STOCKS | Nvidia's Revenue Dependence on a Few Large Customers Deepens

Nvidia CEO Jensen Huang has strong reasons to keep investing in new cloud providers and AI companies to win more customers. According to Sina Finance, the company's sales are becoming increasingly concentrated among a small number of clients, and some of those customers may reduce purchases in the future. In the first half of the fiscal year ended in July, three customers accounted for 44% of total revenue. In the prior fiscal year, two customers accounted for 36% of total revenue, while in fiscal 2023 Nvidia had no single customer contributing more than 10% of revenue. Nvidia's data center business rose from $15 billion in fiscal 2023 to $193.7 billion last year, and revenue from that business is expected to double again this year. The concentration risk has already worried investors. Michael Burry, known for correctly anticipating the 2008 subprime crisis, has recently viewed Nvidia's rising customer concentration as a major risk. The latest filing's three largest customers may include Dell or Foxconn, which buy Nvidia chips to assemble servers and resell them to downstream companies. Dell recently disclosed that second-quarter revenue from its AI-optimized servers doubled to $16.4 billion, helping overall revenue jump 58%. Foxconn also said first-half revenue rose 35% on surging AI equipment sales. Nvidia is also heavily dependent on some cloud providers. In February, CFO Colette Kress said five cloud providers and hyperscalers, likely including Meta and SpaceX, accounted for more than 50% of data center revenue. It is still unclear how many of those five companies are among Nvidia's top three customers. The top three may include SpaceX, Meta, or Microsoft, all of which are building new AI data centers with Nvidia chips. Google and Amazon also buy large volumes of Nvidia chips, but both also use their own AI chips on a large scale, making them less likely to be among Nvidia's biggest customers. Microsoft, Meta, and SpaceX are all developing their own chips, which means they may reduce purchases from Nvidia in the future. That is one reason Nvidia is working to cultivate new customers by investing in chip buyers such as CoreWeave and Nebius. Nvidia is also expanding overseas. On Wednesday, it added several Australian companies to its AI factory project, supplying chips and networking equipment to firms that build and operate data centers. This year, Nvidia has also advanced similar projects in India and Armenia, aiming to build more sovereign project customers and new cloud providers beyond its core U.S. clients. The payment-term risk may be rising even as Nvidia pushes for diversification. In July, the company said 70% of accounts receivable came from five customers. By comparison, three customers accounted for 56% of accounts receivable in January, while only two customers accounted for 33% at the end of fiscal 2025. Dell recently disclosed that accounts payable to suppliers jumped 48% in the first half of the year to $49.7 billion as of the end of July, and much of that amount was likely owed to Nvidia. Nvidia also said at the end of August that accounts receivable rose 64% to $63 billion in the first half because of extended payment terms on large cross-quarter agreements with some investment-grade customers. Last quarter, Nvidia extended payment terms from 45 days to 60 days, and in a concurrent securities filing said it may lengthen them further. The company said customers generally pay shortly after delivery, but for large purchase orders from investment-grade customers it has provided, and may continue to provide, extended payment terms of up to one year, starting at 90 days, based on the scale of data center construction projects. Looser payment terms carry risks: they can reduce Nvidia's quarterly operating cash inflow and make the chipmaker increasingly dependent on the financial health of customers that receive deferred-payment privileges.