NVIDIA is leveraging its robust balance sheet to assist more companies in acquiring its costly AI chips, with the potential to generate profits from these arrangements.
According to executives from GPU cloud service providers Firmus and Sharon AI—both participants in the initiative—along with three other firms that conduct business with NVIDIA, the chipmaker has pledged financial support to emerging cloud providers that rent out its graphics processing units. In return, NVIDIA will receive a portion of the revenue generated. Should these companies struggle to find AI developers to lease the GPUs, NVIDIA will provide support by committing to repurchase the unused hardware.
Insiders familiar with the matter indicate these deals are part of an internal NVIDIA program referred to by some employees as the "AI Computing Partner" initiative. NVIDIA has proposed guaranteeing a price to pay for any unsold GPU capacity held by these cloud service providers.
A spokesperson for NVIDIA confirmed the existence of this program, signaling that the world's most valuable company is increasingly relying on its financial strength in an attempt to act as a "central bank" for the hundreds of enterprises purchasing its chips in bulk.
GPUs are typically the most expensive component within AI data centers. Consequently, NVIDIA's guarantee to cover the lease costs for unsold GPU computing power makes it easier for chip buyers with less-than-ideal credit ratings to secure the necessary financing.
One data center executive remarked that in such transactions, "NVIDIA kills two birds with one stone." If NVIDIA only guaranteed leases related to the physical facility, "you still face the problem of 'how do you finance the GPUs?'" However, if NVIDIA commits to paying for unused computing capacity within the facility, "then both the GPUs and the entire data center can get financing," the executive stated.
Currently, a handful of major cloud providers—including Amazon, Microsoft, SpaceX, Oracle, Meta, and Google—purchase the majority of NVIDIA's chips. NVIDIA aims to alter this landscape—partly because many of these companies are also developing competing AI chips—and over the past few years has been nurturing a series of emerging GPU cloud providers, such as CoreWeave, to reduce its dependence on these industry giants.
Comments