NVIDIA's AI Compute Partnership: From Chip Seller to AI Compute Bank with Revenue Sharing and Financial Guarantees

According to The Information, NVIDIA has launched a new business model that combines revenue sharing and credit support for emerging GPU cloud providers. Under the 'AI Compute Partnership' program, NVIDIA provides financial backstop commitments—promising to buy back unsold GPU capacity at agreed prices—in exchange for a percentage of the cloud providers' revenue. Initial participants include Sharon AI (planning to deploy up to 40,000 GB300 GPUs) and Firmus (building a 360 MW AI factory in Batam, Indonesia with 170,000 GPUs). NVIDIA has already invested billions in equity, lease guarantees, and capacity repurchase deals with CoreWeave and Lambda, aiming to reduce dependency on hyperscalers like Amazon and Microsoft while unlocking capital for AI startups. This marks a strategic shift from hardware vendor to the 'central bank' of AI compute.