NVIDIA Revenue Shared Infrastructure Framework Enables Global AI Factory Growth and Large Scale Grace Blackwell Deployments
As AI workflows transition from initial model development into high volume production inference, the worldwide demand for on demand compute has grown exponentially. To that end, NVIDIA unveiled a new commercial approach to expand infrastructure access to high growth startup environments and low resource regional markets. This approach, known as a revenue shared credit support framework, enables wholesale cloud providers to access high tier hardware without the daunting capital investment barriers traditionally associated with those assets.
This model offers cloud providers the ability to deliver GPU powered services by monetizing the hardware with existing and new customer bases. Under this commercial model, cloud providers partner with the hardware manufacturer to sell services driven by that hardware. In return, the manufacturer receives a recurring revenue share of the cloud revenue generated on the covered capacity. According to official corporate statements, this effort aims to accelerate the adoption of H100 based processing units and greenfield distribution enabling providers to monetize assets without heavy initial investments in land, power, and facilities.
The new commercial approach is being utilized by a host of cloud infrastructure vendors building out localized AI factories. Sharon AI is the first announced partner to implement this strategy with a planned scale of 40,000 NVIDIA Grace Blackwell GB300 nodes. The compound was designed to deliver localized, secure compute globally. James Manning, cofounder and chief executive officer of Sharon AI, noted the industry impact of these components:
"This strategic collaboration with NVIDIA marks a pivotal moment in Sharon AI’s mission to deliver sovereign, large scale AI compute infrastructure."
In another application, Firmus is utilizing this new approach to develop a massive DSX aligned campus located in Batam, Indonesia. The site is poised to reach a scale of 360 megawatts of energy capacity and house 170,000 NVIDIA graphics processing units. Tim Rosenfield, the joint CEO of the platform, explained the shift in demand:
"AI native companies need access to scalable, energy and cost efficient compute infrastructure to compete globally. Firmus AI cloud is building a NVIDIA DSX aligned AI factory, which will enable our cloud to help more customers access the compute they need to build and scale AI."
New industry demand sources such as Baseten, Fireworks AI, or Together AI reflect a shift for capacity that is immediate, highly scalable, and capable of running fine tuning, post training, and high volume agentic inference. This commercial approach supports these new service providers by allowing them to maintain flexibility as their software tools shift from small pilots into scalable production at the token scale.
