U.S. artificial intelligence chip giant NVIDIA (NVDA.US) disclosed in its latest financial filing that its revenue is highly concentrated among a small number of customers. The company's second-quarter filing submitted to the U.S. Securities and Exchange Commission (SEC) on Wednesday showed that just two customers contributed 39% of NVIDIA's total revenue for the July quarter.

"Customer A" accounted for 23% of total revenue, while "Customer B" contributed 16%. This proportion is significantly higher than the same period last year, when the top two customers contributed 14% and 11% of sales respectively. This disclosure has once again sparked market discussions about whether NVIDIA is overly dependent on a few major customers, particularly cloud computing giants such as Microsoft, Amazon, Google, and Oracle.

NVIDIA Chief Financial Officer Colette Kress stated that large cloud service providers account for approximately 50% of the company's data center revenue. In this quarter, the data center business represented 88% of NVIDIA's overall revenue, serving as the core driver of the company's growth. NVIDIA acknowledged in its filing: "We have derived a substantial amount of our revenue from a limited number of customers during certain periods, and this trend may continue."

However, NVIDIA did not reveal the specific identities of Customer A and Customer B. It's worth noting that these two are listed as "direct customers," meaning they are not end users but rather companies that purchase chips from NVIDIA and then assemble them into systems or circuit boards for resale to data centers, cloud service providers, or end customers. Such direct customers include ODM/OEM manufacturers like Foxconn and Quanta, as well as system integrators or distributors like Dell (DELL.US). Meanwhile, cloud computing companies, internet enterprises, and other business customers are classified as "indirect customers," with NVIDIA only able to estimate their contributions through orders and internal sales data.

The filing also revealed that two indirect customers each contributed more than 10% of total revenue, primarily through purchases of systems from Customer A and Customer B. Additionally, an unnamed "AI research and development company" also contributed significant revenue to NVIDIA through both direct and indirect channels. This makes NVIDIA's customer structure even more complex and mysterious.

Analysts generally believe that NVIDIA's future growth potential will be closely tied to global cloud computing capital expenditure. HSBC analyst Frank Lee noted in a Thursday research report: "Unless cloud service provider capital expenditure expectations for 2026 become clearer, NVIDIA's upside earnings potential and stock price catalysts will be limited in the near term." He maintained a "Hold" rating on NVIDIA.

Facing skepticism, NVIDIA management emphasized that the company's demand has expanded beyond traditional cloud service providers. Kress pointed out during the earnings call that in addition to cloud computing giants, enterprise users, "new cloud service providers," and overseas government procurement through "sovereign AI" projects are collectively driving the company's growth. NVIDIA expects to record $20 billion in revenue from the sovereign AI sector this year.

CEO Jensen Huang presented an even grander vision during investor communications: by the end of this century, the global AI infrastructure market could reach $3-4 trillion. NVIDIA can not only capture approximately 70% of the cost share in AI data centers worth $50 billion but also profit from other chips beyond GPUs. He stated that the capital expenditure of the top four hyperscale cloud providers has doubled over the past two years and is expected to reach $600 billion this year, while more enterprises and overseas cloud service providers are joining the AI computing power expansion.