A Securities and Exchange Commission filing from Nvidia revealed a striking concentration: two direct customers together accounted for 39% of the company’s second-quarter revenue, underscoring how a handful of buyers are fueling an otherwise staggering top-line performance.
Nvidia posted record revenue of $46.7 billion for the quarter, a jump of roughly 56% year over year driven largely by sales into AI data centers and enterprise GPU deployments.
What the SEC filing actually shows
The filing identifies “Customer A” as 23% of Q2 revenue and “Customer B” as 16%, while four other named direct customers represented double-digit slices as well: 14%, 11%, 11% and 10% respectively.
Nvidia characterizes those buyers as direct customers—OEMs, system integrators or distributors who purchase chips straight from the company—rather than end users like cloud platforms that often buy through those intermediaries.
Why customer concentration matters
Concentration amplifies both upside and downside: a few large deals can produce blistering growth, but losing or slowing one account would hit revenue disproportionately. Investors and risk teams will be watching renewal cadence and contract terms closely.
Analysts such as Dave Novosel of Gimme Credit told Fortune that while the concentration is a material risk, the likely buyers are cash-rich and expected to invest heavily in data centers—something that could extend Nvidia’s tailwind over several years.
Who could those “mystery” customers be?
The filing stops short of names, but the profile fits major OEMs and integrators that assemble freight‑scale systems: companies like Dell Technologies, HPE, Supermicro and Inspur regularly purchase large volumes of GPUs for enterprise and hyperscale customers.
At the same time, cloud providers—Amazon Web Services, Microsoft Azure, Google Cloud and Oracle Cloud—are central to the AI infrastructure surge, but they often appear as indirect buyers because they source through system builders or sign reseller agreements.
Supply-chain and geopolitical context
Exports, chip allocation and fab capacity shape who can buy what and when. TSMC’s production cadence and U.S. export controls affecting high-end accelerators influence which customers can secure inventory and at what scale.
That dynamic helps explain concentrated orders: a few customers with priority access, large budgets and urgent AI projects can soak up inventory and create outsized revenue pockets for Nvidia in a single quarter.
Strategic implications for Nvidia and the market
For Nvidia, the short-term picture is cash-rich and leverage-enhanced, but the company will want to diversify its buyer base to reduce earnings volatility and bargaining pressure from any single partner.
Competitors and suppliers—AMD, Intel, Marvell, and fabs like Samsung and TSMC—are watching too. Large, concentrated purchases can trigger competitive bids, R&D shifts and inventory plays across the ecosystem.
Regulators and institutional investors will likely monitor subsequent filings and conference calls for more granularity on contract length, renewal risk and whether those direct customers are fulfilling orders for global hyperscalers.
In short: the headline numbers confirm Nvidia’s central role in the AI infrastructure boom, but the concentration disclosed in the SEC filing is a reminder that rapid growth can sit atop a narrow set of customers—an arrangement that brings both power and exposure.
