The latest quarterly report from NVIDIA filed with the U.S. SEC has raised alarms among analysts and investors: two customers accounted for 39% of revenue in Q2 2025. This figure marks a significant increase compared to the same period in 2024, when their combined share was 25%.
The document identifies these companies only as “Customer A” (23% of sales) and “Customer B” (16%). Their identities have not been disclosed, sparking speculation in the market: are they cloud giants like Microsoft, Amazon, Google, or Oracle, or ODM/OEM manufacturers such as Foxconn and Quanta, or systems integrators like Dell?
Growing dependency on a few customers
NVIDIA itself acknowledges in its earnings presentation that it has gone through periods where a significant portion of its revenue comes from a limited group of buyers, and warns that this trend might continue.
Meanwhile, Chief Financial Officer Colette Kress confirmed that hyperscale cloud providers now account for nearly 50% of data center revenue, which today is at the core of the business and represents 88% of total revenue.
HSBC analyst Frank Lee was adamant in a note to investors: “We see limited short-term catalyst potential in NVIDIA’s stock unless we get more clarity on the cloud providers’ capex expectations for 2026.”
Direct vs. indirect: a maze of definitions
The mystery surrounding A and B is further complicated by NVIDIA’s own classification:
- Direct customers: purchase chips to integrate into systems, boards, or servers which they then resell to third parties. This includes ODM/OEM manufacturers like Foxconn or Quanta, as well as systems integrators like Dell.
- Indirect customers: are end users — hyperscalers, digital companies, governments — who acquire these systems through direct customers.
The company even admits that some direct customers use chips for their internal consumption, blurring the lines further.
The report adds that two indirect customers also exceeded 10% of global revenue, but channel their purchases specifically through A and B. Additionally, NVIDIA mentions a significant involvement of an AI research and development company, which reportedly contributed with substantial orders both directly and indirectly.
The context: explosion of AI spending
This level of concentration comes amid a feverish investment in AI infrastructure. CEO Jensen Huang stated that the specialized AI infrastructure market could reach between $3 trillion and $4 trillion by the end of the decade, with NVIDIA aiming to capture up to 70% of the cost of an AI-focused data center (around $50 billion per installation), not only through its GPUs but also through its network chips, CPUs, and software portfolio.
Huang also highlighted that capital expenditure by the four major hyperscalers has doubled in just two years, driven by the global race to train and deploy large-scale generative AI models.
New customers: neo-clouds, companies, and governments
While major cloud providers account for half of the business, NVIDIA emphasizes that demand is diversifying:
- Traditional companies building their own clusters for enterprise AI applications.
- “Neo-clouds”, emerging providers with specialized AI services seeking to differentiate themselves from AWS, Azure, or Google Cloud.
- Foreign governments, which according to the company will contribute $20 billion in 2025 through contracts related to soberan AI projects.
This broadening of buyer profiles could mitigate reliance risks, but currently does not offset the growing influence of the mysterious A and B.
Concentration risk and growth sustainability
The core issue is clear: if major clients reduce their investment pace, NVIDIA could face an immediate impact on income and forecasts. The scenario becomes even more uncertain given that the identities of A and B, and their actual control over the bulk of revenue, remain unknown.
At the same time, NVIDIA’s current strength is undeniable. Its leadership in AI GPUs and ecosystem of software make it almost indispensable for players seeking to compete in the sector.
In other words: NVIDIA is on top, but its dependence on a few strategic buyers creates a vulnerable flank that analysts will continue to monitor quarter by quarter.
Frequently Asked Questions (FAQs)
1. What percentage of NVIDIA’s revenue depends on its two largest customers?
39% in Q2 2025, up from 25% in 2024.
2. Is it known who Customer A and Customer B are?
Not officially. They could be OEM/ODM integrators like Foxconn or Quanta, or large cloud providers purchasing directly.
3. What risk does this concentration pose?
Excessive reliance: any decrease in orders from these customers would have a direct effect on results.
4. What role do governments play in NVIDIA’s growth?
The company expects around $20 billion in 2025 from contracts related to sovereign AI projects, a rapidly expanding segment.
5. Which other clients are gaining importance?
Private companies with their own clusters, specialized neo-clouds, and AI research centers, in addition to traditional hyperscalers.