NVIDIA CORP (NVDA) 2025 Earnings Analysis
NVIDIA CORP2025 Earnings Analysis
73/100
NVIDIA's 71.1% gross margin on $215.9B revenue is the clearest evidence of monopoly-grade pricing power in semiconductors. The moat is CUDA — a software ecosystem with millions of developers and a decade of optimization that no hardware competitor can replicate. But 76.3% ROE at this scale raises cyclicality concerns: when hyperscalers' AI capex normalizes, will NVIDIA's earnings prove structural or cyclical? The $4.5B export control charge hints at moat fragility in geopolitically contested markets.
Core Dimension Scores
Evaluating competitive strength across earnings quality, moat strength, and risk sustainability
Overall Score Trend
Earnings Quality
Gross margin of 71.1% is exceptional for a semiconductor company but declined 3.9 percentage points from 75.0% in FY2025. The 10-K explicitly attributes this to two factors: 'our business model transitioned from offering Hopper HGX systems to Blackwell full-scale datacenter solutions' and 'a $4.5 billion charge associated with H20 excess inventory and purchase obligations.' The filing quantifies inventory provisions at $7.2B for FY2026 vs. $3.7B for FY2025, with the 'net effect on gross margin an unfavorable impact of 2.6%.'
Operating cash flow of $102.7B against net income of $120.1B yields a 0.86x ratio — below 1.0x, which warrants attention. The gap is partially explained by the $9.0B in unrealized gains on equity investments (non-marketable and publicly-held securities) included in Other income, which boost net income but do not generate cash. The 10-K notes 'Other income, net, compared to fiscal year 2025, was primarily driven by unrealized gains in non-marketable and publicly-held equity securities.' Excluding this non-cash income, cash conversion of core operations is strong.
Revenue surged 65% from $130.5B to $215.9B. The 10-K breaks down the drivers: 'Data Center revenue was up 68% from a year ago' with 'Blackwell architectures represented the majority of Data Center revenue.' Networking grew even faster: 'Revenue from Data Center networking grew 142% driven by the introduction and continued ramp of NVLink compute fabric for GB200 and GB300 systems.' Gaming was also strong at +41% 'driven by strong Blackwell demand,' and Professional Visualization rose 70% 'driven by exceptional demand for Blackwell as well as the launch of DGX Spark.'
Operating cash flow of $102.7B represents a 47.6% OCF margin — extraordinary for a semiconductor company. NVIDIA crossed the $100B OCF threshold for the first time, joining an elite group of only 3-4 companies globally. The 10-K notes operating expenses grew 41% 'driven by higher compensation and benefits expenses due to employee growth and compute and infrastructure costs,' yet OCF still expanded dramatically from the prior year's level.
Operating expenses at 10.7% of revenue (down from 12.6% in FY2025) demonstrate extreme operating leverage. R&D at 8.6% of revenue (vs. 9.9% prior year) and SGA at 2.1% (vs. 2.7%) both declined as percentages despite growing in absolute dollars. The 10-K discloses R&D increased 43% to $18.5B 'driven by a 29% increase in compensation and benefits expense reflecting employee growth' and '79% increase in compute and infrastructure costs.' The revenue base simply grew faster than spending.
Earnings quality scores 82/100 — phenomenal top-line growth and cash generation with two important caveats. The 65% revenue surge to $215.9B is driven by the Blackwell architecture transition that the 10-K calls 'the major platform shifts accelerated computing and AI,' with Data Center networking growing 142% on NVLink fabric demand. However, the 3.9-point gross margin decline to 71.1% reflects the structural reality that full-scale datacenter solutions carry lower margins than component sales, compounded by the $4.5B H20 charge. The 0.86x CF/NI ratio is distorted by $9B+ in unrealized investment gains — core operational cash conversion remains strong at $102.7B OCF.
Moat Strength
The 10-K states NVIDIA 'pioneered accelerated computing to help solve the most challenging computational problems' and has evolved into 'a data center scale AI infrastructure company reshaping all industries.' The Compute & Networking segment generated $193.5B in revenue (67% growth), with operating income of $130.1B. Data Center networking revenue grew 142% on NVLink adoption for GB200 and GB300 systems. This is not just chip selling — NVIDIA is providing full-stack datacenter solutions that create deep platform lock-in.
The 10-K states NVIDIA is 'bringing new advanced architectures on a one-year product cadence, including our Rubin platform.' The filing confirms 'we began shipping production units of our new Blackwell Ultra platforms including GB300 in the second quarter of fiscal year 2026.' This annual architecture refresh — Hopper to Blackwell to Blackwell Ultra to Rubin — creates a relentless competitive treadmill that makes it extremely difficult for AMD, Intel, or custom silicon to catch up.
Despite declining from 75.0%, a 71.1% gross margin for a semiconductor company selling full-scale datacenter solutions is remarkable. The 10-K notes that 'Blackwell architectures represented the majority of Data Center revenue,' confirming strong ASPs for the new platform. Excluding the $4.5B H20 charge, the underlying gross margin would be approximately 73%, demonstrating sustained pricing power. Customers are paying premium prices because NVIDIA's CUDA ecosystem creates prohibitive switching costs.
The 10-K discloses significant customer concentration: 'sales to one direct customer represented 22% of total revenue and sales to another direct customer represented 14% of total revenue' in FY2026, up from 12% and 11% in FY2025. The filing also notes that 'one AI research and deployment company contributed to a meaningful amount of our revenue purchasing cloud services from our customers.' This concentration — 36% from two direct customers alone — creates meaningful revenue risk if any key relationship changes.
Moat strength scores 88/100 — dominant platform position with a notable concentration vulnerability. The 10-K paints NVIDIA as having successfully transitioned from a chip company to 'a data center scale AI infrastructure company' with full-stack solutions. The one-year architecture cadence (Hopper -> Blackwell -> Blackwell Ultra -> Rubin) creates a competitive moat through relentless innovation. Data Center networking's 142% growth on NVLink for GB200/GB300 systems shows the moat extending beyond compute into networking fabric. However, the red flag is customer concentration: 36% of revenue from just two direct customers, with the filing acknowledging 'a limited number of partners and distributors' drive the business. The open-source AI risk is also real — the 10-K warns that 'the rise in high-quality open-source foundation models is making advanced AI capabilities broadly accessible' and could 'reduce demand for our products' if deployed on competing platforms.
Capital Allocation
Free cash flow of $96.7B represents a 44.8% FCF margin — among the highest of any company at any scale, ever. The gap between OCF ($102.7B) and FCF ($96.7B) reflects only $6.0B in capex, confirming NVIDIA's fabless model where TSMC bears the manufacturing capital burden. This $96.7B in annual distributable cash provides enormous strategic flexibility.
The 10-K reveals NVIDIA 'invested $17.5 billion in private companies and infrastructure funds, primarily to support early stage startups' in FY2026. The filing warns these 'investments include AI model makers that purchase our products directly or through CSPs' and are 'illiquid and non-marketable' with 'no assurance that we will realize a return.' Additionally, NVIDIA 'provided $3.5 billion in land, power, and shell guarantees to early stage companies.' This $21B in total commitments to startups that are also customers raises concerns about circular revenue and investment quality.
ROE of 76.3% is exceptional and driven primarily by operating efficiency rather than financial leverage (debt ratio only 23.9%). This reflects the fabless business model's capital efficiency — NVIDIA generates $120.1B in net income without owning fabrication facilities. The 10-K's disclosure of a 3.7% employee turnover rate (with 42,000 employees, 31,000 in R&D) suggests the human capital driving this ROE is stable and retained.
A 23.9% debt ratio is conservatively low for a company generating $96.7B in annual FCF. NVIDIA maintains a fortress balance sheet with minimal leverage, providing maximum strategic flexibility for acquisitions, investments, and navigating potential downturns. This conservative approach contrasts sharply with Apple's 79.5% leverage strategy.
Capital allocation scores 72/100 — the fabless model generates elite FCF but the ecosystem investment strategy raises questions. The $96.7B FCF (44.8% margin) on only $6.0B capex is the most capital-efficient model among mega-cap technology companies. The conservative 23.9% debt ratio and 76.3% ROE confirm genuine operating efficiency. However, the 10-K's disclosure of $17.5B invested in 'illiquid and non-marketable' startup investments — many of which are 'AI model makers that purchase our products' — is a significant concern. Combined with $3.5B in datacenter guarantees to early-stage companies, NVIDIA has committed $21B to entities whose revenue depends partly on buying NVIDIA products. This circular dynamic and the 10-K's own warning that 'there can be no assurance that we will realize a return' warrant the lower score despite extraordinary FCF generation.
Key Risks
The 10-K provides detailed disclosure on China export restrictions: 'In April 2025, the USG informed us that a license is required for exports of our H20 product into the China market. As a result, we incurred a $4.5 billion charge associated with H20 for excess inventory and purchase obligations.' The filing notes that while limited licenses were granted ('approximately $60 million in H20 revenue'), the H200 licensing program has generated zero revenue and requires 'an inspection process in the United States prior to any shipment' plus a '25% tariff upon importation.' Revenue from customers outside the U.S. declined from 41% to 31% of total revenue year-over-year, directly reflecting export control impact.
The 10-K warns: 'Our revenue is concentrated among a limited number of direct and indirect customers and this trend may continue.' In FY2026, two direct customers represented 22% and 14% of total revenue respectively — up from 12% and 11% in FY2025. The filing also estimates 'one AI research and deployment company contributed to a meaningful amount of our revenue purchasing cloud services from our customers.' This increasing concentration on hyperscaler demand creates binary risk if any single customer reduces orders, develops competitive alternatives, or shifts to custom silicon.
The 10-K documents gross margin declining from 75.0% to 71.1%, a 3.9-point compression. Beyond the $4.5B H20 charge, the structural driver is the transition 'from offering Hopper HGX systems to Blackwell full-scale datacenter solutions' — selling complete datacenter solutions inherently carries lower margins than selling GPU modules. Inventory provisions totaled $7.2B (vs. $3.7B prior year). The filing warns of potential 'revenue volatility, quality issues, increased inventory provisions, decreases in product yields and higher material costs' from the complexity of product transitions.
The 10-K discloses $17.5B invested in private companies, explicitly warning they are 'illiquid and non-marketable' with 'related early-stage startups may not become profitable in the near term, or at all.' The filing adds $3.5B in 'land, power, and shell guarantees to early stage companies' where NVIDIA 'may elect to assume the underlying leases' if partners default. Combined, this $21B exposure to early-stage AI companies — many of whom are also NVIDIA customers — represents a concentration of risk in an ecosystem whose long-term profitability is unproven.
The 10-K states 'our supply chain is mainly concentrated in Asia' but NVIDIA is 'expanding into the U.S. and Latin America' to 'add resiliency and redundancy.' The filing warns that 'our ability to increase manufacturing capabilities will depend on the local region's manufacturing ecosystem's capacity to ramp production supply to the required volume and on a timely basis.' Additionally, 'long manufacturing lead times' extend 'beyond twelve months for some products,' and NVIDIA places 'non-cancellable inventory orders for certain product components in advance of historical lead times.' This creates demand mismatch risk in a rapidly evolving market.
Risk profile scores 48/100 (higher = safer) — the lowest module score, reflecting multiple elevated and interconnected risks. The 10-K's disclosure of the $4.5B H20 charge from U.S. export controls demonstrates that government action can eliminate entire product lines overnight. International revenue dropped from 41% to 31% of total, quantifying the China impact. Customer concentration worsened — two customers now account for 36% of revenue, up from 23% last year. The $17.5B in startup investments ('illiquid and non-marketable') plus $3.5B in datacenter guarantees create circular ecosystem risk. Gross margin compression of 3.9 points signals the structural cost of transitioning to full-scale datacenter solutions. The 10-K's repeated warnings about 'revenue volatility,' 'supply constraints,' and 'complex product transitions' paint a picture of a company growing at extraordinary rates but accumulating risk at a comparable pace.
Management
Ask about this section
This analysis is for educational purposes only and does not constitute investment advice.