en.Wedoany.com Reported - Nvidia has introduced a new business model for its AI accelerators. Small cloud data center operators, known as "Neoclouds," can purchase Nvidia hardware and lease it to startups and other end users. The core of this model is a "credit support model": Nvidia commits to leasing back unused computing capacity if operators fail to find enough customers. In return, operators must share revenue with Nvidia. It remains unclear whether this model comes with discounts on hardware purchase prices.

Under this model, Neoclouds can access the latest hardware with reduced financing risk. According to the announcement, the initial two Australian suppliers involved are Sharon AI and Firmus Technologies. Sharon stated it is building a data center with a capacity of up to 72 megawatts, equipped with up to 40,000 GB300 motherboards (Grace Blackwell). Firmus plans to open a campus in Indonesia that can scale to 360 megawatts and 170,000 Nvidia GPUs.

Particularly advantageous for Nvidia: all AI data centers under this program are built to Nvidia's specifications. Thus, in addition to AI accelerators, Nvidia also sells processors, switches, and other components.

Chief Financial Officer Colette Kress believes this can generate "recurring, usage-based revenue streams." Meanwhile, this program is another way for Nvidia to fill its order book. Nvidia has previously invested billions of dollars in AI companies, which in turn purchase or lease Nvidia hardware. However, this cyclical model only applies to large customers like OpenAI.

The new credit support model does not require such large-scale investments and is more flexible. End users can obtain the computing power they need without having to consider data center construction.