NVIDIA
What it does
NVIDIA reports 2 businesses. In FY2026, compute and networking brought in 90% and graphics brought in 10% of revenue.
By region, United States made up 69%, then Taiwan at 20%.
From NVIDIA Corporation (SEC EDGAR), Form 10-K for the fiscal year ended 25 Jan 2026.
Our coverage
Read it through a lens
How it makes money
Where NVIDIA's FY2026 revenue came from, what it cost to earn, and what was left for shareholders.
- Income
- Costs and tax
- Profit
- % = share of revenue
Source: NVIDIA Corporation (SEC EDGAR), Form 10-K for the fiscal year ended 25 Jan 2026 (25 Feb 2026), statement of income, Note 16. FY2026 is the year to 25 Jan 2026. Figures as reported, in US$ millions in the tooltips and table.
Technical: gross margin was 71.1%, operating margin 60.4% and net margin 55.6%. Operating costs and cost of revenue together took 39.6% of revenue.
Plain: of every dollar of sales, about 29 cents paid for making or delivering what was sold, about 11 cents went on running and growing the business, and about 56 cents ended up as profit for shareholders.
Revenue by market
Revenue by business
Revenue by region
See the numbers
| FY2026 | US$M | % of revenue |
|---|---|---|
| Revenue | 215,938 | 100.0% |
| Cost of revenue | −62,475 | 28.9% |
| Gross profit | 153,463 | 71.1% |
| Research and development | −18,497 | 8.6% |
| Sales, general and administrative | −4,579 | 2.1% |
| Operating income | 130,387 | 60.4% |
| Other income (expense), net | 11,063 | 5.1% |
| Profit before tax | 141,450 | 65.5% |
| Income tax | −21,383 | 9.9% |
| Net profit to shareholders | 120,067 | 55.6% |
- In the first quarter, NVIDIA took a US$4.5B charge for excess H20 inventory and purchase obligations after the US required a licence to export H20 chips to China.
- Other income of US$11,063M is interest income of US$2,300M, less interest expense of US$259M, plus other income, net of US$9,022M.
- Data centre revenue splits into compute (US$162,361M) and networking (US$31,376M).
- Segment operating income adds up to US$139,297M. NVIDIA keeps US$8,910M of costs outside its two segments.
- NVIDIA assigns revenue to the country of the customer's headquarters. The end customer and shipping location may be elsewhere.
Moat rating
Wide moat
8.8 / 10
ROIC averaged 152.1% over FY2022 to FY2026, far above a 10.55% cost of capital, and beat it every year. FY2023's 20.4% shows how cyclical chips can be before the AI boom.
NVIDIA's returns are extraordinary right now. They were much lower three years ago, so the question is how long demand stays this strong.
The numbers
10.0 / 10Return on invested capital against cost of capital, semiconductors
- Average ROIC
- 152.1%
- Average excess return
- +141.5 pp
- Years above the hurdle
- 5 of 5
Spread 10 × 60% + durability 10 × 40% = 10.0
The sources
7.6 / 10Where NVIDIA's advantage comes from
- 1. Brand and pricing power9 / 10
Gross margin of 71.1% in FY2026, even after a US$4.5B charge for excess H20 inventory and purchase obligations.
NVIDIA's AI chips are in such demand that it keeps over 70 cents of every sales dollar after making them.
- 2. Barriers to entry8 / 10
Data Center revenue of US$193.7B, 90% of sales. Research and development was US$18.5B in FY2026.
Designing chips like these takes years of work and billions of dollars before the first sale.
- 3. Switching costs8 / 10
AI developers build on NVIDIA's CUDA software platform. Moving workloads to another chip means rewriting, retesting and retuning code.
AI software is written for NVIDIA's tools. Switching chips means redoing a lot of that work.
- 4. Network effects6 / 10
More CUDA developers attract more software libraries and models built for NVIDIA hardware, which draws more developers.
The more people build for NVIDIA, the more ready-made tools there are, which pulls in more people.
- 5. Economies of scale7 / 10
Revenue of US$215.9B in FY2026, with chips made by outside foundries rather than NVIDIA's own factories.
NVIDIA's size helps it book scarce factory capacity, but it depends on partners to make its chips.
Return on invested capital: NVIDIA Corporation (SEC EDGAR), Form 10-K for the fiscal year ended 30 Jan 2022; NVIDIA Corporation (SEC EDGAR), Form 10-K for the fiscal year ended 29 Jan 2023; NVIDIA Corporation (SEC EDGAR), Form 10-K for the fiscal year ended 28 Jan 2024; NVIDIA Corporation (SEC EDGAR), Form 10-K for the fiscal year ended 26 Jan 2025; NVIDIA Corporation (SEC EDGAR), Form 10-K for the fiscal year ended 25 Jan 2026. Hurdle: Aswath Damodaran, NYU Stern, Cost of capital by sector (US), data as of January 2026.
How we worked out ROIC: operating income after tax, divided by the capital invested at the start of the year. Tax is at the year's effective rate, or the 21% US federal rate where a pre-tax loss or a tax credit makes the effective rate meaningless. Invested capital is shareholders' equity plus borrowings, less cash and short-term investments, from the opening balance sheet in the same 10-K; lease liabilities are left out. This follows Damodaran's definition. The ratios are our calculation from the filed figures, not numbers the company reports. NVIDIA tags its January 2025 marketable securities (US$34,621M) with its own label, so we took that line from the FY2026 10-K balance sheet directly.
Lens scores are rasinbunlover's judgment, reviewed 19 Oct 2026. A wide moat is a reason to dig deeper, not a verdict on the price. For education only. Not financial advice.
How we rate moats ›Value NVIDIA with dbot
dbot has NVIDIA's latest figures (12 months to Jul 2026 (fiscal Q2 2027); cash includes $42.8B of equity securities; debt includes leases; price at 16 Oct 2026 close). It gives a value per share from Damodaran's methods and shows what today's price assumes. Every input can be changed.
Open NVIDIA in dbot ›Price data: TradingView, may be delayed according to exchange rules. A ticker on this site is not a recommendation. For education only. Not financial advice.