How does Nvidia make money?
Nvidia designs chips and has them manufactured by partners; it does not run its own chip factories. It started with graphics processing units (GPUs) for PC games, then found that the same chips, built to do many calculations at once, suit AI. Today it sells whole platforms: GPUs, central processors, networking, software and complete rack-scale systems.
In fiscal 2026 Nvidia reported five market platforms. Data Center: $193.7 billion, about 89.7% of revenue. Gaming: $16.0 billion, about 7.4%. Professional Visualization, workstation graphics for designers and engineers: $3.2 billion, about 1.5%. Automotive, chips and software for driver assistance and self-driving: $2.3 billion, about 1.1%. OEM and Other: $0.6 billion, about 0.3%. Its two reportable segments were Compute & Networking, $193.5 billion, and Graphics, $22.5 billion.
- GPU (graphics processing unit)
- A chip built to run many simple calculations in parallel. First made for video game graphics, it is now the main chip used to train and run AI models in data centers.
What does Nvidia’s data center business sell?
Nvidia splits Data Center revenue into compute and networking. Compute is the GPUs, processors and the systems built around them. Networking is the switches, network cards and interconnects that let thousands of GPUs work as one machine, using Ethernet, InfiniBand and Nvidia’s own NVLink.
In fiscal 2026 compute brought in $162.4 billion, up about 59% from $102.2 billion, and networking $31.4 billion, up about 142% from $13.0 billion. Networking is the smaller part but grew faster, because the newest systems are sold as connected racks rather than single chips.
Who are Nvidia’s biggest customers?
A small number of very large buyers. Nvidia’s 10-K says two direct customers each accounted for more than 10% of fiscal 2026 revenue: one for 22% and one for 14%, both mainly in Compute & Networking. That is about $47.5 billion and $30.2 billion, or about 36% of revenue from two buyers. Nvidia does not name them, and direct customers are often the companies that assemble servers, which then sell to cloud providers and others. It also estimates that some indirect customers, buying through those partners, individually reached 10% or more.
In the fourth quarter of fiscal 2026, according to Nvidia’s CFO commentary, hyperscalers, the largest cloud and internet companies, were its largest customer group at slightly over half of Data Center revenue. Concentration cuts both ways: a few buyers can add a lot of revenue quickly, and a pause by one of them can take a lot away.
- Hyperscaler
- One of the largest cloud and internet companies, which build and run data centers at enormous scale and buy chips and servers in very large volumes.
- Customer concentration
- When a large share of a company’s revenue comes from a few customers. US companies disclose any customer that brings in 10% or more of revenue, usually without naming it.
Why does hyperscaler capex matter for Nvidia stock?
Capital expenditures, or capex, are what a company spends on long-lived assets such as buildings and equipment; see free cash flow. For the big cloud and internet companies, a large part of that money now goes to AI data centers, and much of a data center’s cost is chips and networking.
Those companies publish their spending plans in their own earnings reports, often weeks before Nvidia reports. When they raise their capex outlook, traders read it as future demand for Nvidia; when they trim it, the reverse. Several of them are fellow members of the Magnificent Seven, so their reports can move Nvidia and the Nasdaq-100 together. A higher plan is not a promise of orders, and part of it goes to buildings, power and other suppliers.
When does Nvidia’s fiscal year end, and how did its reporting change?
Nvidia’s fiscal year ends on the last Sunday of January, so its quarter names run about a year ahead of the calendar. Fiscal 2026 ended on January 25, 2026, and fiscal 2027 runs from late January 2026 to late January 2027. Its first quarter ended on April 26, 2026 and was reported on May 20, 2026; later quarters end in late July, late October and late January, with results about a month after each.
From the first quarter of fiscal 2027 Nvidia reports two market platforms instead of five: Data Center, split into Hyperscale (public clouds and the largest consumer internet companies) and ACIE (AI clouds, industrial and enterprise), and Edge Computing, which covers PCs, game consoles, workstations, AI-RAN base stations, robotics and automotive. In that quarter Data Center brought in $75.2 billion and Edge Computing $6.4 billion, of $81.6 billion in total. Use the new categories when you compare recent quarters and the old ones for fiscal 2026 and earlier.
What should you watch in Nvidia’s earnings report?
Nvidia reports after the US market close, and the release comes with a CFO commentary that breaks revenue down further. Because of its weight in the indexes, the reaction can spread to chip stocks and QQQ. Read these each quarter:
- Data Center revenue and its split, Hyperscale against ACIE since fiscal 2027, and compute against networking where given.
- The revenue outlook for the next quarter, which the market often weighs more than the quarter just ended.
- Gross margin. It was 71.1% in fiscal 2026 against 75.0% in fiscal 2025, held down partly by a $4.5 billion first-quarter charge for H20 excess inventory and purchase obligations.
- Export rules: recent outlooks assumed no data-center compute revenue from China.
- Customer concentration in the 10-Q and 10-K, and inventory and supply commitments on the balance sheet; see how to read earnings reports.