Designs the processors and systems that run most artificial-intelligence computing; almost nine-tenths of its revenue now comes from data centres.
Revenue (FY2026)$215.9 billionFiscal year ended 25 Jan 2026
Net income (FY2026)$120.1 billionFiscal year ended 25 Jan 2026
Net margin (FY2026)55.6%Net income ÷ revenue
Founded1993
Overview
NVIDIA designs graphics processing units (GPUs) and the surrounding hardware and software used to run demanding computing workloads. Its chips were originally built for video-game graphics; the same parallel-processing design proved well suited to training and running artificial-intelligence models, and sales to data-centre operators now dominate its business.
The company does not manufacture its own chips. Its annual report describes a "fabless" model: it designs products and relies on contract manufacturers, TSMC and Samsung, to fabricate them, on SK hynix, Micron and Samsung for memory, and on other partners for assembly. It describes this supply chain as mainly concentrated in Asia.
Full fiscal years from each year's annual report (Form 10-K) filed with the U.S. SEC, in US dollars under US GAAP. These are not quarterly or trailing-twelve-month figures. Latest: FY2026, filed 25 Feb 2026.
Values in US dollars as originally filed (US GAAP). Net income is the profit attributable to the company's shareholders. Net margin is net income divided by revenue, calculated by Kazamuki. † A later annual report restated this figure; see the note below the table.
NVIDIA's fiscal year ends in late January, so "fiscal 2026" covers roughly February 2025 to January 2026. Fiscal years are labelled as the company names them.
Share price, market value and valuation ratios are not shown: they require licensed market data.
How it makes money
NVIDIA sells processors, complete server systems and networking equipment to cloud providers, enterprises and computer makers, and graphics cards to consumers through partners. A long-developed software platform (CUDA) encourages developers to build for its hardware, which reinforces demand. Sales are concentrated: two direct customers accounted for 22% and 14% of revenue in fiscal 2026.
Main activities
Data centre. AI accelerators, server systems and networking for cloud providers and enterprises. About 90% of fiscal 2026 revenue.
Gaming. GeForce graphics cards and related software for PCs. About 7% of revenue.
Professional visualisation. Workstation graphics for design, engineering and media.
Automotive. Computing platforms for driver assistance and in-vehicle systems.
Business mix
Share of revenue, Fiscal 2026 (year ended 25 January 2026), as reported by the company.
NVIDIA: share of revenue by segment, Fiscal 2026 (year ended 25 January 2026)
Segment
Share
Proportion
Compute & Networking Data-centre computing and networking platforms, automotive and related software.
89.6%
Graphics GeForce GPUs for gaming and PCs, workstation graphics and related software.
10.4%
NVIDIA reports two segments. The four activities above are its "market platforms", a different way of dividing the same revenue.
Source: Form 10-K, fiscal 2026. Percentages are calculated by Kazamuki from the reported figures.
Recent developments
Dated events from the company's own announcements and filings. This is a record of what was reported, not a news feed, and it is updated when the profile is reviewed.
Third-quarter results were signed by chief financial officer Colette Kress as interim chief executive. The company referred to a search for a new chief executive and an ongoing investigation by an independent committee of the board, without giving details.
Structural sensitivities described in the company's own reporting and official sources. These are not predictions.
Manufacturing concentration
Chips are fabricated mainly in Taiwan and memory comes largely from South Korea, so production depends on a small number of sites and suppliers.
Export controls
US rules restrict sales of the most capable AI chips to China. In April 2025 a new licence requirement for one product led to a $4.5 billion charge; the rules have since been revised more than once.
Customer concentration
Demand depends heavily on the capital-spending plans of a few very large buyers, which are themselves borrowing to fund it.
Leadership
The company's November 2026 results were issued under an interim chief executive, with a search for a successor and an investigation by an independent board committee under way. It has given no further detail.
Four US technology companies spent more than $200 billion on plant and equipment in a single quarter of 2026. Their filings show where the money comes from, where it goes, and why regulators are watching.
Chips now make up about half of South Korea's exports, Samsung and SK hynix are earning record profits from AI memory, and the Bank of Korea has raised rates twice. The boom is real; so is the dependence on one product cycle.
An advanced processor is designed in the United States, printed with Dutch machines, fabricated in Taiwan, paired with Korean memory and made with Japanese materials. Who does what, how concentrated each step is, and what the estimates can and cannot tell us.