Nvidia says two anonymous customers drove 40% of Q2 sales

Nvidia revealed on Wednesday that two unnamed customers accounted for 39% of its total revenue in the second quarter of its fiscal year, according to a regulatory filing with the U.S. Securities and Exchange Commission.

The company identified the buyers only as “Customer A” and “Customer B.” Customer A represented 23% of sales, while Customer B contributed 16% during the three months ending in July. Together, the two accounted for nearly \$6 billion of Nvidia’s quarterly revenue.

That concentration is much higher than in the same quarter a year earlier, when the two largest customers made up 14% and 11% of revenue, respectively.

The sharp increase is drawing closer attention to who is driving the surge in demand for Nvidia’s AI chips, and what that could mean for the company’s future revenue stability.

Although speculation has focused on major cloud providers such as Amazon, Microsoft, Google, or Oracle, Nvidia did not disclose the identities of the clients.

Nvidia conceals key buyers behind supply chain layers

In its filing, Nvidia described Customer A and Customer B as “direct customers,” though that does not necessarily mean they use the chips themselves.

These direct buyers purchase Nvidia’s hardware to build complete systems or circuit boards, which are then resold to end users such as cloud providers, corporations, and government agencies.

Possible intermediaries include original design manufacturers and equipment makers such as Foxconn and Quanta, as well as system integrators like Dell.

Nvidia also said it has indirect customers, meaning the end users of its chips, though these companies do not buy directly from Nvidia.

These indirect buyers include cloud infrastructure operators, technology companies, and large enterprises developing internal AI platforms. Nvidia explained that it can only estimate the scale of its indirect business based on purchase orders and internal sales data.

The filing also disclosed that two indirect customers each represented more than 10% of total revenue, both accessed through either Customer A or Customer B.

That has fueled speculation over whether these are established cloud providers or emerging players rapidly expanding in AI.

During an earnings call, CFO Colette Kress said about 50% of Nvidia’s data center revenue came from major cloud service providers. Data center sales accounted for 88% of the company’s total revenue in Q2. Kress told analysts:

“We have experienced periods where we receive a significant amount of our revenue from a limited number of customers, and this trend may continue.”

Analysts track 2026 cloud capex for growth signals

Wall Street has taken note of Nvidia’s growing dependence on a small group of large buyers. HSBC analyst Frank Lee wrote Thursday that the stock is unlikely to see “further earnings upside revision or share price catalyst in the near-term unless we have increasing clarity over upside in 2026 [cloud service provider] capex expectations.” He currently rates Nvidia shares as a hold.

Nvidia also mentioned that an “AI research and development company” contributed a significant share of revenue through both direct and indirect purchases, but did not name the customer.

The company emphasized that demand extends beyond major cloud providers. It pointed to a broader customer mix, including enterprises building their own AI infrastructure, foreign governments, and what it calls “neoclouds” — newer infrastructure providers focused on AI workloads and positioned as competitors to the dominant four.

CEO Jensen Huang projected that AI infrastructure spending could reach \$3 to \$4 trillion by 2030. He said Nvidia could capture around 70% of the cost of a \$50 billion AI data center, through GPUs as well as networking, accelerators, and software.

Huang also highlighted the scale of current spending. “As you know, the capex of just the top four hyperscalers has doubled in two years as the AI revolution went into full steam,” he told analysts, referencing Amazon, Microsoft, Google, and Oracle.