Abstract:
Nvidia recently suspended its revenue sharing and credit support plans for cloud service providers. The move stems from internal concerns within the company that related cooperation may trigger antitrust scrutiny and involves discussions about the scope of Nvidia's influence on customer financing, operations and business decisions. The plan was suspended last week but has not been officially canceled and could still be revived in the future, people familiar with the matter said.
Nvidia announced in July this year that it plans to jointly build an "AI factory" with cloud service providers through revenue sharing and credit support models. Partners include Sharon AI and Firmus: the former plans to deploy 40,000 NVIDIA Blackwell AI GPUs to provide sovereign AI infrastructure; Firmus plans to deploy up to 170,000 NVIDIA GPUs in the AI factory park in Batam, Indonesia.
According to the original arrangement, NVIDIA can not only obtain direct income from selling GPUs, but also share part of the income after the relevant infrastructure is put into operation and services begin to be provided. However, this cooperation method that is more deeply involved in customer financing and production capacity operations is becoming the focus of potential regulatory risks.
In its latest financial report, Nvidia also disclosed that in order to support AI cloud computing and data center construction, the company has made approximately US$108 billion in guarantees and other commitments, involving the provision of financial guarantees, credit support, financing arrangements, and data center leasing and other commercial cooperation to customers and partners.
Nvidia also stated that the company has reached an agreement with an AI cloud service provider to expand the market’s use of its data center infrastructure products. Such an agreement could require Nvidia to buy back unsold committed computing capacity. The company warned that while it may generate some revenue from sales of supported capacity, related earnings may also be compressed if demand for AI computing power or market prices are lower than expected.
Comments