Executive Summary

$215.9B
FY2026 Revenue
$120.1B
FY2026 Net Income
$102.7B
FY2026 Op. Cash Flow
$62.6B
Cash & Securities
89.2%
FY2025 ROIC

NVIDIA Corporation has executed the most dramatic corporate transformation in modern business history, evolving from a diversified gaming GPU company into the dominant infrastructure provider for the artificial intelligence revolution. The numbers are staggering by any measure: FY2026 revenue reached $215.9 billion, up 65.5% year-over-year, with net income of $120.1 billion and operating cash flow of $102.7 billion. Four years earlier, in FY2022, total revenue was $26.9 billion. This is not a company that grew — it metamorphosed into an entirely different entity operating at a scale that no semiconductor company has approached in history.

The engine of this transformation is Data Center, which accounted for 88.3% of FY2025 revenue ($115.2 billion) and grew at a 121% CAGR from FY2022 to FY2025. Within Data Center, the GPU compute sub-segment — the H100, H200, and now Blackwell accelerators — represents 78% of total company revenue, making NVIDIA's financial performance a near-direct proxy for global AI infrastructure investment. The Compute & Networking segment generated $82.9 billion in operating income at a 71.3% operating margin in FY2025. For reference: Intel's entire revenue was $52.9 billion in FY2025 with a negative operating income.

The financial quality underpinning these results is exceptional. Gross margin expanded from 56.9% in FY2023 to 75.0% in FY2025, before dipping to 60.5% in Q1 FY2026 during the Blackwell production ramp (driven by a $4.5 billion H20 inventory charge) and recovering to 75.0% by Q4 FY2026. The Piotroski F-Score of 8/9 signals STRONG BUY across two consecutive years. The Altman Z-Score reached 6.57 in FY2025 — a figure that places NVIDIA firmly in the upper tier of financial safety among all public companies. The Beneish M-Score flags as LIKELY MANIPULATOR at -0.95 (threshold: -2.22), but this is a well-documented false positive for hypergrowth companies: the dominant driver is the Sales Growth Index of 2.14x, and forensic analysis confirms the CFO/Net Income ratio of 0.88 provides genuine cash backing for reported earnings.

ROIC reached 89.2% in FY2025 — arguably the highest sustained return on capital in semiconductor history. The balance sheet is fortress-grade: $43.2 billion in liquid assets (cash plus marketable securities) against $8.5 billion in fixed-rate debt, with interest coverage of 329.8x. Jensen Huang has created a business that generates more cash than it can reinvest or return to shareholders: FY2025 operating cash flow of $64.1 billion funded $33.7 billion in buybacks, $6.9 billion in tax withholding on equity vesting, and still left the company with a growing cash position.

The CUDA ecosystem moat is central to the investment thesis. AMD's FY2025 10-K explicitly names CUDA as a competitive barrier that influences AMD's own customers. Intel's FY2025 10-K records a $5 billion equity sale to NVIDIA — an extraordinary reversal where the former dominant processor company sold itself a stake to a GPU maker. The competitive position score of 88/100 from transcript analysis reflects a five-layer moat: CUDA ecosystem (1.5 million trained models), full-stack data center integration, networking ($44 billion annualized), annual architecture cadence (Hopper → Blackwell → Vera Rubin), and ecosystem investments ($10 billion in Anthropic, Groq acquisition).

However, this extraordinary performance carries material risks demanding clear-eyed assessment. First, concentration risk is severe and worsening: Data Center is 90% of Q3 FY2026 revenue, and the top four hyperscalers (Microsoft, Amazon, Google, Meta) represent an estimated 50%+ of Data Center revenue. The entire $216 billion business is a single-segment bet on AI infrastructure investment. A 40-50% revenue decline in an AI capex downturn is mathematically plausible — it has happened before in gaming (FY2023 -27%). Second, the April 2025 H20 export ban permanently eliminated approximately $8 billion per quarter in China Data Center compute revenue. NVIDIA has more than offset this with growth elsewhere, but the $17.1 billion China geography in FY2025 at 13.1% revenue share carries escalating regulatory risk. Singapore's explosive growth from $2.3 billion (FY2023) to $23.7 billion (FY2025), a 222% CAGR in two years, warrants scrutiny as potential indirect China demand routing: no city-state can organically absorb $23.7 billion in AI GPU spend.

CEO Jensen Huang's insider selling totals $1.0 billion in open-market sales from June 2025 through March 2026, conducted via a 10b5-1 pre-planned trading arrangement in 361 transactions across 83 trading days. This is systematic, continuous selling at an average rate of approximately $80 million per month. The appropriate forensic interpretation: the 10b5-1 structure removes any inference of real-time bearish conviction, but the scale — nearly $1 billion in 10 months — is notable context. CFO Colette Kress sold $81.2 million over the same period. Seventeen insiders collectively sold $1.83 billion in the period with zero open-market purchases.

Management credibility scores 87/100 from transcript forensics covering Q1-Q4 FY2026. CFO Kress's guidance has been accurate within 10-33 basis points on gross margins and consistently 4-5% above the midpoint on revenue — a systematic sandbagging pattern (score: 8.5/10). The 'compute equals revenues' thesis repeated by Jensen Huang across all four quarters is the most consequential unfalsifiable claim in NVIDIA's coverage universe: no analyst challenged it across 31 Q&A pairs. The China $50B TAM narrative was invoked aggressively in Q1 for policy advocacy, then silently dropped as non-China growth compensated. Vera Rubin demand visibility remains the critical unknown: CFO Kress's most hedged answer of the year was 'too early to know' on Rubin demand in Q4, and if this answer persists through Q2 FY2027, it would signal a potential air pocket in the next product cycle.

At current prices around $110-115 (as of report date), NVIDIA trades at approximately 22-24x FY2026 revenue and 40-50x GAAP earnings. The reverse DCF implied growth rate to justify current valuation at a 10% discount rate requires sustained earnings growth of approximately 20-25% annually for a decade — achievable if AI infrastructure investment continues compounding, but historically rare for any company. The FY2027 Q1 guidance of $78 billion in revenue (if annualized at $312 billion) implies the market is pricing continued 40-50% growth. The valuation is not irrational given the earnings power trajectory, but it discounts significant execution of an uncertain demand curve for AI that no analyst is stress-testing.

Verdicts

Business Quality
A
Score: 92/100
Exceptional — AI infrastructure monopoly with 75% gross margin and 89% ROIC

NVIDIA is among the highest-quality businesses in public market history at scale. FY2026 gross margin of 71.1% and net margin of 55.6% are unprecedented for a hardware company at $216 billion in revenue. The CUDA software ecosystem functions as a multi-year switching cost barrier — AMD's own 10-K cites it as a barrier influencing AMD's customers. ROIC of 89% in FY2025 reflects extraordinary capital efficiency in a fabless model. The Piotroski F-Score of 8/9 and Altman Z-Score of 6.57 confirm textbook financial strength. The one-point deduction from A+ reflects extreme concentration risk: a single segment at 89% of revenue creates structural vulnerability to AI capex cyclicality that has not been tested at this scale.

Management Quality
B+
Score: 79/100
Strong with specific concerns — exceptional operator, governance sound, selling pattern notable

Jensen Huang is a visionary founder-CEO who has steered NVIDIA through three platform transitions over 32 years. CEO comp CAGR of 27% dramatically trails revenue CAGR of 67% and earnings CAGR of 103% — shareholder favorable. The board is 92% independent with single-class shares, no poison pill, annual elections. Management credibility scores 87/100 on transcript analysis; CFO Kress earns 92/100 for guidance precision. B+ rather than A- reflects: (1) CEO and CFO combined insider selling of $1.08B via 10b5-1 over 10 months; (2) all performance targets at maximum payout for 2 consecutive years; (3) 54% of board with 10+ year tenure; (4) Jensen's escalating absolutism on compute-equals-revenues without qualifier.

Current Trajectory
B+
Score: 78/100
Decelerating but still explosive — Blackwell ramp complete, Rubin visibility uncertain

FY2026 added $85.4 billion in revenue — the largest single-year revenue addition in semiconductor history. Q4 FY2026 at $68.1 billion (+73% YoY) signals reacceleration. Q1 FY2027 guide of $78 billion implies 14.5% sequential growth. Hyperscaler CapEx correlation of 0.96 with NVIDIA revenue at one-year lag supports FY2027 above $300B. B+ rather than A- reflects: (1) revenue growth mathematically decelerating (126% → 114% → 65% YoY); (2) H20 ban permanent $32B+ annual revenue hole; (3) Vera Rubin demand genuinely opaque with both Jensen and Kress hedging across multiple quarters; (4) efficiency-gain wildcard (DeepSeek-style inference optimization) unpriced by any analyst.

Valuation Context

Trailing Multiples (at ~$112/share)

MultipleValueContext
EV/Revenue12.43xExpensive for hardware
EV/EBITDA19.8xPremium to sector
P/FCF27.4xReflects growth premium
P/E (GAAP)22.7xModerate for 65% growth
Forward P/E18.5xReasonable on FY2027 est.

Intrinsic Value Models

ModelPer Sharevs Current ~$112
Earnings Power Value (EPV)$62-45% overvalued
Graham Number$38.50-66% overvalued
Residual Income$95-15% overvalued
Reverse DCF (implied CAGR)22%/yr for 10yrAchievable but rare

Risk / Return Profile

MetricValue
Beta (5yr)1.72
Annualized Volatility52.4%
Sharpe Ratio1.45
Max Drawdown-42.3%
VaR (95%, 1-day)-8.2%

Technical Momentum

IndicatorValue
RSI (14-day)44.0
50-day SMA$120.50
200-day SMA$118.80
Trend StateBelow both SMAs — Bearish
12-Month Return-26.3%
Analyst Valuation Assessment

NVIDIA's valuation defies traditional semiconductor framework analysis. At approximately 22-24x FY2026 revenue and 40-50x GAAP earnings (at $110-115 share price), the stock demands investors accept one of two propositions: either AI infrastructure spending will continue compounding at rates that sustain 30-50% revenue growth for multiple years, or the current valuation will compress meaningfully as growth decelerates to semiconductor-cycle norms of 10-20% annual growth.

The bull case rests on the hyperscaler CapEx correlation (0.96 with NVIDIA revenue at one-year lag) and the Q1 FY2027 guide of $78 billion implying $312 billion annualized revenue — a 44% increase above FY2026. Three hyperscalers alone spent an estimated $260 billion on CapEx in CY2025, up 68% year-over-year. The Vera Rubin architecture in the pipeline provides the next demand cycle, and sovereign AI ($30 billion+ tripled year-over-year in FY2026 transcript analysis) adds genuine demand breadth beyond the hyperscaler oligopoly.

The bear case centers on law-of-large-numbers mathematics combined with structural risks. The reverse DCF model implies a 22% annual earnings CAGR for ten years to justify current prices at a 10% discount rate — achievable in the AI bull case, unprecedented in most market environments. The H20 export ban proved regulatory risk is real and swift. Singapore's $23.7 billion in revenue (18.1% of FY2025 total) growing 247% year-over-year suggests demand routing that could face its own regulatory scrutiny. The efficiency-gain wildcard — AI models getting dramatically cheaper to run per inference — has no precedent in NVIDIA's publicly presented demand models.

Verdict: Expensive at current prices; requires sustained 20-25% earnings CAGR for a decade to justify valuation at 10% discount rate. Intrinsic value models anchor at EPV $62, Graham $38.50, residual income $95 — growth premium is very real and priced in.

Financial Overview & Quarterly Trends

Annual Financial Summary (FY2023–FY2026)

Metric FY2023 FY2024 FY2025 FY2026
Revenue$26.97B$60.92B$130.50B$215.94B
Gross Profit$15.36B$44.30B$97.86B$153.46B
Operating Income$4.22B$32.97B$81.45B$130.39B
Net Income$4.37B$29.76B$72.88B$120.07B
Operating Cash Flow$5.64B$28.09B$64.09B$102.72B
EPS (Diluted)$0.17$1.19$2.94$4.90
Gross Margin56.9%72.7%75.0%71.1%
Operating Margin15.7%54.1%62.4%60.4%
Net Margin16.2%48.9%55.9%55.6%
ROE19.8%69.2%91.9%76.3%
ROIC65.7%89.2%
R&D Expense$7.34B$8.68B$12.91B$18.50B
Total Assets$41.18B$65.73B$111.60B$206.80B
LT Debt$9.70B$8.46B$8.46B$7.47B
Cash & Securities$43.21B$62.56B
Share Buybacks$9.50B$33.71B$40.09B

Quarterly Trends (FY2025 Q4 – FY2027 Q1 Guide)

Quarter Revenue Gross Margin Data Center Key Event
FY2025 Q4 (Jan 2025)$39.3B73.5%$35.6BBlackwell launch quarter
FY2026 Q1 (Apr 2025)$44.1B60.5%$39.1BH20 ban — $4.5B charge
FY2026 Q2 (Jul 2025)$46.7B72.4%$41.1BMargin recovery begins
FY2026 Q3 (Oct 2025)$57.0B73.4%$51.2BBlackwell acceleration
FY2026 Q4 (Jan 2026)$68.1B75.0%$62.3BFull Blackwell ramp
FY2027 Q1 (Guide)$78.0B~73%+14.5% QoQ implied

The Q1 FY2026 gross margin trough of 60.5% — driven by the H20 export control inventory write-down — was transitory. Margin recovery to 75.0% by Q4 FY2026 confirms Blackwell manufacturing normalization. The Q1 FY2027 guide implies $312B annualized at current trajectory.

Forensic Models

Beneish M-Score
Earnings manipulation detector | Threshold: -2.22 (flagged above)
-0.948
-3.0 (Safe)-2.22 threshold0 (Danger)
Classification: LIKELY MANIPULATOR — HIGH FALSE POSITIVE PROBABILITY
SGI (Sales Growth Index): 2.142 — dominant driver
DSRI: 1.077 | DEPI: 1.216 | GMI: 0.970 | TATA: 0.079
Counter-signals: CFO/NI ratio 0.88 confirms cash backing; GMI<1 means margins improved; academic literature notes >90% false positive rate when SGI>1.5x. Forensic verdict: FALSE POSITIVE — growth-driven, not manipulation-driven.
Altman Z-Score
Bankruptcy predictor | Safe zone: above 2.99
6.574
0 (Distress)1.81 / 2.996+ (Safe)
SAFE — bankruptcy risk is effectively zero
FY2025: 6.57 | FY2024: 4.97 | FY2023: 2.52 (Grey Zone)
X1 (Working Capital/Assets): 0.556 | X2 (RE/Assets): 0.610
X3 (EBIT/Assets): 0.730 | X4 (Equity/Liabilities): 2.458
Trend: DRAMATICALLY IMPROVING — from grey zone to upper safe tier in 2 years
Piotroski F-Score
9-signal financial strength (8-9 = Strong Buy)
8/9
0 (Weak)4-59 (Strong)
STRONG BUY
FY2025: 8/9 | FY2024: 8/9 | FY2023: 6/9
Failed signal: F4 (CFO < Net Income)
Context: F4 failure is structural to hypergrowth — working capital expansion from accelerating receivables. Not a quality concern; 8 of 9 signals pass cleanly for 2 consecutive years.
Sloan Accrual Ratio
Earnings quality — lower is better (cash-backed)
0.330
-0.5 (Best)0+0.5 (Risk)
ELEVATED — but investing-driven, not accrual-driven
FY2025: 0.330 | FY2024: 0.229 | FY2023: -0.203
Context: Elevated Sloan ratio driven by security purchases in investing activities. Operating accrual ratio (NI-CFO/Assets) is ~8% — benign. Cash backs 88% of earnings per CFO/NI ratio.
Cash Conversion Quality
CFO / Net Income — measures earnings cash backing
0.856
0 (Poor)0.751.0+ (Excellent)
STRONG — cash backs 86% of reported earnings (FY2026)
FY2026: 0.856 | FY2025: 0.879 | FY2024: 0.944
Cash taxes EXCEEDED GAAP taxes in FY2024-FY2025 by $2.5-4.0B — positive quality signal, no aggressive sheltering. Piotroski F4 fail (CFO<NI) is the only minor drag; all other quality signals strong.

Transcript Intelligence

Composite Scores (Q1-Q4 FY2026)

DimensionScore
Overall Transcript Score86/100
Management Credibility87/100
Guidance Accuracy92/100
Competitive Position88/100
Growth Quality85/100
Narrative Consistency0.82/1.0
Revenue Guidance Accuracy95/100
Gross Margin Accuracy90/100
Product Timeline Delivery93/100
Linguistic Honesty78/100

100% revenue beat rate — average 4.7% above midpoint. Sandbagging score 8.5/10. Guidance is a managed floor, not a true forecast.

Guidance Accuracy Detail

QuarterGuidedActualBeat
Q1 FY2026$43.0B$44.1B+2.6%
Q2 FY2026$45.0B$46.7B+3.8%
Q3 FY2026$54.5B$57.0B+4.6%
Q4 FY2026$65.0B$68.1B+4.8%

Cumulative beats total approximately $4.4B across the year. Gross margin guidance was accurate within 10-33 basis points each quarter.

Cross-Layer Investigative Flags

XFLAG-001: Vera Rubin Demand Visibility Gap HIGH

Both Jensen Huang and CFO Colette Kress hedged on Vera Rubin demand across 3 consecutive quarters. CFO Kress's most hedged phrase of FY2026: "too early to know" on commercial trajectory in Q4. If this answer persists through Q2 FY2027 (August 2026), it signals a potential product cycle air pocket that the sell-side is not modeling.

XFLAG-002: China $50B TAM Disappeared Without Resolution MEDIUM

The $50B China TAM narrative was invoked aggressively in Q1 FY2026 for policy advocacy purposes. By Q4, it had been silently retired without resolution as non-China growth compensated. No analyst pressed management on this discontinuity across 31 Q&A pairs.

XFLAG-003: HBM Memory Cost Exposure Never Quantified MEDIUM-HIGH

High Bandwidth Memory (HBM) is estimated to represent 40-60% of COGS for Blackwell GPUs. Despite four quarters of earnings calls and 31 Q&A pairs, no analyst demanded a breakdown and management never quantified it. This is the single most important unasked cost question in NVIDIA's coverage universe.

XFLAG-004: Compute Equals Revenues Thesis Unchallenged MEDIUM-HIGH

Jensen's conviction on the "compute equals revenues" thesis rose from hedged (Q1) to absolute (Q4 FY2026) — a confidence escalation from 0.4 to 1.0 on a normalized scale — with zero analyst challenge across all four quarters. The thesis is unfalsifiable as stated and represents the most consequential unexamined assumption in AI semiconductor analysis.

XFLAG-005: Analyst Community Capitulation to Unanimous Bullishness MEDIUM

The 100% guidance beat rate has conditioned pavlovian optimism among sell-side analysts covering NVDA. Bear case scenarios (AI efficiency improvements, CapEx cycle reversal, ASIC competition) are not modeled by any major sell-side firm based on transcript Q&A analysis. This creates systematic information risk for institutional investors.

Revenue & Segments

Segment Revenue (FY2022–FY2025)

SegmentFY2022FY2023FY2024FY20253-Yr CAGRFY2025 Margin
Compute & Networking$11.0B$15.1B$47.4B$116.2B119.7%71.3%
Graphics$15.9B$11.9B$13.5B$14.3B-3.4%35.6%

Concentration risk CRITICAL: Compute & Networking is 89% of total revenue and 94% of segment operating income. Graphics — the historical core business — is now 11% of revenue.

Product Revenue Breakdown (FY2025)

Data Center 88.3%
Gaming 8.7%
Other 3.0%
ProductFY2022FY2025Share FY2025CAGR
Data Center$10.6B$115.2B88.3%121.2%
Gaming$12.5B$11.4B8.7%-3.0%
Professional Visualization$1.9B1.4%
Automotive$0.6B$1.7B1.3%44.1%
OEM & Other$0.4B0.3%

Geographic Revenue (FY2025)

GeographyFY2025 RevenueShareCAGRNotes
United States$61.3B46.9%141.2%Hyperscaler demand core
Singapore$23.7B18.1%221.7% (2yr)⚠ Demand routing scrutiny
Taiwan$20.6B15.8%Manufacturing partners
China (incl. HK)$17.1B13.1%H20 ban Apr 2025 — compute eliminated
Other$7.9B6.0%

Singapore Flag: A 221.7% CAGR over two years to $23.7 billion in annual revenue cannot be explained by organic demand in a city-state of 5.6 million people. This warrants scrutiny as potential indirect China demand routing — a geography that could face its own regulatory attention.

Governance & Compensation

Entrenchment Score: 2/10

2

0 = Shareholder Friendly — 10 = Entrenched

Governance AttributeAssessment
Share StructureSingle-class (1 vote/share)
Poison PillNone
Staggered BoardNone — annual elections
Majority VotingYes
Proxy AccessYes
Supermajority RemovalBoard-sponsored FY2025 elimination
Clawback PolicyActive
Board Independence92.3% (12/13 directors)

CEO Compensation vs. Value Creation

MetricFY2021FY2025CAGR
CEO Total Comp (SCT)$19.3M$49.9M+26.7%
Company Revenue$16.7B$130.5B+67.3%
Net Income$4.3B$72.9B+103.0%

CEO comp CAGR of 27% dramatically trails revenue CAGR of 67% and earnings CAGR of 103%. CEO Compensation at Actual Pay (CAP) was NEGATIVE in FY2023 (-$4.1M) — real downside alignment demonstrated. This is one of the most favorable pay-for-value alignment profiles among mega-cap technology CEOs.

Concern: All performance targets at maximum payout for 2 consecutive years (FY2024, FY2025). Target-setting insufficiently ambitious given revenue trajectory.

Board Composition (FY2025 DEF 14A)

AttributeValue
Total Directors13
Independent Directors12 (92.3%)
Lead Independent DirectorStephen C. Neal
Women Directors5 (38.5%)
Diverse Directors4
Directors with 10+ year tenure7 (54%) — potential loyalty dynamic

Tenure Concern: Three directors have served since the 1993 founding era (32+ years). While 92.3% independence is excellent, multi-decade relationships with the founder-CEO can create unquantifiable loyalty dynamics that independence metrics do not capture.

Insider & Institutional Activity

Aggregate Insider Activity: June 2025 – March 2026

$1,832,015,390 Total Sold $0 Total Purchased 99 Form 4 Filings 17 Unique Sellers

Key Insider Activity Detail

InsiderRole$ SoldShares SoldPlanAssessment
Jensen HuangCEO & Co-Founder $1,003,717,158 55,189,560 10b5-1 361 trades over 83 days — $80M/month avg. Pre-planned; no real-time bearish signal. Scale notable.
Mark StevensDirector $292,489,711 1,641,979 Non-10b5-1 Discretionary sales — informationally significant. Not pre-planned.
Ajay PuriEVP, Worldwide Sales $183,002,962 1,000,000 10b5-1 Pre-planned; routine executive diversification.
Colette KressCFO $81,233,232 449,500 10b5-1 Pre-planned; routine executive diversification.
Timothy TeterEVP General Counsel +4,691 net N/A Only insider with net positive position change in period.
Forensic Interpretation

The 10b5-1 structure for CEO Jensen Huang and CFO Colette Kress means the selling was pre-planned and does not reflect real-time bearish views about NVIDIA's prospects. However, the aggregate scale ($1.83 billion in 10 months from 17 insiders) and the absence of any open-market purchases from any insider deserves monitoring context.

Director Mark Stevens' $292.5 million in non-10b5-1 sales is the most informationally significant item in this dataset — these are discretionary transactions that reflect real-time judgment rather than pre-programmed plans. Three simultaneous selling windows (Nov 2025-Mar 2026: 7 insiders; Jul-Dec 2025: 6 insiders; Aug-Oct 2025: 5 insiders) suggest coordinated awareness of company circumstances.

Context: Jensen Huang also gifted 119.2 million shares in the period, suggesting estate planning alongside cash diversification. The overall pattern is consistent with long-term holders executing planned diversification from concentrated positions — not an insider bear thesis. Monitor if selling pace accelerates or if discretionary (non-10b5-1) transactions increase.

Peer Comparison

NVDA vs. AMD vs. Intel (Latest Annual)

Metric NVDA (FY2026) AMD (FY2025) INTC (FY2025)
Revenue$215.9B$34.6B$52.9B
Net Income$120.1B$4.3B-$0.3B
Gross Margin71.1%49.5%34.8%
Operating Margin60.4%10.7%-4.2%
Net Margin55.6%12.5%-0.5%
R&D % Revenue8.6%23.4%26.1%
ROE76.3%6.9%-0.2%
Total Assets$206.8B$76.9B$211.4B

Dominance Metrics

vs. AMD Revenue
6.2x
NVDA wins
vs. AMD Net Income
27.7x
NVDA wins
vs. INTC Revenue
4.1x
NVDA wins
Gross Profit vs. peers
4.3x
NVDA wins

NVDA's $153.5B gross profit exceeds AMD+INTC combined gross profit ($35.5B) by 4.3x.

Industry Rankings (CY2024)

Ranking DimensionNVDA Rank
Revenue (all US filers)19th
Net Income (all US filers)5th
FY2026 Net Income (projected global)1st or 2nd

FY2026 net income of $120.1B would rank 1st or 2nd globally, exceeding Alphabet's $100.1B. This positions NVIDIA as the most profitable non-financial company in public market history by net income.

Competitive Intelligence from Peer Filings

AMD FY2025 10-K takes a defensive tone, explicitly citing CUDA ecosystem as a competitive barrier that influences AMD's own customers' purchasing decisions. AMD also flags the NVIDIA-Intel partnership as a material risk factor. This is remarkable: AMD's SEC filing is, in effect, a testimonial to NVIDIA's competitive position.

Intel FY2025 10-K records a $5 billion equity stake sale to NVIDIA — Intel received 215 million NVDA shares at $23.28/share in exchange for its equity stake. Intel is collaborating on CPU+GPU integrated products while having publicly conceded the AI accelerator market to NVIDIA. This is one of the most extraordinary competitive capitulations in semiconductor industry history.

Text Analysis

Growth Signals in 10-K Text

Signal / TermCountSentiment
Blackwell5Strongly Positive
AI (pervasive)ManyStrongly Positive
Data Center5Strongly Positive
Sovereign AI0ABSENT — despite $30B+ in transcripts

Flag: "Sovereign AI" appears zero times in the FY2025 10-K despite being cited as a $30B+ revenue category in Q1-Q4 FY2026 earnings transcripts. The legal/risk disclosure framework does not acknowledge this as a distinct revenue category — a disclosure gap.

Risk Signals in 10-K Risk Factors

Risk CategorySeverityNew in FY2025?
Export Control (general)CRITICALNo — escalating since Oct 2022
AI Diffusion IFR (Jan 2025)CRITICALYES — worldwide licensing
China (no licenses granted)CRITICALNo
Supply Chain (TSMC, CoWoS)HIGHNo
Taiwan-China geopoliticalHIGHNo
TariffMODERATENo

Notable XBRL Taxonomy Extensions

ExtensionForensic Interpretation
nvda:PrepaidSupplyAndCapacityAgreementsNoncurrentLong-term TSMC capacity prepayments — reflects supply chain concentration and strategic commitment to CoWoS packaging
nvda:ExcessInventoryPurchaseObligationsCurrentPotential write-down risk indicator — monitor quarterly for changes in obligation size
nvda:AccruedCustomerProgramsCurrentChannel incentive provisions — relevant to revenue recognition timing
nvda:BusinessCombinationAdvancedConsiderationWrittenOffLikely ARM deal deposit write-off from the failed 2022 acquisition attempt ($66B deal blocked by regulators)

Export Control Timeline

October 2022
Initial A100/H100 Export Restrictions
BIS restricts advanced AI chips to China — NVIDIA creates A800/H800 variants at lower performance thresholds.
July 2023
Expanded Restrictions to A800/H800
BIS closes the threshold gap — NVIDIA creates H20 chip designed to comply with new thresholds.
January 2025
AI Diffusion IFR — Worldwide Licensing
Critical: worldwide licensing requirement imposed on ALL AI products. Three-tier country system creates new compliance complexity. NEW disclosure in FY2025 10-K.
April 2025
H20 Export Ban — China Data Center Eliminated
H20 chip banned — $4.5B inventory charge in Q1 FY2026. Permanently eliminates ~$8B/quarter in China Data Center compute revenue.

Advanced Analysis

Tax Forensics — ETR Trend

YearETRKey Driver
FY20221.9%Tax benefits dominating small income base
FY2023-4.5%Net tax benefit — R&D credits exceed taxes
FY202412.0%Income scale absorbs fixed credits
FY202513.3%Normalizing as income grows
FY202615.1%Rising but still below statutory 21%

Quality signal: Cash taxes EXCEEDED GAAP taxes by $2.5-4.0B in FY2024-FY2025 — positive indicator of no aggressive tax sheltering. Israel reduced-rate benefit: $3.064B in FY2026. FDDEI benefit: $4.208B. R&D credit: $1.933B.

SBC & Dilution Analysis

YearSBC Expense% Revenue% Net Income
FY2022$2.00B7.4%
FY2025$4.74B3.6%6.5%
FY2026$6.39B3.0%5.3%

ANTI-DILUTIVE: Buybacks ($40.1B in FY2026) consistently exceed share issuances — share count declining ~0.7% annually (173M net shares in FY2026). Unrecognized SBC of $14.8B represents approximately 2.6 years of forward expense at current rates.

Debt Quality Assessment

MetricFY2026
Total Debt$8.47B
Debt TypeFixed-rate unsecured senior notes
Floating Rate Exposure0%
Interest Coverage502.2x
Debt/Equity0.054x
Commercial Paper Program$25B (Jan 2026 increase)
Commercial Paper Outstanding$0

Fortress balance sheet — debt is negligible. The $25B commercial paper program increase in January 2026 provides liquidity flexibility while currently undrawn.

Hyperscaler CapEx Correlation

MetricValue
Correlation (CapEx vs NVDA revenue, 1yr lag)r = 0.96
CY2025 3-hyperscaler CapEx estimate~$260B
YoY CapEx growth+68% YoY
Implied FY2027 NVDA revenue>$300B

A 0.96 correlation coefficient is near-perfect co-movement. If hyperscaler CapEx continues expanding and NVIDIA maintains its current capture rate of AI infrastructure spend, FY2027 revenue above $300B is mathematically consistent with the historical relationship.

Risk & Convergence Matrix

Red Flags Identified

China Export Control Escalation HIGH

3-year escalation from Oct 2022 → Jul 2023 → Jan 2025 → Apr 2025 H20 ban. The trajectory is unambiguously tightening. The AI Diffusion IFR creates worldwide licensing complexity. China was 13.1% of FY2025 revenue ($17.1B) — further restrictions are a tail risk priced only partially by markets. Source: Text analysis, SEC risk factors.

Singapore Revenue (221% CAGR) — Potential Demand Routing HIGH

Singapore revenue grew from $2.3B (FY2023) to $23.7B (FY2025) — a 221.7% CAGR that cannot be explained by organic domestic demand. This pattern is consistent with third-country demand routing as Chinese entities purchase through Singapore entities. If regulators identify and restrict this routing, it would eliminate a material and fast-growing revenue stream. Source: Segment analysis, geographic data.

Data Center Concentration — 89% of Revenue HIGH

The entire $216 billion business is effectively a single-segment bet on AI infrastructure investment. The top four hyperscalers are estimated to represent 50%+ of Data Center revenue. NVIDIA experienced a -27% revenue decline in FY2023 when gaming demand cyclicality reversed — a similar event in AI capex would be mathematically devastating at current scale. Source: Segment analysis.

Vera Rubin Demand Gap (XFLAG-001) HIGH

The next major product cycle (Vera Rubin architecture) has genuine demand opacity. Both Jensen and CFO Kress hedged on Rubin across 3 consecutive quarters. If orders do not formalize by Q2 FY2027, a product cycle air pocket becomes plausible — a risk no analyst in the coverage universe is currently modeling. Source: XFLAG-001 transcript forensics.

HBM Memory Cost Exposure — Unquantified (XFLAG-003) MEDIUM-HIGH

High Bandwidth Memory estimated at 40-60% of Blackwell COGS. SK Hynix and Samsung hold pricing leverage. HBM pricing dynamics are not disclosed by NVIDIA and have not been questioned in 4 quarters of analyst calls. A 10% HBM price increase would reduce gross margin by approximately 400-600 basis points. Source: XFLAG-003 transcript forensics.

CEO Insider Selling — $1.0B in 10 Months LOW-MEDIUM

Pre-planned 10b5-1 selling by CEO Jensen Huang ($1.003B, 361 transactions, 83 trading days) — approximately $80M/month. The 10b5-1 structure eliminates real-time bearish inference, but the scale is notable context. Seventeen total insiders sold $1.83B aggregate with zero open-market purchases. Source: Form 4 analysis.

Convergent Signals (Cross-Validated)

  • ✅ XBRL financials + transcripts both confirm demand durability: $500B+ Blackwell backlog confirmed
  • ✅ Ratios + forensic models confirm exceptional earnings quality (Piotroski 8/9, Z-Score 6.57)
  • ✅ Segments + transcripts agree on Data Center monopoly — both acknowledge concentration risk
  • ✅ Governance analysis + comp data confirm shareholder-friendly structure (entrenchment 2/10)
  • ✅ Advanced analysis + ratios confirm fortress balance sheet — interest coverage 502x, zero floating debt

Divergent Signals (Cross-Validated Conflicts)

  • ⚠ Beneish M-Score flags LIKELY MANIPULATOR while all other forensic models confirm exceptional quality — likely growth-driven false positive
  • ⚠ Management cited $50B China TAM in Q1 for policy advocacy; filings show regulatory harm already occurred and accelerating
  • ⚠ Transcripts cite no material ASIC threat; AMD's filing cites CUDA moat as their own competitive barrier — NVDA moat is real but the ASIC threat dismissal in transcripts is concerning
  • ⚠ Jensen's Q4 absolute conviction on compute-equals-revenues (10/10) conflicts with CFO Kress's hedged Rubin demand "too early to know" answer in the same quarter

Watchlist & Predictions

Monitoring Watchlist

Q1 FY2027 Revenue (vs $78B guide)
Next date: May 2026 earnings
A miss below $76B would signal Blackwell demand curve is peaking ahead of expectations. The first credible evidence of AI infrastructure supercycle inflection.
Vera Rubin Order Commentary (CFO Kress)
Next date: May 2026 earnings
If Kress still says "too early to know" on Rubin demand in Q1 FY2027 call, monitor for Q2 confirmation. Persistent opacity on the next product cycle is the highest-probability downside catalyst.
Singapore Geography Quarter-by-Quarter
Frequency: each quarter
Monitor trajectory for demand routing regulation signals. Growth should moderate if organic; acceleration suggests entity routing.
HBM Pricing from SK Hynix / Samsung
Frequency: ongoing, quarterly filings
HBM is est. 40-60% of Blackwell COGS. Price changes from memory suppliers flow directly to NVDA gross margins. The most important unmonitored cost driver.
Hyperscaler CapEx Guidance Updates
Frequency: each quarter
0.96 correlation with NVDA revenue at 1-year lag. Any reduction in Microsoft/Amazon/Google/Meta CapEx guidance is a leading indicator of NVDA demand risk.
New Export Control Actions (BIS/Commerce)
Frequency: ongoing
3-year escalation trajectory continues. Singapore routing could attract regulatory attention. AI Diffusion IFR enforcement has not yet been tested publicly.

Forward Predictions

FY2027 Revenue Will Exceed $300B Confidence: 70%

Basis: Q1 FY2027 guide $78B implies $312B annualized; Vera Rubin ramp + Blackwell Ultra + sovereign AI support continued expansion. Trigger: Q1 FY2027 below $76B would reduce confidence to 50%.

Gross Margins Stabilize 73-76% in FY2027 Confidence: 75%

Basis: FY2026 Q4 at 75%; Vera Rubin architecture transition may cause 2-3pp temporary compression similar to Blackwell ramp. HBM pricing is the key variable. Trigger: Two consecutive quarters below 72%.

Data Center Concentration Remains Above 88% Confidence: 90%

Basis: No credible diversification vector in 12-month horizon. Gaming, Automotive, and ProViz cannot grow fast enough to dilute Data Center share materially.

Export Controls Will Not Worsen from H20 Ban Levels Confidence: 55%

Basis: Trump administration generally business-favorable; Singapore routing creates political complexity. Uncertainty remains highest of all predictions — binary outcome with high-impact tails on both sides.

Share Count Will Decline 1-2% Annually Through FY2028 Confidence: 85%

Basis: $58.5B remaining on existing buyback authorization plus $60B additional authorization; buybacks have exceeded issuances for 3 consecutive years. Anti-dilutive buyback program firmly established.

Final Synthesis — The Complete Picture

After analyzing 6 annual 10-K filings (FY2021-FY2026), 8 quarterly 10-Q filings, 2 DEF 14A proxy statements, 99 Form 4 insider filings, 12 8-K events, 4 earnings call transcripts (FY2026 Q1-Q4), 5 forensic models, and 10 transcript investigation layers, here is what we know about NVIDIA Corporation.

NVIDIA is the most profitable hardware company in history at scale, having executed a transformation from gaming GPU supplier to AI infrastructure monopolist that created approximately $2.5 trillion in shareholder value in three years. The CUDA ecosystem lock-in is real — AMD acknowledges it as a competitive barrier to their own customers. The financial quality is exceptional: 89% ROIC, 75% gross margins, $64B operating cash flow in FY2025. The trajectory remains explosive but is mathematically decelerating, and the company faces three genuine structural risks: (1) extreme concentration in AI infrastructure that is subject to cyclicality and hyperscaler CapEx cycles; (2) escalating export controls that have already permanently eliminated billions in China revenue; (3) valuation that requires sustained 20-25% earnings CAGR for a decade to justify at 10% discount rates. The most important question investors are not asking: what happens to compute demand if AI efficiency gains make models dramatically cheaper to run? This is the only scenario that breaks the core investment thesis — and no analyst in NVIDIA's coverage universe is modeling it.

What the Numbers Say

  • FY2026 revenue $215.9B, net income $120.1B, operating cash flow $102.7B — 5th most profitable company in US history
  • Gross margin V-shape: 56.9% (FY2023 trough) → 75.0% (FY2025 peak) → 60.5% (Q1 FY2026 ramp) → 75.0% (Q4 FY2026 recovery)
  • ROIC 89.2% (FY2025) — extraordinary capital efficiency; interest coverage 330x
  • Piotroski F-Score 8/9 (STRONG BUY), Altman Z-Score 6.57 (firmly SAFE)
  • Data Center revenue: $10.6B (FY2022) → $115.2B (FY2025) → 88%+ of total revenue — unprecedented segment concentration
  • Singapore revenue 221% CAGR (FY2023-FY2025) to $23.7B — warrants geographic end-customer scrutiny

What Management Says

  • Management credibility 87/100; CFO guidance accuracy 92/100 (100% revenue beat rate, avg 4.7% above midpoint)
  • Systematic sandbagging score 8.5/10 — guidance is a managed floor, not a true forecast
  • Vera Rubin demand: both Jensen and Kress hedged across 3 quarters; "too early to know" on commercial trajectory
  • China $50B TAM invoked in Q1 for policy advocacy; silently retired by Q4 without resolution
  • HBM memory cost (est. 40-60% of COGS) has NEVER been quantified in 4 quarters of calls — most important unasked question
  • Analyst community capitulation to unanimous bullishness — no stress-testing of bear case across 31 Q&A pairs

Where They Agree

  • Both financial data and management confirm demand durability: $500B+ backlog, hyperscaler CapEx 0.96 correlation
  • All forensic models except Beneish confirm exceptional quality (Piotroski 8/9, Z-Score 6.57, cash conversion 88%)
  • Segments and transcripts both confirm Data Center monopoly and acknowledge concentration risk
  • Governance data and proxy confirm shareholder-friendly structure (entrenchment 2/10, single-class shares)

Where They Conflict

  • Beneish flags LIKELY MANIPULATOR while all other models confirm exceptional quality — growth-driven false positive, not genuine manipulation signal
  • Management's China optimism ($50B TAM, policy advocacy) conflicts with filing evidence of regulatory harm already occurred and H20 ban in Q1 FY2026
  • Jensen's absolute conviction on compute-equals-revenues conflicts with CFO Kress's hedged Rubin demand commentary in the same Q4 — internal management divergence
  • Insider selling ($1.83B collective in 10 months) contrasts with management's publicly stated long-term confidence narrative about AI infrastructure permanence

The Single Most Important Thing to Watch

Q1 FY2027 Revenue Currently: $78B guided Threshold: Below $76B or Kress still "too early to know" on Rubin

The guided $78B for Q1 FY2027 is the single most important near-term datapoint. A miss below $76B would signal the Blackwell demand curve is peaking ahead of expectations. Simultaneously, if CFO Kress still cannot characterize Rubin order demand by Q2 FY2027 (August 2026), it would suggest a product cycle gap that the sell-side is not modeling. Both conditions together would represent the first credible evidence that the AI infrastructure supercycle is approaching an inflection.