Nvidia rolls out revenue-sharing and credit-support model aimed at accelerating AI cloud buildouts

Nvidia is introducing a new business model designed to make it easier for AI startups and other users to access large-scale, accelerated computing, pairing traditional hardware sales with an economics-sharing mechanism tied to cloud revenue. The goal, according to Nvidia, is to address a bottleneck that emerges when demand shifts from building AI models to running them at production scale, where inference services can operate continuously and consume infrastructure at “token-scale” speeds.

Under the model described by Nvidia, AI clouds can procure NVIDIA infrastructure for their customers through what the company calls “economic alignment” using revenue sharing and credit support. In Nvidia’s framing, the credit-support component is intended to help solve a financing problem that can prevent compute providers from funding capacity-intensive AI “factories,” even when customers make long-term commitments. Nvidia said this can otherwise slow site selection, power procurement, construction, and hardware bring-up, delaying when infrastructure is available to serve real workloads.

Nvidia said the structure works by letting AI clouds sell NVIDIA-powered cloud services while Nvidia earns both standard product revenue and a share of the cloud revenue on the supported capacity. For AI model builders, inference providers, agent platforms, and enterprises scaling AI, Nvidia characterized the arrangement as a faster path to full-stack accelerated computing without waiting for the full procurement and construction cycle of a new data center footprint.

In describing how the initiative is already taking shape, Nvidia said AI cloud companies are building “DSX AI factories” designed to serve customers and workloads across regions. Nvidia also pointed to named examples of participants, including Sharon AI and Firmus, while describing the broader intent to open compute access to startups, regional AI players, research organizations, and enterprise users moving toward production deployments.

Market coverage of the announcement also indicated that Nvidia would be taking its first revenue-sharing steps with two data center operators, highlighting that the effort is meant to move beyond a pure chip-supply relationship. The reporting suggested the cloud revenue share is directly linked to capacity that Nvidia supports, which would extend Nvidia’s earnings beyond one-time hardware transactions into a recurring, usage-linked stream if deployments scale as expected.

The initiative reflects a strategic shift that Nvidia has been pursuing across its ecosystem: positioning itself not only as a supplier of GPUs (graphics processing units, specialized chips used for AI training and inference) but as an orchestrator of the infrastructure economics behind AI services. By aligning incentives with AI clouds, Nvidia is effectively trying to reduce the financial friction that can slow down new compute availability, while giving cloud providers a reason to invest in NVIDIA-powered capacity that can be utilized quickly and kept highly utilized.

Even with the outlined mechanics, important details were not fully spelled out in the accessible company description, including the precise revenue-share formula, how “supported capacity” is defined in contractual terms, and what credit-support limits and conditions apply in practice. Nvidia also did not provide the size of the expected program footprint, the commercial terms for individual participants, or any quantitative guidance tied to how much incremental revenue such arrangements could generate.

What to watch next is whether additional AI clouds and data center operators sign on to the revenue-sharing and credit-support structure, and whether Nvidia provides further documentation on pricing, contract durations, and measurement of capacity and cloud revenue. For the market, the key question is how quickly this model translates into higher utilization of NVIDIA systems in production inference, where revenue is recurring and tied to continuous demand rather than occasional training runs.