GPU Immediate in Exchange for a Share of Future Revenues: NVIDIA Invests Without Investing
NVIDIA wants to further strengthen its position in artificial intelligence with a new collaboration model aimed at high-growth startups. The company has announced a partnership program that will allow emerging businesses to access the power of its GPUs without immediately bearing all the infrastructure costs. In exchange, NVIDIA will receive a share of the revenues generated by the participating companies, complementing its traditional hardware sales model with a source of recurring revenue linked to the use of its platforms.
The new initiative arises from the transformation taking place in the artificial intelligence market. In recent years, demand for computing was primarily focused on training models; today, the sector is rapidly shifting its focus towards production inference. Indeed, companies are implementing large-scale operational AI services capable of continuously generating tokens for millions of users, giving rise to what NVIDIA calls "AI Factory": infrastructures that must remain operational continuously, ensure high levels of utilization, and provide immediately available computing capacity.
According to NVIDIA, this change is making it increasingly difficult for startups to procure the necessary infrastructure. The young companies in the sector, despite having promising models, often lack access to the required capital to finance data centers and large GPU clusters. Even multi-year agreements for purchasing computing capacity do not always allow the necessary funding for the implementation of infrastructure.
For this reason, the company has developed an economic model that aims to align the interests of NVIDIA, cloud providers, and companies developing AI solutions. The program targets startups specializing in model development, inference service providers, AI agent platforms, software houses, businesses, and research organizations.
From an operational standpoint, NVIDIA will provide credits and financial support to the cloud providers participating in the program, enabling them to purchase NVIDIA infrastructure intended for end customers. Partners will sell cloud services based on the U.S. company's GPUs, while NVIDIA will continue to collect the normal revenue from hardware sales, but will also add a share of the revenues generated from cloud services built on the capacity supported by the program. The declared goal is to accelerate the adoption of NVIDIA platforms among the fastest-growing AI companies while simultaneously creating a recurring revenue stream directly linked to the actual utilization of the infrastructure.
For companies developing language models, inference platforms, or enterprise services based on artificial intelligence, the new approach should significantly reduce the time required to obtain computing capacity. This way, it will be possible to avoid the long phases normally required to identify a suitable site, secure energy availability, build the data center, and complete the installation of the hardware.
The initiative has already entered the operational phase through the establishment of new AI Factories distributed across various geographical areas. NVIDIA has announced two initial partners that will utilize this business model. The Australian company Sharon AI will create infrastructure that could accommodate up to 40,000 NVIDIA Grace Blackwell GB300 GPUs, intended to provide large-scale computing capacity to the program's clients. CEO James Manning called the agreement a strategic step towards building sovereign and large-scale AI infrastructures.
At the same time, Firmus Technologies, a company based in Singapore, is developing a campus dedicated to AI Factory DSX on the island of Batam, Indonesia. The complex is expected to achieve an electrical capacity of 360 MW and house up to 170,000 NVIDIA GPUs. According to the company, the aim is to offer scalable, energy-efficient, and cost-effective infrastructure, enabling AI businesses to compete on a global scale.