The two best clients of mystery in NVIDIA represent 39 % of its revenues in the second quarter

two Nafidia The company unveiled 39 % of NVIDIA revenues in the quarter of July, and the company revealed in financial presentation On Wednesday, it raised concerns about the focus of the chip maker’s clients.

The “Customer A” constitutes 23 % of the total revenue, and the “Customer B” consists of 16 % of the total revenue, according to the provision of the second quarter of the company to the Securities and Stock Exchange Committee.

This is higher than the same quarter of last year when the two customers in NVIDIA were the best 14 % and 11 % sales, according to the deposit.

The company regularly publishes information on a quarterly basis about its best customers, but the disclosure this week provides renewal of a discussion on whether the explosive growth in NVIDIA is driven by a handful of large cloud service providers like Microsoftand Amazonand Google and Oracle.

“The large cloud service providers” make up about 50 % of the company’s data center revenue. This is important because the data center works make up 88 % of the total NVIDIA revenue in the second quarter.

“We have witnessed periods as we receive a large amount of our revenues from a limited number of customers, and this trend may continue,” Nafidia wrote in the deposit.

Increasingly, analysts look forward to the obligations of cloud capital expenditures for the model of the future growth of NVIDIA.

“We see a limited space for more upscale review or stock rate stimulating in the short term unless we have increased clarity on the upward trend in 2026 (CAPEX). It has a shares suspension classification.

But Nvidia A and customer B are not necessarily cloud service providers. It is a bit puzzle, and the NVIDIA representative refused to share the identities of the customer A and the customer B.

In his file, Nvidia says she has “direct customers” and “indirect clients”. Customer A and Customer B as “Direct Customers” are listed.

The direct customers are not the final users of NVIDIA P there. They are companies that buy chips to build in complete systems or departments panels that then sell them to data centers, cloud services providers and final users. Some of these direct customers are the original design manufacturers or original equipment manufacturers such as Foxconn or Quanta. Others are distributed or complementary to the system, such as Dell.

Meanwhile, indigo customers include cloud service providers, internet and institutions, which usually buy systems from direct NVIDIA customers. NVIDIA says it can only estimate revenues for indirect customers based on purchase orders and internal sales data.

The deciphering if any of these cloud service providers are NVIDIA’s ambiguity agents, and this is partly due to the fact that the chips maker has a space for maneuver in the definitions of its direct and non -direct customers.

For example, NVIDIA wrote in the deposit that some direct customers buy chips to create systems for their own use.

In addition, NVIDIA noted that two of its indirect customers represent more than 10 % of their total revenues, and buy purchase systems primarily through customers A and B.

NVIDIA contributed more to the mystery of everything, that the “Artificial Intelligence Research and Development Company” contributed to the amount of “meaning” of revenues through the direct and non -direct customers.

NVIDIA told investors on Wednesday that the demand for the company’s artificial intelligence systems is still high, not only among cloud service providers, but among other types of customers, including institutions purchase systems for artificial intelligence and “Neooclouds”, companies that handle the largest service providers with services that are more seized than artificial intelligence. NVIDIA also included foreign governments, saying that they would record $ 20 billion revenues this year for “Sovereign Ai”. Chris told analysts in the profit call.

Nvidia Jensen Huang CEO also said that the company has new expectations ranging from $ 3 to $ 4 trillion in Amnesty International infrastructure by the end of the contract. He said that it may take about 70 % of the total cost of a data center focusing on $ 50 billion, not only for its graphics processing units but for other chips that it sells as well.

Huang told the investors that it was a reasonable goal over the next five years because of the extent of what was spent on Easter and committed to spending – $ 600 billion this year, according to Hanging. He also said that new types of customers, such as companies or cloud service providers abroad, were joining construction.

“As you know, the depth has multiplied to the four most important four of the excessive numbers within two years, as the artificial intelligence revolution entered the full steam,” Huang said.

He watches: The risk of relying on the concentration of Nafidia