Nvidia’s Top Two Mystery Customers Account for 39% of Q2 Revenue, Raising Concerns Over Client Concentration
Santa Clara, California – August 29, 2025 – Nvidia Corporation (NASDAQ: NVDA), the dominant force in AI chip technology, disclosed in its latest SEC filing that two unnamed direct customers—referred to as “Customer A” and “Customer B”—accounted for a staggering 39% of its total revenue in the fiscal second quarter ending July 2025. This revelation, buried in the company’s 10-Q report submitted on August 28, has reignited investor worries about the chipmaker’s heavy reliance on a narrow base of clients amid the explosive growth of the AI sector. As Nvidia’s market capitalization hovers around $4 trillion, the concentration risk underscores the vulnerabilities in its business model, even as demand for its GPUs remains insatiable.
The filing specifies that Customer A contributed 23% of Nvidia’s total revenue, while Customer B made up 16%, a significant increase from the prior year when the top two customers represented just 14% and 11%, respectively. Combined, these two entities drove nearly $6 billion in sales during the quarter, based on Nvidia’s reported Q2 revenue of approximately $46.74 billion—up 50% year-over-year and beating analyst expectations of $46.06 billion. The company’s data center segment, which powers AI workloads and constitutes 88% of overall revenue, was the primary driver, with large cloud service providers accounting for about 50% of that division’s sales.
Nvidia’s CFO Colette Kress addressed the trend during the earnings call on August 27, stating, “We have experienced periods where we receive a significant amount of our revenue from a limited number of customers, and this trend may continue.” Kress highlighted robust demand not only from cloud giants but also from enterprises, “neoclouds”—emerging AI-focused providers—and even foreign governments. However, the identities of Customer A and B remain undisclosed, fueling speculation among analysts and investors. Nvidia classifies them as “direct customers,” which typically include original design manufacturers (ODMs) like Foxconn or Quanta, original equipment manufacturers (OEMs) such as Dell, or system integrators that purchase chips to assemble into systems sold to end-users like data centers and cloud providers.
The Mystery Behind the Customers and Speculation on Identities
Adding layers to the enigma, Nvidia’s filing notes that two “indirect customers”—likely end-users such as cloud service providers or internet companies—each exceeded 10% of total revenue, primarily through purchases routed via Customer A or B. This structure allows Nvidia flexibility in reporting, as some direct customers may build systems for their own use. An Nvidia spokesperson declined to elaborate on the identities, but the disclosure has sparked widespread conjecture.
Wall Street analysts point to major hyperscalers as prime suspects. Microsoft, Amazon, Alphabet (Google), and Meta Platforms have been aggressively ramping up capital expenditures (capex) on AI infrastructure. For instance, Microsoft allocated $55.7 billion to capex in its fiscal 2024, with plans for even more in 2025, much of it directed toward Nvidia GPUs for Azure’s AI capabilities. Amazon’s capex exceeded $60 billion in calendar 2024, while Meta forecasts up to $40 billion for AI projects this year, expanding its GPU capacity to 600,000 H100 equivalents by year-end. Alphabet’s spending hit $50 billion in 2024, and Oracle plans to double its AI capex to $13.8 billion in fiscal 2025. Other possibilities include Tesla, which committed $11 billion to AI infrastructure, or OpenAI, the creator of ChatGPT, though these are often indirect buyers.
Speculation extends beyond traditional tech giants. Some observers, including users on platforms like Reddit and Hacker News, suggest involvement from U.S. intelligence agencies, the military, or even “Singaporean” shell companies potentially linked to governments. Emerging players like CoreWeave, a neocloud provider backed by Nvidia, have been floated as candidates due to their rapid scaling in AI services. HSBC analyst Frank Lee noted in a research note that without clearer visibility into 2026 capex from cloud providers, “further earnings upside revision or share price catalyst in the near-term” seems limited, maintaining a hold rating on the stock.
The filing also mentions an “AI research and development company” contributing a “meaningful” amount through both direct and indirect channels, further obscuring the picture. This opacity is not new—Nvidia has long anonymized major clients in filings—but the escalating percentages highlight growing dependence as AI adoption accelerates.
Q2 Financial Highlights and Broader AI Demand
Nvidia’s Q2 results, reported on August 27, showcased continued momentum in the AI boom. Adjusted earnings per share came in at $1.05, surpassing estimates of $1.01, driven by record data center revenue. CEO Jensen Huang emphasized during the call that “Hopper demand remains strong, and the anticipation for Blackwell is incredible,” referring to the company’s next-generation architecture. Blackwell samples are already shipping, with high-volume production of the GB200 GPU expected in the fiscal fourth quarter, promising 30 times faster AI inference than the H100 at similar pricing ($30,000–$40,000 per unit).
The company forecasts Q3 revenue at $54 billion (±2%), implying over 50% growth, though this excludes potential sales of the H20 chip to China due to U.S. export restrictions. Gross margins are projected at 74.4% (GAAP) and 75.0% (non-GAAP), with full-year operating expenses growing in the mid- to upper-40% range. Huang’s recent global engagements, including trips to China, Taiwan, and Saudi Arabia, underscore Nvidia’s efforts to navigate geopolitical tensions. Notably, a deal with President Trump allows H20 exports to China in exchange for 15% of related revenue going to the U.S. government, potentially unlocking $8 billion in quarterly sales.
Despite the beat, Nvidia’s stock dipped in after-hours trading on August 27 before recovering, closing up slightly on August 28 amid the filing’s release. Year-to-date, shares have risen 35%, following a tripling in 2024, as the company hit a $4 trillion market cap milestone last month.
Implications for Nvidia and the AI Ecosystem
The 39% concentration poses tangible risks. If one or both top customers reduce spending—perhaps due to economic slowdowns, AI hype cooling, or shifts to in-house chips from rivals like AMD or custom silicon from hyperscalers—Nvidia’s growth could falter sharply. Customer A alone spent an estimated $10.8 billion in the first half of fiscal 2025, a level few entities can sustain. Analysts warn that while current demand is “full throttle,” as Huang described, over-reliance on a handful of buyers could amplify volatility.
On the positive side, Nvidia’s full-stack platform, including Spectrum-X Ethernet and AI Enterprise software, is gaining traction beyond hyperscalers. The company projects AI infrastructure spending could reach $3–$4 trillion by decade’s end, driven by diverse segments. Innovations like Blackwell Ultra (shipping H2 2025) and Rubin (2027) position Nvidia to maintain leadership, even as competition intensifies.
For investors, this disclosure serves as a cautionary tale in the AI gold rush. While Nvidia’s dominance in accelerated computing is unchallenged, diversifying its customer base will be key to mitigating risks. As the sector evolves, clarity on these mystery clients—and their capex trajectories—could dictate the stock’s next moves. Wall Street remains bullish long-term, but short-term catalysts hinge on sustained AI fervor from these unnamed powerhouses.