Nvidia's top three direct customers drive 44% of revenue

Three Nvidia direct customers generated 16%, 15% and 13% of revenue in the first half of fiscal 2027, or 44% combined. A year earlier, only two direct customers crossed the company's 10% disclosure threshold, at 20% and 15%. At July 26, five customers also accounted for 70% of Nvidia's accounts receivable.

The disclosure measures direct purchasing relationships, which can be mistaken for dependence on individual Big Tech end users. Nvidia defines direct customers to include cloud service providers and AI model makers, but also add-in-board manufacturers, distributors, original-design manufacturers, original-equipment manufacturers and system integrators. The entity placing the order can be an intermediary serving several downstream users.

Nvidia's separate market-platform disclosure shows how different the two measures are. Hyperscale revenue was $91.761 billion in the first half of fiscal 2027 against $177.837 billion of total revenue, or 51.6%. In the recast prior-year period, hyperscale revenue was $46.428 billion against $90.805 billion, or 51.1%. Hyperscale's share moved by less than half a percentage point while direct-customer concentration became much more visible.

Concentrated orders increase credit and bargaining exposure

An ODM can buy GPUs for a cloud operator, a system integrator can serve several model developers, and a distributor can aggregate smaller buyers. One direct account can represent multiple end users and workloads.

The three purchasing relationships account for about $78.2 billion, or 44% of first-half revenue. Calling that "$78 billion at risk" assumes the buyers are three ultimate customers that could independently withdraw all that demand. The disclosure does not establish this.

Cash collection is concentrated too. Nvidia reported five direct customers at 22%, 14%, 13%, 11% and 10% of accounts receivable at July 26, or 70% combined. At the January year end, the top three disclosed customers represented 25%, 18% and 13%, or 56%. The identity and composition of the group can change between periods.

Nvidia also says that, for some large data-center builds involving investment-grade customers, it may provide payment terms ranging from 90 days to one year. Longer terms can be commercially rational when a customer's infrastructure build is large and credit quality is strong. They also move more working-capital exposure onto Nvidia as the absolute revenue base expands.

First-half revenue nearly doubled to $177.8 billion from $90.8 billion, increasing the dollars exposed even when credit terms stay unchanged. Large buyers can also gain bargaining power over delivery schedules, configuration, support and payment timing, particularly when they coordinate data-center programs over several quarters.

Direct-customer shares do not measure custom-chip substitution

The same distinction matters in the ASIC debate. Hyperscalers are designing more custom AI chips, but a direct customer's 16% share of Nvidia revenue does not reveal how much of that buyer's downstream workload could shift to an internal accelerator. The customer might be an ODM serving several hyperscalers, or a cloud provider whose Nvidia and custom-silicon spending both rise because total AI infrastructure expands.

Hyperscalers still account for roughly half of Nvidia's revenue, while orders and receivables have become more concentrated among direct buyers. That increases sensitivity to their purchasing schedules, bargaining and payment behavior. Identifying those accounts and the programs they serve would be necessary to assess whether ultimate demand is similarly concentrated.

Sources

• Nvidia fiscal Q2 2027 Form 10-Q: https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm
• Nvidia accounts-receivable concentration and payment-term disclosure: https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/R14.htm
• Nvidia recast market-platform revenue: https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/R20.htm