Nvidia Faces Rising Credit and Customer Concentration Risks as OpenAI Burns Cash
While OpenAI runs more than 90 percent of its workloads on Nvidia silicon, widening operating losses and a $105 billion lease guarantee test the chipmaker's financial exposure.
OpenAI depends on Nvidia hardware for more than 90 percent of its active computing capacity even as its quarterly operating losses outpace top-line revenue, intensifying financial concentration and credit risks for the semiconductor maker, according to an analysis reported by Yahoo Finance (https://www.tikr.com/blog/nvidias-problem-openai-is-90-plus-percent-nvidia-and-its-the-lab-burning-cash?ref=yahoofinance&tlnk=yf-261115-6qwrm2q7&.tsrc=rss).
The operational divergence between leading artificial intelligence laboratories is drawing scrutiny toward how downstream software unit economics affect hardware suppliers. While OpenAI is reportedly negotiating a fresh $15 billion funding round, its deep reliance on Nvidia infrastructure stands in contrast to competitors that have diversified their compute architectures across custom application-specific integrated circuits and rival cloud platforms.
OpenAI generated $10 billion in third-quarter revenue, marking roughly 50 percent growth, according to investor documents reviewed by The Wall Street Journal. However, its adjusted operating loss widened to $10.9 billion over the same period, translating to an adjusted operating margin of negative 109 percent, as reported by CNBC's Kate Rooney. In contrast, rival Anthropic recorded an adjusted operating margin of roughly positive 15 percent, although its public IPO filing revealed an unadjusted operating loss of $567 million on $17 billion in third-quarter revenue.
The underlying infrastructure configurations of both frontier research labs differ substantially. Speaking on the TBPN broadcast, SemiAnalysis chief analyst Dylan Patel observed that although OpenAI has developed internal chip designs, more than 90 percent of its compute footprint remains powered by Nvidia graphics processing units. Conversely, Patel estimated that Anthropic runs one-third or less of its computing workloads on Nvidia hardware, securing the bulk of its capacity through Alphabet and Amazon cloud infrastructure.
That infrastructure split has channeled significant hardware expenditure into alternative supply networks. Alphabet relies heavily on custom Tensor Processing Units co-developed with Broadcom, a product line Morgan Stanley estimates could generate $84 billion in TPU system sales to external customers by 2027. Amazon has simultaneously scaled its proprietary silicon footprint, announcing that its internal chip business surpassed an annualized run rate of $25 billion by the second quarter.
Beyond direct silicon deliveries, Nvidia has deployed its own balance sheet to support OpenAI's operational expansion. The chipmaker agreed to guarantee portions of the lease and utility commitments for OpenAI's data center campus in Ohio. That guarantee is capped at $105 billion, begins phasing in during 2028, and becomes payable only in the event of an OpenAI default. Nvidia is now preparing to moderate its pace of artificial intelligence startup investments and credit guarantees, according to reporting from The Information.
Nvidia retains broad structural demand across global hyperscalers, with executive management confirming that the five largest cloud providers account for slightly over half of total data center revenues. Wall Street consensus estimates tracked by TIKR project Nvidia's annual revenue will rise from $216 billion in the fiscal year ended January 2026 to $896 billion by the fiscal year ending January 2029, though rising memory costs are forecast to compress gross margins toward a low of 71 to 72 percent in the January 2027 quarter.
Nvidia is scheduled to report quarterly financial results on Nov. 17. Market analysts are expected to focus closely on the company's accounts receivable figures and executive commentary regarding customer financing arrangements for signals on whether funding strains at major research labs are spilling over into chip suppliers.
Sources
Written by
The Company Wire
Inside the companies building what’s next. Reporting on startups, technology, funding and the people shaping them.