NVIDIA Revenue Sharing AI Cloud Debuts With 210,000 GPUs: Flywheel or Vendor Finance Risk

NVIDIA has launched a GPU revenue-sharing and credit-support model for AI cloud operators, beginning July 1, 2026, that allows the chipmaker to take a recurring share tied to deployed GPU capacity. The article says NVIDIA published a blog post on that date, co-authored by CFO Colette Kress and Raj Mirpuri, describing how NVIDIA backstops GPU infrastructure buildouts by smaller cloud providers in exchange for a cut of resulting cloud revenue. It points to early participants, including Australian firms Sharon AI and Firmus Technologies, with commitments covering up to 210,000 GPUs. The report argues the structure is designed to solve a financing problem driven by uncertain GPU residual value as new architectures arrive, noting NVIDIA’s successor platform to Blackwell is already in production. By providing buyback-like guarantees for unsold capacity, NVIDIA effectively sets a value floor to make deployments bankable for lenders. The article frames the model as both “flywheel” acceleration and a vendor-finance risk pattern.