Nvidia halts revenue-sharing agreements with AI cloud providers

NVDA PAUSES REVENUE-SHARING DEALS WITH AI CLOUD PROVIDERS The company stepped back from the program last week, less than two months after announcing it, per people familiar with the matter. Some employees told current and potential customers the program could draw antitrust scrutiny, and cited sensitivities around how far Nvidia can dictate the way its customers do business. The precise reason for the decision could not be learned. An Nvidia spokeswoman said the business model introduced in July that opens up compute access to the AI ecosystem is still in place and continues to evolve on high demand. How the AI Compute Partnership worked: - Nvidia promised to rent GPU capacity itself if a provider could not find another customer, giving smaller providers guaranteed revenue to borrow against - Nvidia and the provider would set a base hourly GPU rate covering costs including chip depreciation, data center expenses and staffing - Nvidia would take 50% of any revenue above that threshold - Sharon AI and Firmus Technologies were the first two named participants Nvidia disclosed the scale for the first time in this week's quarterly filing: $36B of commitments under agreements typically running six years, decreasing as providers sell capacity to other customers. CFO Colette Kress told investors Wednesday the revenue stream could generate billions over the medium to long term. Some providers pushed back during the first weeks. Nvidia told some they could rent chips only to approved customers, and indicated it preferred capacity spread across several smaller AI companies rather than one large customer. Nvidia also recently scaled back a proposed financial backstop for OpenAI's data center project in Ohio. Per WSJ.