NVDA 10-K Deep Dive — NVIDIA Corporation FY2026

Skill: financial-report-analyst · Filing: Annual Report on Form 10-K · Fiscal year ended: January 25, 2026 Report date: July 12, 2026 · Generated by Claude Code using the InvestSkill us-stock-analysis plugin. Data source: All figures extracted directly from NVIDIA’s FY2026 Form 10-K (SEC filing). US dollars, in millions unless noted. ⚠️ Market price not independently verified. Valuation commentary below is illustrative, derived from prices disclosed in the filing (Q4 FY2026 buyback range $179–$199). Verify a live quote before any decision. Educational analysis only — not investment advice.

Table of Contents

1. Executive Summary
2. Financial Health Dashboard
3. Visual Dashboard — Charts
4. Key Insights — Professional Investor Lens
5. Document Orientation
6. MD&A Deep Read — Revenue, Segments & Geography
7. Income Statement Analysis
8. Balance Sheet Analysis
9. Cash Flow & Capital Return
10. Deeper Analysis — Growth, Margins & Returns
11. Scenario Analysis — FY2027 (illustrative)
12. Competitive Context
13. Risk Factors — What Changed This Year
14. Footnotes & Accounting Quality
15. Management Credibility & Tone
16. Leading Indicators to Monitor
17. Key Positives
18. Key Negatives / Red Flags
19. Accounting Quality Score
20. Illustrative Valuation Context
21. Final Suggestions — Professional Investor View
22. Thesis Invalidation & Investment Signal

Executive Summary

NVIDIA delivered one of the largest single-year absolute earnings expansions in corporate history: revenue grew +65.5% to $215.9B and net income reached $120.1B (diluted EPS $4.90), driven almost entirely by Data Center (89.7% of revenue, +68% YoY) as Blackwell scaled. The two things a professional should hold in tension: (1) the business is a cash machine — $96.7B of free cash flow, 60.4% operating margin, and a $54B net-cash balance sheet with an unqualified PwC audit and no restatements; and (2) the quality of that growth now carries three material overhangs — effective foreclosure from China’s data-center market (a $4.5B H20 charge this year), customer concentration (two direct customers = 36% of revenue), and a ~390 bps gross-margin decline during the Hopper→Blackwell transition. This is a best-in-class compounder whose fundamentals are unambiguously bullish, but whose risk profile has become more concentrated and more politically exposed.

Financial Health Dashboard

Revenue Growth:    +65.5% YoY   (FY26 $215.9B vs FY25 $130.5B; substantially all organic)
Gross Margin:      71.1%        (vs 75.0% prior, Δ −390 bps; mostly Blackwell mix — ex-provisions ≈ 73.7%)
Operating Margin:  60.4%        (vs 62.4% prior, Δ −200 bps)
Net Margin:        55.6%        (vs 55.8% prior, Δ −20 bps)
FCF Margin:        44.8%        ($96.7B FCF; conversion 80.5% of net income, ~86% ex-investment gains)
Net Cash:          +$54.1B      (cash & securities $62.6B − total debt $8.5B; Debt/EBITDA ≈ 0.06x)
SBC / Revenue:     3.0%         ($6.4B — below the 5% watch line)
Management Tone:   CONFIDENT / POSITIVE  (with explicit, repeated China-export hedging)

Visual Dashboard — Charts

Revenue trajectory — the largest three-year ramp in tech history

REVENUE ($B, fiscal year)                            each █ ≈ $10B
FY2024   $60.9   ██████
FY2025  $130.5   █████████████                        +114% YoY
FY2026  $215.9   ██████████████████████               +65% YoY

Net income & diluted EPS

NET INCOME ($B)                                      each █ ≈ $5B
FY2024   $29.8   ██████            EPS $1.19
FY2025   $72.9   ███████████████   EPS $2.94
FY2026  $120.1   ████████████████████████  EPS $4.90  (+67%)

Revenue mix by end market (FY2026) — a Data Center company

Data Center  89.7%  ███████████████████████████████████████████  $193.7B
Gaming        7.4%  ███▌                                          $16.0B
Pro Viz       1.5%  ▊                                             $3.2B
Automotive    1.1%  ▌                                             $2.3B
OEM & Other   0.3%  ▏                                             $0.6B
                    (each █ ≈ 2% of revenue)

Margin trend — record scale, first margin give-back

%  (scale starts at 45% to show the moves)
75 ┤            ╭── 75.0 ─╮
   │            │          ╰─ 71.1   ← Gross margin
70 ┤            │
   │       ╭─ 62.4 ─────── 60.4      ← Operating margin
60 ┤       │
55 ┤       │    ╭─ 55.8 ── 55.6      ← Net margin
   │  ╭ 54.1
50 ┤  │ 48.8
45 ┼──┴────┬──────────┬─────────
     FY2024        FY2025      FY2026

Operating vs free cash flow

CASH FLOW ($B)                                       each █ ≈ $5B
        OCF                          FCF
FY2024  $28.1  ██████               $27.0  █████▌
FY2025  $64.1  █████████████        $60.9  ████████████▏
FY2026 $102.7  █████████████████████ $96.7 ███████████████████▍

Geographic revenue mix (customer-HQ basis) — China halved

                 FY2025      FY2026     Shift
United States    59.4%   →   69.3%      ▲ +9.9 pts
Taiwan           18.1%   →   19.6%      ▲ +1.5 pts
China (incl HK)  19.2%   →    9.1%      ▼▼ −10.1 pts   (revenue −21% YoY)
Other             3.3%   →    2.0%      ▼ −1.3 pts

Segment scorecard (FY2026)

SEGMENT                Revenue   YoY    Op. margin   Op. income
Compute & Networking   $193.5B   +67%   67.3%  █████████████████  $130.1B
Graphics                $22.5B   +57%   40.8%  ██████████         $9.2B

Capital returned to shareholders

CAPITAL RETURN ($B)                                  each █ ≈ $4B
FY2024  Buybacks  $9.5  ██▍          Dividends $0.4
FY2025  Buybacks $33.7  ████████▍    Dividends $0.8
FY2026  Buybacks $40.4  ██████████   Dividends $1.0
        → $60B new authorization (Aug 2025); $58.5B still remaining

Key Insights — Professional Investor Lens

1. The growth is real and cash-backed, not accrual-driven. Operating cash flow was $102.7B against $120.1B net income; DSO held flat at ~65 days despite +65% revenue, and goodwill is only 10% of assets. There is no evidence of aggressive revenue recognition or balance-sheet inflation. This is the single most important quality check — and it passes.

2. Data Center is now the whole story — and so is its concentration. At 89.7% of revenue, NVIDIA is effectively a Data Center company with a Gaming side-business. Within that, Networking (+142% to $31.4B) has emerged as a credible second pillar, which partially diversifies the “GPU-only” narrative.

3. China is now a subtracted market, not a growth market. China (incl. HK) revenue fell −21% to $19.7B, and management states outright it is “effectively foreclosed from competing in China’s data center… market” (p10). The $4.5B H20 write-down is the accounting expression of that policy shift. Upside from any China re-opening is optionality, not a base-case.

4. Margin normalization is mostly mix/transition, not competitive erosion. Ex inventory/purchase provisions, gross margin was ~73.7% (vs ~77.3% prior) — a ~360 bps decline that is overwhelmingly the shift from selling Hopper boards to full Blackwell rack-scale systems (more third-party content passed through), not the H20 charge (provisions worsened only ~30 bps YoY) and not price competition. Watch whether Blackwell/Rubin margins recover toward the mid-70s as the ramp matures — management guided to that intent.

5. Capital-return firepower is enormous. $40.4B of buybacks in FY2026 plus a fresh $60B authorization ($58.5B remaining), against a token $0.04 dividend. This is a buyback story, and at ~24.3B shares the repurchases are already shrinking the count.

6. The $95.2B supply/purchase-commitment stack is the double-edged sword. It signals confidence in demand and locks in scarce TSMC/CoWoS capacity — but in a 1-year product cadence (Blackwell → Blackwell Ultra → Rubin) it is also the source of the inventory-obsolescence risk that PwC flagged as the sole critical audit matter.

Document Orientation

Verdict: Clean orientation. A Big-4 unqualified opinion, no restatements, a long-tenured auditor, and a single, well-understood CAM. Nothing here is a red flag on its own.

MD&A Deep Read — Revenue, Segments & Geography

Revenue by reportable segment

Revenue by end market (specialized market)

Revenue by geography (basis changed FY2026 → customer HQ location)

⚠️ Segment-reporting change: In Q3 FY2026 NVIDIA switched geographic revenue to customer-headquarters location (prior periods recast). The former bill-to-Singapore concentration is no longer shown; the ship-to caveat now attaches to Taiwan. This is disclosed and reasonable, but it reduces comparability with older filings — note it when trend-lining geography.

Margin trend: Gross margin compressed from 75.0% → 71.1% (−390 bps), of which ~260 bps is the direct H20 charge/inventory impact and the remainder is the Hopper→Blackwell rack-scale mix shift. Operating margin −200 bps to 60.4% as R&D grew +43% (investing through the cycle). Direction: compressing, but for identifiable, largely transitional reasons — not price competition.

Management tone: Confident and specific on demand (“driven by the major platform shifts – accelerated computing and AI,” p37), with deliberate hedging isolated to China/export policy and near-term Gaming supply (“We expect supply constraints to be a headwind to Gaming in the first quarter of fiscal 2027 and beyond,” p37). Forward guidance on capex is explicit and directional: “We expect to increase capital expenditures in fiscal year 2027” (p43).

Income Statement Analysis

Quality-of-earnings notes:

• Growth is organic. No large revenue-contributing acquisitions; the increase is unit/ASP-driven Data Center demand.
• SBC is well-contained at $6.4B (3.0% of revenue) — below the 5% watch threshold and falling as a percentage.
• ⚠️ Other income, net jumped to $9.0B (from $1.0B) — driven largely by gains on non-marketable equity investments. This is non-operating and lower-quality. It added roughly $7.6B after-tax (~$0.31 of EPS). Strip it out and pre-tax income still rose ~+56%, so the operating story doesn’t need it — but headline EPS growth is modestly flattered by investment marks that may not repeat.
• Effective tax rate rising (12.0% → 13.3% → 15.1%) — a modest structural headwind to net-income growth going forward, but no one-time distortions.

Balance Sheet Analysis

• Net cash position: +$54.1B ($62.6B liquidity − $8.5B debt); Debt/EBITDA ≈ 0.06x. Leverage is a non-issue.
• DSO discipline is the standout tell: AR grew +66.8% vs revenue +65.5% → days-sales-outstanding essentially unchanged (~65 days). No channel-stuffing signature.
• ⚠️ Two line items grew far faster than revenue and deserve monitoring: inventory (+112%) and non-marketable equity securities (+557%). The first is the obsolescence/CAM risk; the second converts operating cash into illiquid, mark-to-model strategic stakes whose gains now flow through Other income.

Cash Flow & Capital Return

• FCF conversion 80.5% clears the >80% quality bar. The optical dip vs prior years is caused by (a) the $11.3B inventory build (working-capital drag) and (b) the $9B non-cash investment gain inflating the net-income denominator. Adjusting for the investment gains, “operating” conversion is ~86% — healthy.
• Capital return: $40.4B repurchased (282M shares) in FY2026; a new $60B authorization was approved Aug 26, 2025 with $58.5B remaining. Dividend is symbolic ($0.04/share; $974M total). NVIDIA also withheld ~$7.9B of shares for RSU tax settlement.
• Capital deployment is shifting toward strategic investments ($17.5B non-marketable equity + $13.0B Groq license). This is empire-building through the ecosystem — potentially value-accretive (locking in demand/IP), but it raises related-party and earnings-quality scrutiny.

Deeper Analysis — Growth, Margins & Returns

Growth decomposition — where the $85B came from

Of the +$85.4B of incremental revenue in FY2026, Data Center supplied $78.6B (92%). Within Data Center, Compute added $60.2B and Networking added $18.4B (+142%) — the fastest-growing line in the company and the clearest evidence of a diversifying second engine.

INCREMENTAL REVENUE FY2025 → FY2026 (+$85.4B total)      each █ ≈ $4B
Data Center — Compute     +$60.2B  ███████████████
Data Center — Networking  +$18.4B  █████
Gaming                     +$4.7B  █▏
Pro Viz / Auto / OEM       +$2.1B  ▌

Quality of growth: organic (no material acquired revenue), volume/ASP-driven, and cash-backed (OCF $102.7B vs net income $120.1B). This is the highest-quality growth signature a filing can show.

Gross-margin bridge (FY2025 → FY2026)

GROSS MARGIN BRIDGE
FY2025 reported                     75.0%
  Blackwell rack-scale mix shift    −3.6 pts   Hopper boards → full systems (pass-through content)
  Inventory / purchase provisions   −0.3 pts   2.6% drag vs 2.3% prior (incl. the $4.5B H20 charge)
FY2026 reported                     71.1%
  Memo — ex-provisions level:      ~73.7%   (vs ~77.3% prior)

The decline is a business-model story, not a pricing story: selling full liquid-cooled racks passes more third-party content (memory, networking, integration) through revenue at a lower percentage margin, even as gross-profit dollars rose +57% to $153B.

Returns on capital — best-in-class

DuPont ROE (FY2026, on average equity $118.3B)
  Net margin        55.6%
  × Asset turnover   1.36x
  × Equity multiplier 1.35x
  = ROE             ~102%
ROIC (NOPAT $110.7B / invested capital ~$103B)  ≈ 107%

Returns this high are not sustainable indefinitely (they invite competition and mean-reversion), but they finance the R&D and buybacks internally with zero net leverage.

Working capital & cash conversion cycle

CASH CONVERSION CYCLE (days)         FY2025   FY2026     Δ
Days sales outstanding (DSO)          64.5     65.0     +0.5   clean — in line w/ revenue
Days inventory outstanding (DIO)     112.7    125.0    +12.3   pre-build for Blackwell/Rubin
Days payable outstanding (DPO)        70.6     57.3    −13.3   paying suppliers faster
Cash conversion cycle (CCC)          106.6    132.7    +26.1

CCC lengthened ~26 days — NVIDIA is tying up more cash in inventory ahead of the ramp and paying suppliers faster (securing scarce TSMC/CoWoS capacity). This is a growth investment, not distress, but it is the working-capital face of the obsolescence risk PwC flagged.

Efficiency — Rule of 40 (and then some)

Revenue growth  65.5%
+ FCF margin    44.8%
= 110.3   (threshold for "healthy" scale-up is 40)

R&D intensity is 8.6% of revenue ($18.5B) — heavy absolute spend, modest as a percentage, and self-funded.

Scenario Analysis — FY2027 (illustrative)

Not guidance and not in the filing — an illustrative framework for how a professional would bracket the next year, anchored on FY2026 revenue of $215.9B.

FY2027 REVENUE SCENARIOS (illustrative)               vs FY2026 $215.9B
Bear   +15%   ~$248B   China stays closed; hyperscaler capex digestion; Rubin slips
Base   +35%   ~$291B   Blackwell/Ultra ramp + networking strength; Rubin on time (2H)
Bull   +55%   ~$335B   Rubin cycle + sovereign-AI demand + partial China re-open

The spread is wide because ~90% revenue concentration in one end-market makes NVIDIA a high-beta expression of the AI-capex cycle. Position sizing should reflect that the bear case is a plateau at record levels — not a collapse.

Competitive Context

• The moat is the software, not just the silicon. CUDA + CUDA-X + the installed base is the switching-cost barrier; competitors can match FLOPs faster than they can match the developer ecosystem (a point management makes explicitly re: China foreclosure aiding rivals’ ecosystems, p10).
• Merchant-GPU competition: AMD (Instinct MI-series) is the closest merchant alternative; Intel is a distant third in AI accelerators.
• The real structural threat is custom silicon: hyperscaler in-house ASICs (Google TPU, Amazon Trainium/Inferentia, Microsoft Maia, plus Broadcom/Marvell design wins). This is why Networking (NVLink/InfiniBand/Ethernet) and full rack-scale systems matter — NVIDIA is selling the whole computer, which is harder to displace than a chip.
• Bottom line: NVIDIA is the clear leader with a widening software moat, but the ~$4T+ valuation already prices leadership; the competitive question is margin durability, not market-share loss, over the next 1–2 years.

Risk Factors — What Changed This Year

The biggest signal in any 10-K is what’s new or newly emphasized. This year that is unambiguously China / export controls.

New strategic disclosures: a non-exclusive Groq IP license and a still-unsigned OpenAI investment/partnership — “There is no assurance that we will enter into an investment and partnership agreement with OpenAI or that a transaction will be completed” (p22).

Footnotes & Accounting Quality

• Critical Audit Matter (inventory). PwC’s single CAM is the valuation of inventory reserves and excess-purchase-commitment provisions. NVIDIA recorded $4.0B of inventory provisions in cost of revenue (vs $1.6B prior); total provisions for inventory + excess purchase obligations were $7.2B including the $4.5B H20 charge, a net 2.6% unfavorable gross-margin impact. This is disclosed transparently — but it confirms obsolescence is the key accounting judgment to watch.
• Purchase & supply obligations: $95.2B (mostly payable through FY2027), plus $27B of multi-year cloud commitments and $11.4B of investment commitments. These off-balance-sheet obligations are large relative to even NVIDIA’s cash — a demand bet, not a solvency risk, but sizeable.
• Other income quality. The $9.0B Other income (largely non-marketable equity gains) is the main earnings-quality caveat; it is non-recurring in nature and should be normalized out when modeling forward earnings.
• Related-party / ecosystem investments. Rapidly growing stakes (Groq, OpenAI pending, $22.3B non-marketable equity) blur the line between customer, supplier, and investee — a governance/earnings-quality area to monitor, though currently well-disclosed.
• Revenue recognition: no policy changes flagged; DSO stability corroborates clean recognition.

Management Credibility & Tone

• Track record: NVIDIA has consistently met or exceeded its own guidance through the AI ramp; FY2026 continued the pattern with +65% revenue growth.
• Language discipline: Good news is stated in active voice with specifics; the only sustained hedging (“headwind,” licensing uncertainty, “no assurance”) is reserved for China policy and the OpenAI deal — i.e., hedging is targeted and credible, not diffuse.
• Capital-allocation signaling: A fresh $60B buyback authorization during a period of record valuation signals management confidence in forward cash generation. Rising capex guidance signals conviction in the Rubin cycle.

Assessment: HIGH credibility. The tone is confident where the data supports it and cautious precisely where the exogenous (political) risk lives.

Leading Indicators to Monitor

The metrics that will confirm or break the thesis before they show up in headline EPS:

Key Positives

• ✅ Revenue +65.5% to $215.9B; net income $120.1B; diluted EPS $4.90 (+66.7%) — organic, unit-driven growth.
• ✅ Elite profitability: 71.1% gross / 60.4% operating / 55.6% net margin.
• ✅ $96.7B free cash flow, 44.8% FCF margin, conversion ~86% ex-investment gains.
• ✅ Fortress balance sheet: +$54B net cash, Debt/EBITDA ~0.06x, goodwill just 10% of assets.
• ✅ Clean accounting: unqualified PwC opinion, no restatements, flat DSO, low SBC (3.0%).
• ✅ Second growth pillar emerging: Networking +142% to $31.4B.
• ✅ Huge, disciplined capital return: $40.4B FY26 buybacks; $58.5B remaining authorization.
• ✅ Product roadmap intact: Blackwell scaled; Blackwell Ultra shipping; Rubin unveiled (production 2H FY2027, up to 10× cost/token improvement).

Key Negatives / Red Flags

• 🔴 HIGH — China foreclosure: effectively locked out of China DC; $4.5B H20 charge; revenue −21% in the region; competitor ecosystems strengthening.
• 🔴 HIGH — Customer concentration: two direct customers = 36% of revenue; loss/slowdown of one would be material.
• 🟠 MED-HIGH — Policy/tariff overhang: 15% USG revenue-share expectation, 25% H200 import tariff, “GAIN AI Act.”
• 🟠 MED — Margin normalization: −390 bps gross margin; monitor recovery through the Blackwell/Rubin ramp.
• 🟠 MED — Inventory/commitment risk: inventory +112%; $95.2B purchase obligations in a 1-year product cadence.
• 🟠 MED — Earnings quality: $9.0B non-operating investment gains flatter EPS; growing illiquid strategic stakes.
• 🟡 LOW-MED — Legal/antitrust: multi-jurisdiction probes, China Mellanox preliminary finding, remanded securities suit — no accruals yet.

Accounting Quality Score

8.5 / 10 — High.

Deductions are for earnings-quality noise from investment gains and the scale of purchase commitments — not for any sign of manipulation. The statements themselves are clean and cash-corroborated.

Illustrative Valuation Context

The 10-K does not set a price, but it discloses Q4 FY2026 buybacks executed at average prices of $179–$199. Using ~$185 as a reference and 24.3B shares:

These multiples are illustrative only and depend on the live price, which must be verified. The takeaway: even a world-class compounder priced at ~37× trailing earnings / ~46× FCF leaves little margin for a growth deceleration or a demand air-pocket — which is exactly why the concentration and China risks matter for sizing and entry, not just for the narrative.

Final Suggestions — Professional Investor View

1. Own the quality, respect the price. The business earns a bullish fundamental verdict — durable moat (CUDA + full-stack), 60% operating margins, $97B FCF, net cash. But at ~37× trailing earnings the stock has priced in continued hyper-growth. Treat this as a buy/accumulate on a fundamentals basis, with price-disciplined entry, not a chase.
2. Size for concentration. With two customers at 36% of revenue and Data Center at ~90%, position size should reflect single-customer / single-end-market risk. This is not a “set-and-forget” full-weight holding.
3. Underwrite China as optionality, not base case. Model near-zero China DC revenue; any re-opening is upside. Do not pay for a China recovery.
4. Watch the margin trajectory as the tell. A gross-margin recovery toward the mid-70s through the Blackwell/Rubin ramp would confirm the −390 bps was transitional. Continued compression would be the first sign of pricing pressure — the thing to actually worry about.
5. Normalize EPS. Strip the $9B investment gains when setting forward estimates and valuation; don’t extrapolate mark-to-model windfalls.
6. Track the commitment stack. $95.2B in purchase obligations is the demand barometer and the obsolescence risk simultaneously — rising commitments = confidence; a build in inventory provisions = the early warning.

Thesis Invalidation & Investment Signal

This BULLISH (fundamentals) thesis breaks if:

• Gross margin continues to compress for a second year without an identifiable one-time cause (signals pricing pressure / competition, not transition).
• A top-2 customer materially cuts orders, or Data Center growth stalls (hyperscaler capex digestion / in-house silicon substitution).
• Auditor issues a going-concern or material restatement (none currently) — immediate exit.
• Export/tariff regime worsens further and spreads beyond China (e.g., broad allied-market restrictions).
• Free-cash-flow conversion falls below ~70% on a sustained, non-working-capital basis.

Re-run this analysis when:

• [ ] Next quarterly (10-Q) / annual (10-K) filing
• [ ] Any change in China export-license status or a signed OpenAI agreement
• [ ] A top customer disclosure changes (concentration ↑/↓)
• [ ] Price moves ±15% or 60 days elapse

╔══════════════════════════════════════════════╗
║              INVESTMENT SIGNAL               ║
╠══════════════════════════════════════════════╣
║ Signal:      BULLISH (fundamentals)          ║
║ Confidence:  HIGH (quality) / MED (near-term)║
║ Horizon:     LONG-TERM                       ║
║ Score:       8.5 / 10  (business quality)    ║
╠══════════════════════════════════════════════╣
║ Action:      BUY / ACCUMULATE (price-disc.)  ║
║ Conviction:  MODERATE                         ║
╚══════════════════════════════════════════════╝

Score Guide: 8.0–10.0 Strongly Bullish | 6.0–7.9 Moderately Bullish | 4.0–5.9 Neutral | 2.0–3.9 Moderately Bearish | 0.0–1.9 Strongly Bearish. Score reflects business/filing quality; it is not a valuation-adjusted price target — verify the live quote and apply your own entry discipline.

Disclaimer: Educational analysis of a public SEC filing only. Not financial advice. All figures sourced from NVIDIA’s FY2026 Form 10-K (fiscal year ended January 25, 2026); market-price commentary is illustrative and unverified. Do your own research and verify live data before making any decision.