Nvidia offers start-up customers chance to swap compute power for revenue share - CNBC

NVIDIA has launched a new partnership program aimed at supporting rapidly growing AI startups by providing access to critical computing infrastructure through its network of AI cloud service providers... Read original →

NVIDIA’s revenue-sharing compute model effectively financializes GPU access, transforming chips from hardware into leased production infrastructure for AI startups. This accelerates consolidation of model training within top-tier cloud providers and subtly discourages foundries from supporting non-NVIDIA architectures at sub-3nm nodes. Geopolitically, deep integration with TSMC’s (Taiwan, China) EUV capacity heightens supply chain fragility and may invite antitrust scrutiny in the U.S. and EU over compute monopolization. Competitors like AMD and Amazon will likely respond with compute derivatives or OpenAI-backed alternatives, yet CUDA’s ecosystem dominance remains unshaken short-term. Over the next 18 months, the market will bifurcate: revenue-share-dependent vertical AI firms versus hyperscalers building custom chiplet-based clusters. The era of buying AI infrastructure is ending—NVIDIA has already positioned itself atop the new value chain.