October 22, 2025
Nvidia Corp. is reportedly in discussions to guarantee a portion of the loans that OpenAI plans to secure for the construction of its own data centers, according to a Wall Street Journal report. This potential arrangement, part of a broader $100 billion investment commitment from Nvidia, could expose the chipmaker to billions of dollars in debt obligations should OpenAI default on its repayments.
The talks follow a letter of intent signed by the two companies in September, in which OpenAI committed to deploying at least 10 gigawatts of Nvidia's advanced computing systems in its new data centers. This ambitious deployment could involve Nvidia leasing up to five million of its AI chips to OpenAI, an order valued at approximately $350 billion at current market prices.
To help fund this colossal order, Nvidia has pledged to invest up to $100 billion in OpenAI as the new systems are progressively deployed. The Wall Street Journal notes that investment discussions were initially sparked soon after OpenAI and SoftBank announced their $500 billion "Stargate" data center joint venture in January. However, those early talks reportedly lost momentum shortly thereafter.
The negotiations regained urgency following a June report from The Information, which revealed that OpenAI had begun utilizing Google’s Tensor Processing Units (TPUs). This move by OpenAI to diversify its chip suppliers appears to have prompted Nvidia to re-engage seriously, culminating in the deal announced just months later.
Despite this significant commitment to Nvidia, OpenAI has continued its strategy of broadening its compute infrastructure. Mere weeks after the Nvidia agreement, OpenAI signed a separate 6GW deal with AMD, a transaction that could ultimately grant OpenAI a ten percent stake in the competing chip designer.
Subsequently, OpenAI officially unveiled its long-anticipated custom AI hardware project, developed in partnership with Broadcom. The company announced plans to deploy up to 10GW of these custom chips by 2029.
Concurrently, Nvidia has pursued a multi-faceted strategy to secure demand for its flagship products. The company is a known investor in several "neocloud" providers, including together.ai, Nscale, Lambda, Nebius, CoreWeave, Crusoe, and Firmus. In many of these deals, Nvidia also leases back a portion of the GPUs it sells to the companies, a mechanism that simultaneously boosts demand and revenue.
A particularly notable arrangement is with CoreWeave, where Nvidia has committed to purchasing any of the company's unsold GPU capacity through the year 2032, effectively guaranteeing a market for its own hardware.
SOURCE DCD
