization $1.3B +159% 2% Quadro / Omniverse
Automotive $0.6B +6% 1% DRIVE / autonomous driving

[Sources: NVIDIA Q4 FY2026 Earnings (NVIDIA Newsroom) · ServeTheHome Q4 FY2026 Analysis]

FY26 Full-Year Results

Metric FY26 Full Year YoY Notes
Total Revenue $215.9B +65% All-time high
Data Center $193.7B +73% 89.7% of total revenue
GAAP Gross Margin 71.1% vs FY25 75% Blackwell ramp transition costs
GAAP Net Income $120.1B Net margin 55.6%
Operating Cash Flow $102.7B
Shareholder Returns $41.1B Buybacks + dividends

[Source: NVIDIA FY2026 Annual Results]

FY27 Q1 Official Guidance (issued with Q4 FY26 earnings)

Metric FY27 Q1 Guidance Notes
Revenue $78.0B ± 2% ~+15% QoQ, +77% YoY (vs Q1 FY26 $44.1B)
GAAP Gross Margin ~71% Vera Rubin ramp impact expected to be manageable
China Revenue $0 (excluded from guidance) Conservative assumption under export controls

[Sources: NVIDIA Q4 FY26 Earnings · Fortune Coverage]


2. Supply Chain Position

Upstream

Supplier Relationship Risk
TSMC Sole foundry (N4/N5 Blackwell → N3 Rubin) Geopolitical risk across the Taiwan Strait; capacity competition
SK Hynix / Samsung HBM3e (Blackwell) / HBM4 (Rubin) HBM4 mass production not until 2026H2; potential bottleneck
Broadcom Networking silicon (partial NVLink Switch) Dual-sourced
CoWoS Packaging (TSMC) 2.5D advanced packaging Capacity-constrained; 18-month expansion cycle

Product Roadmap:

Generation GPU Process HBM Production Key Features
Blackwell B200/B300 N4 HBM3e (192GB) In production NVL72 liquid-cooled; sold out through 2026H1
Vera Rubin R100 N3 HBM4 (288GB) 2026H2 336B transistors, 5x Blackwell, 50 PFLOPS FP4
Vera Rubin Ultra N3 HBM4e 2027H1 Next-generation upgrade

[Sources: CNBC GTC 2026 Coverage · NVIDIA Vera Rubin Platform · Tech-Insider Rubin Analysis]

Downstream (4 Customer Types)

Customer Type Key Names Revenue Share Trend Pricing Power
Hyperscaler Microsoft, Google, Meta, Amazon ~60–70% of Data Center revenue Low (NVIDIA holds pricing power)
Neocloud CoreWeave, Lambda, Nebius, Nscale Fastest-growing segment Medium
Sovereign AI G7 governments, Middle East sovereign funds Highest margin High
Enterprise Fortune 500 Growing Medium

Key Customer AI CapEx Plans (2026):

Customer 2026 CapEx Plan YoY
Amazon $200B ~2x
Alphabet (Google) $175–185B ~2x
Microsoft $120B+ ~2x
Meta $115–135B ~2x
Big Four Total ~$700B +77%

Approximately 75% of hyperscaler CapEx (~$450B) is allocated to AI-related infrastructure.

[Sources: Fortune AI CapEx Report · CNBC Report · Tom's Hardware]


3. Sector Cycle Assessment

AI Compute Cycle: Mid-Acceleration Phase

Signal FY26/2026 Data Assessment
Hyperscaler AI CapEx $700B (2026 plan, +77% YoY) Accelerating, no deceleration signals
NVIDIA Data Center Revenue $193.7B FY26 (+73%) Primary growth engine intact
Blackwell Capacity Sold out through 2026H1 Demand exceeds supply
Vera Rubin Orders GTC 2026: $1T backlog visibility announced Demand far exceeds capacity
FY27 Q1 Guidance $78B (+77% YoY) Growth rate sustained
Inference Inflection Jensen: "the inference inflection has arrived" New addressable market opening

Core Assessment: AI compute demand is transitioning from training-driven to inference-driven, expanding rather than contracting the total addressable market. Agentic AI requires continuous 24/7 inference capability, opening a TAM potentially 3–5x larger than training alone.

[Sources: Fortune — Jensen Huang $1T · Quartz — Inference Inflection]

Counter-Signals (Cycle Risk Factors)

  • ⚠️ Hyperscaler FCF pressure: Amazon may generate negative FCF of $17–28B in 2026 (Morgan Stanley/BofA estimates)
  • ⚠️ Gross margin declined from FY25 75% to FY26 71.1% (generational transition costs)
  • ⚠️ China market share fell from 95% to ~0% (export controls)
  • No hyperscaler has signaled CapEx reductions as of this writing
  • No signs of an "AI winter" in current data

FQ Period End Rev (B) GM% OI (B) OM% NI (B) EPS Notes
Q1 FY25 2024-04-28 $26.0 78.4% $16.9 65.0% $14.9 $0.61 Hopper at peak
Q2 FY25 2024-07-28 $30.0 75.1% $18.6 62.0% $16.6 $0.68 DC $26.3B
Q3 FY25 2024-10-27 $35.1 75.0% $21.9 62.4% $19.3 $0.78 Then-record high
Q4 FY25 2025-01-26 $39.3 ~72%¹ $24.3² ~62% $22.1 $0.89 FY25 close
Q1 FY26 2025-04-27 $44.1 69.8% $27.4 62.1% $24.1 $0.98 Blackwell ramp begins
Q2 FY26 2025-07-27 $49.0 69.0% $29.8 60.8% $26.5 $1.08 GM% trough
Q3 FY26 2025-10-26 $56.3 70.7% $35.3 62.7% $30.0 $1.22 Margin recovery
Q4 FY26 2026-01-25 $66.6 72.2% $37.9² 56.9% $29.2 $1.19 Quarterly all-time high

¹ FY25 Q4 GM% derived from FY25 full-year 75% minus Q1–Q3 weighted average ² DERIVED (full year minus first three quarters)

Key Observations

  1. Q4 FY26 revenue of $66.6B = all-time quarterly record, +18% QoQ, +69% YoY
  2. Gross margin V-shaped recovery: Q2 FY26 69.0% (trough) → Q4 FY26 72.2% (+3.2pp). Initial Blackwell line adjustment costs have been absorbed
  3. Full-year net margin of 55.6%: Exceptional for the semiconductor industry (compare TSMC ~39%, AMD ~7%)
  4. Revenue growth trajectory: Q1 FY25 $26B → Q4 FY26 $66.6B = 156% growth over 7 quarters
  5. Q4 FY26 operating margin dipped to 56.9% (vs Q3 62.7%), potentially related to accelerated Vera Rubin R&D and one-time charges

FY26 Full-Year Summary

Metric FY26 FY25 YoY
Revenue $215.9B $130.5B +65%
GAAP Gross Margin 71.1% 75.0% -3.9pp
Net Income $120.1B $72.9B +65%
Operating Cash Flow $102.7B
Diluted EPS $4.90 $2.94 +67%

Shareholder Returns (FY26)

Item FY26 Source
Total Returns (buybacks + dividends) $41.1B NVIDIA IR
Remaining Buyback Authorization $58.5B Q4 FY26 earnings

[Sources: NVIDIA FY2026 Annual Results · CNBC Shareholder Returns Report]


5. Balance Sheet

Metric Q4 FY26 (2026-01-25) Notes
Total Assets $206.8B
Total Liabilities $49.5B
Total Debt $8.5B Very low leverage
Shareholders' Equity $157.3B Strong
Debt-to-Equity 5.4% Near-zero leverage
Cash & Short-Term Investments Exceeds total debt Net cash position

Key Takeaways

  • Net cash position: NVIDIA's cash holdings far exceed total debt, in stark contrast to companies like DELL (which has negative shareholders' equity). This is a nearly unleveraged balance sheet.
  • $157.3B shareholders' equity: Provides substantial capital cushion even in extreme downside scenarios.
  • Total assets of $206.8B: Grew rapidly during FY26, driven by retained earnings and inventory build-up (Vera Rubin pre-production).
  • Context: In late 2022 (before the AI wave), NVIDIA's total assets were ~$41B — a 5x increase in three years.

[Sources: StockAnalysis NVDA Balance Sheet · MacroTrends NVDA Balance Sheet]


6. Data Quality Summary

Dimension Status Notes
8Q Continuity Complete Q1 FY25 → Q4 FY26, no gaps
Q4 Derived Values (income) 2 items (Q4 FY25 OI, Q4 FY26 OI) FY minus Q1–Q3; margin of error is manageable
Segment Data Source NVIDIA IR press release L2 (not in statutory filings)
FY27 Q1 Guidance L2 (NVIDIA IR original text) $78B ± 2%
GTC $1T Forecast L2–L3 (CNBC/Fortune coverage) Jensen Huang keynote
Analyst Consensus L3 (MarketBeat/StockAnalysis) FY27 revenue $370B / EPS $8.34
Blocking Issues 0 Core data cross-verification passed

Overall Data Quality Assessment: NVIDIA's information disclosure quality is exceptionally high. Quarterly earnings press releases include complete segment revenue, gross margin, and OCF data. The annual GTC conference provides forward-looking guidance. The primary limitation is the absence of precise FactSet/Bloomberg consensus figures.


7. Peer Comparison

Dimension NVIDIA AMD Intel Broadcom TSMC
Market Cap $5,229B ~$240B ~$130B ~$1,100B ~$1,200B
TTM Revenue $215.9B ~$30B ~$50B ~$55B ~$95B
Gross Margin 71.1% ~50% ~42% ~67% ~57%
Net Margin 55.6% ~7% ~1% ~30% ~39%
AI GPU Market Share ~86% ~5–7% <1% N/A N/A (foundry)
PE (TTM) ~44x ~60x+ 100x+ ~45x ~28x
1Y Return +83.4% ~+15% ~-20% ~+40% ~+30%

Key Differentials

Dimension Interpretation
Net margin 55.6% Industry-leading — comparable only to payment networks like Visa/Mastercard. AMD's 7% net margin is 1/8th of NVIDIA's
AI GPU share ~86% Near-monopoly. AMD 5–7% + Google/Amazon custom ~10% = NVIDIA still dominates the vast majority
PE 44x vs 65% growth PEG ~0.68. For a company growing at this rate, sub-1.0 PEG is within reasonable range
Market cap $5.2T Top 3 globally; enters territory where "too big to grow" questions arise, yet revenue growth still exceeds 65%

[Sources: Multiple cross-referenced (MarketBeat/StockAnalysis/Yahoo Finance)]


8. Valuation Framework

Current Valuation Snapshot

Metric Value
Diluted Shares ~24.3B
TTM Revenue $215.9B (FY26 full year)
TTM Net Income $120.1B
TTM OCF $102.7B
TTM EPS $4.90
Total Debt $8.5B
Net Debt Negative (net cash)
PE (TTM) 43.9x
PS (TTM) 24.2x
EV/OCF ~50.9x
FCF Yield ~1.96%

Forward Valuation (Based on FY27 Consensus)

Metric Value
FY27 Consensus Revenue $370B (+71.4% YoY)
FY27 Consensus EPS $8.34
Forward PE 25.8x
Forward PS 14.1x
PEG (fwd PE / growth%) 0.36

Valuation Methods Comparison

Method Result Interpretation
TTM PE (43.9x) Appears elevated But with TTM growth of 65%, PEG of 0.68 is within reasonable range
Forward PE (25.8x) Moderate Against FY27 expected 71% growth, PEG 0.36 suggests room relative to growth
FCF Yield (1.96%) Well below risk-free rate vs 10Y Treasury 4.41%; FCF yield risk premium is -245 bp
PS TTM (24.2x) High in absolute terms But with 55.6% net margin, PS 24x is margin-adjusted equivalent of PE 44x

Valuation Context

NVIDIA's valuation reflects a "growth pricing" regime — the market is pricing in the following core assumptions:

  1. FY27 revenue of $370B (+71%) will be achieved
  2. Gross margin recovers from 71% toward 73–75%
  3. Inference market explosion expands TAM from ~$300B (training) to $1T+

FCF Yield of 1.96% vs 10Y at 4.41% = negative risk premium. This means investors are paying for future growth, not current cash flow. If FY27 growth falls below 50% (vs the 71% consensus), valuations could compress sharply.


9. Growth Catalysts

Catalyst 1: $1T Backlog Visibility (Blackwell + Vera Rubin through 2027)

  • Source: CNBC GTC 2026 · Fortune
  • Jensen Huang announced at GTC 2026 that backlog visibility doubled from $500B the prior year. Revenue visibility for FY27–FY28 is exceptionally high.

Catalyst 2: Inference Inflection = TAM Expansion

  • Source: Quartz — Inference Inflection
  • Agentic AI requires continuous 24/7 inference, unlike one-time training workloads. Inference TAM could be 3–5x training TAM. NVIDIA's TensorRT and Triton inference optimization tools lead the market.

Catalyst 3: Vera Rubin 2026H2 Mass Production

Catalyst 4: Hyperscaler CapEx Reaching $700B (2026, +77% YoY)

  • Source: Fortune · Tom's Hardware
  • Amazon $200B, Google $175–185B, Microsoft $120B+, Meta $115–135B. Even if NVIDIA captures only 30% of AI CapEx (~$210B), this exceeds FY26 Data Center revenue of $194B.

Catalyst 5: 20-Year CUDA Ecosystem Moat

Catalyst 6: Gross Margin V-Shaped Recovery

  • Source: NVIDIA Q4 FY26 Earnings
  • Q2 FY26 69.0% (trough) → Q4 FY26 72.2% (+3.2pp); Q4 non-GAAP reached 75.2%. Management guides FY27 back to mid-70s. Each 1pp of gross margin ≈ ~$3.7B additional profit (based on FY27 $370B revenue).

Catalyst 7: May 20, 2026 Earnings (FY27 Q1) — Near-Term Inflection Point


10. Risk Factors & Counter-Evidence

Risk 1: Custom Silicon Eroding Inference Market (Primary Medium-Term Risk)

  • Data: Google TPU v8 (+80% inference efficiency/dollar vs v7), AWS Trainium 3–4 (triple-digit growth, sold out), Microsoft Maia 200, Meta MTIA
  • Sources: Nerd Level Tech Custom Chip Analysis · Motley Fool
  • Analysts project NVIDIA's inference market share may decline from 90%+ to 20–30% by 2028
  • Monitor: Quarterly hyperscaler custom chip deployment volumes + NVIDIA Data Center growth rate deceleration

Risk 2: China Export Controls (Realized Loss)

Risk 3: Structural Gross Margin Decline

  • Data: FY25 75.0% → FY26 71.1% (-3.9pp). While Q4 recovered to 72.2%, Vera Rubin ramp may pressure margins again
  • Trigger: FY27 full-year GAAP GM below 68%
  • Monitor: Quarterly gross margin trends; HBM4 costs (288GB/GPU, significantly higher than HBM3e)

Risk 4: Valuation Concentration at $5.2T

  • Data: NVIDIA market cap is top 3 globally. FCF Yield of 1.96% is well below the 4.41% risk-free rate
  • Any growth deceleration below expectations could trigger sharp multiple compression

Risk 5: HBM4 Supply Bottleneck

  • Data: Each Vera Rubin GPU requires 288GB HBM4. SK Hynix + Samsung mass production begins 2026H2. Estimated 2026 capacity ceiling: 200,000–300,000 Rubin GPUs
  • Source: Introl Blog Rubin Capacity Analysis
  • Trigger: Low HBM4 yields → Rubin shipment delays → FY27 H2 revenue miss

Risk 6: TSMC Capacity & Geopolitical Risk

  • NVIDIA is 100% dependent on TSMC for fabrication. N3 capacity is shared with Apple and AMD
  • Extreme scenario: Taiwan Strait escalation → TSMC production halt (low probability but non-zero)

11. Upcoming Events & Monitoring Calendar

Timeframe Event Key Focus
2026-05-20 NVDA FY27 Q1 Earnings Revenue vs $78B consensus / gross margin trend / Q2 guidance (consensus $86.6B) / China revenue
2026-08 (est.) NVDA FY27 Q2 Earnings Initial Vera Rubin shipment feedback / gross margin / Blackwell backlog
2026-11 (est.) NVDA FY27 Q3 Earnings Vera Rubin production ramp / FY27 full-year guidance / inference revenue share
2027-02 (est.) NVDA FY27 Q4 Earnings FY27 actuals vs $370B consensus / FY28 guidance (critical inflection)

Interim Events

Timeframe Event Key Focus
2026-06 COMPUTEX 2026 Vera Rubin product details / new customer announcements
2026 H2 Vera Rubin First Shipments Actual ASP / yield rates / customer feedback
Quarterly Hyperscaler Earnings AI CapEx maintenance vs cuts

12. CUDA Software Moat — Deep Dive

Why CUDA Is a Deeper Moat Than Hardware

Dimension CUDA Ecosystem AMD ROCm Google TPU (JAX)
Ecosystem Age 20 years (since 2006) ~8 years ~10 years
Developer Base Millions Tens of thousands Primarily Google-internal
Optimized Libraries 5,000+ ~300 N/A
PyTorch Native Support First-class Second-class (requires ROCm compilation) JAX-primary (not PyTorch)
Migration Cost 3–6 months of engineering Requires code rewrite
Inference Optimization TensorRT + Triton Server MIGraphX (immature) TPU Inference (GCP-only)

Competitive Progress

  • AMD ROCm: As of 2026, PyTorch/JAX have native support; ZLUDA project enables CUDA code compatibility. But production stability remains far behind CUDA.
  • Google TPU v8: +80% inference efficiency/dollar (vs v7), but available only on GCP — hardware cannot be purchased.
  • AWS Trainium 3–4: Sold out, triple-digit growth. Available only on AWS.
  • Meta MTIA: Internal use only, not sold externally.

Assessment: In the near term (2026–2027), the CUDA moat remains unbreakable. In the medium term (2028–2030), the inference market may see 10–30% erosion from custom silicon, but NVIDIA will likely continue to dominate training.

[Sources: GPUnex Blog · Thundercompute ROCm vs CUDA · Rayhan Press CUDA Moat]


13. China Export Controls — Impact Analysis

Timeline

Date Event Impact
2022-10 First U.S. AI chip export controls A100/H100 banned for sale to China
2023-10 Second round tightening Downgraded H800/A800 also banned
2024 H20 (further downgraded) still exportable NVIDIA China revenue ~15%
2025-04 H20 exports banned + $4.5B inventory write-down NVIDIA China revenue approaches zero
2026-01 DOC partially eases restrictions (H200 exportable) Jensen announces Chinese H200 orders received
2026 (ongoing) 15% tariff on AI chip exports to China Reduced economic viability

Financial Impact Quantification

Scenario Est. China Revenue % of Total Notes
FY24 Peak ~$17B ~26% Pre-controls
FY25 ~$8–10B ~6–8% H800/H20 era
FY26 ~$0 ~0% $4.5B write-down + full ban
FY27 Guidance $0 (excluded) 0% Management's conservative assumption

Impact Assessment: The ~$17B/year China revenue loss has been fully offset by the global AI CapEx explosion (FY26 Data Center revenue of $194B set a new record even with zero China contribution). However, if controls expand to Middle Eastern and Southeast Asian third countries, a new downside risk emerges.

[Sources: Tom's Hardware Jensen Comments · Computer Weekly $4.5B Write-down]


Sources & Citations

Official Sources (L1–L2)

  1. NVIDIA Q4 FY2026 Earnings Press Release (2026-02-25) — NVIDIA Newsroom
  2. NVIDIA Vera Rubin Platform Announcement — NVIDIA Newsroom
  3. Fortune: Jensen Huang $1T AI Infrastructure — Fortune
  4. CNBC GTC 2026 Coverage — CNBC
  5. Fortune: Big Tech $700B AI CapEx — Fortune

Cross-Referenced Sources (L2–L3)

  1. ServeTheHome Q4 FY2026 Analysis — ServeTheHome
  2. Tom's Hardware: Jensen on China — Tom's Hardware
  3. Computer Weekly: $4.5B Write-down — Computer Weekly
  4. Tech-Insider: Vera Rubin Analysis — Tech-Insider
  5. Nerd Level Tech: Custom Chip Race — Nerd Level Tech
  6. Stratum Review: CUDA Moat — Stratum Review
  7. Introl Blog: Rubin Capacity — Introl