NVIDIA’s New Revenue Sharing Model Could Change How AI Clouds Buy GPUs

NVIDIA is expanding its role in the AI infrastructure market with a new model that combines credit support for cloud providers with ongoing revenue sharing from the AI services built on its hardware. The approach could help smaller AI cloud companies finance large GPU deployments while giving NVIDIA a recurring source of income after the initial hardware sale.

Under the traditional model, a cloud provider buys GPUs, builds data center capacity, rents out compute power, and keeps the revenue generated from those services. NVIDIA’s new arrangement reportedly changes that structure by allowing the company to support purchases through credit while also taking a share of the revenue generated by NVIDIA-powered cloud infrastructure.

The model gives NVIDIA a closer financial connection to how well its partners perform after their systems are deployed.

NVIDIA Is Moving Beyond One-Time GPU Sales

NVIDIA has built much of its AI business around selling accelerators, networking hardware, software tools, and complete data center platforms. The new revenue-sharing approach could extend that role further.

Instead of earning only when a cloud provider buys hardware, NVIDIA could continue receiving payments as customers use AI services for training, fine-tuning, inference, and agent-based workloads.

This could be especially attractive for newer AI cloud companies that want to compete with larger providers but do not have the capital needed to build massive GPU clusters on their own.

AI Cloud Demand Is Creating a Financing Challenge

Building AI infrastructure requires huge upfront investment. GPU clusters, networking, cooling, power systems, land, and data center construction can cost billions of dollars before a provider earns meaningful revenue.

That creates a difficult situation for companies trying to enter the AI cloud market. They may have customer interest and future contracts, but still struggle to secure financing for the hardware needed to deliver those services.

NVIDIA’s credit support could help bridge that gap. The company would be helping partners buy more NVIDIA systems while gaining a stake in the services those systems later provide.

The strategy also reflects the growing importance of inference. Training large AI models remains expensive, but inference is becoming a major source of ongoing demand as more companies deploy chatbots, AI agents, search tools, coding assistants, and enterprise automation services.

Several AI Cloud Providers Are Linked to the New Model

The initial partners mentioned include Sharon AI, Firmus Technologies, Baseten, Fireworks AI, and Together AI.

Sharon AI reportedly plans to deploy as many as 40,000 Grace Blackwell GB300 systems. Firmus Technologies is said to be building an AI factory campus in Batam, Indonesia, with plans that could eventually scale to 360 megawatts and around 170,000 NVIDIA GPUs.

These numbers show how quickly AI infrastructure plans are growing. Even a single large deployment can involve tens of thousands of high-end accelerators.

The Strategy Could Make NVIDIA Even More Influential

The new model could help NVIDIA sell more hardware while making customers more dependent on its ecosystem. Cloud providers would not only use NVIDIA GPUs, but could also become financially linked to NVIDIA through shared revenue arrangements.

That could strengthen NVIDIA’s position against rival chip makers and cloud platforms. It may also raise questions about how much influence one hardware company should have over AI infrastructure financing, cloud pricing, and service availability.

For AI cloud providers, the offer could be difficult to ignore. Access to funding and hardware is one of the biggest barriers in the market. NVIDIA’s new model may give them a path to scale faster, but it also means sharing a larger part of their future business with the company supplying the GPUs.