As AI strikes from mannequin improvement to manufacturing inference, compute demand is accelerating and shifting towards constantly working AI factories that generate tokens at scale. This shift requires entry to giant‑scale, multi‑tenant accelerated computing that may come on-line rapidly, keep extremely utilized and help the economics of token‑scale AI companies.
Rising AI firms traditionally have had restricted entry to capital-intensive infrastructure, with even long-term commitments inadequate to unlock financing for compute.
To handle this, NVIDIA is introducing a brand new enterprise mannequin that opens up compute entry to the quick‑rising AI ecosystem of startups, mannequin builders, enterprises, analysis organizations and regional AI gamers.
This new mannequin permits AI clouds to acquire NVIDIA infrastructure for AI-native, enterprise and ISV clients by financial alignment with a revenue-sharing and credit-support mannequin. By the partnership, AI clouds will promote NVIDIA-powered cloud companies, with NVIDIA incomes each normal product income and a share of the cloud income on the supported capability. This construction accelerates adoption of NVIDIA platforms among the many high-growth, high-conviction AI native sector, and supplies NVIDIA with a recurring, usage-linked earnings stream.
For mannequin builders, inference suppliers, agent platforms and enterprises scaling AI, it could actually imply quicker entry to full-stack accelerated computing with out ready by web site choice, energy procurement, development and {hardware} bring-up.
NVIDIA AI Manufacturing facility Capability Constructed Round Demand
The initiative is already taking form, with AI cloud firms constructing DSX AI factories designed to serve clients and workloads throughout areas.
Sharon AI and Firmus are among the many first firms to work with NVIDIA on this new enterprise mannequin.
Sharon AI is deploying as much as 40,000 NVIDIA Grace Blackwell GB300 GPUs.
“This strategic collaboration with NVIDIA marks a pivotal second in Sharon AI’s mission to ship sovereign, large-scale AI compute infrastructure,” mentioned James Manning, cofounder and CEO of Sharon AI.
Firmus is constructing a DSX AI manufacturing facility campus in Batam, Indonesia. The campus is predicted to scale to 360 megawatts and as much as 170,000 NVIDIA GPUs.
“AI-native firms want entry to scalable, energy- and cost-efficient compute infrastructure to compete globally,” mentioned Tim Rosenfield, co-CEO of Firmus Applied sciences. “Firmus AI cloud is constructing a NVIDIA DSX-aligned AI manufacturing facility, which is able to allow our cloud to assist extra clients entry the compute they should construct and scale AI.”
AI natives akin to Baseten, Fireworks AI and Collectively AI present the place compute demand is headed: they want speedy entry to AI cloud capability to run mannequin coaching, post-training, fine-tuning and high-volume agentic inference for builders, digital natives and enterprises constructing with AI.
Their clients want dependable entry to large-scale NVIDIA accelerated computing as utilization grows, however additionally they want business flexibility as merchandise transfer from pilot to manufacturing.
To safe compute capability and construct and deploy AI fashions, contact Sharon AI and Firmus.
Study extra about NVIDIA Cloud Companions and AI factories.
