Nvidia is pulling the plug on revenue-sharing arrangements with AI cloud providers, a move that could reshape how smaller players in the artificial intelligence infrastructure space access the company’s most coveted chips.
What Nvidia built, and why it mattered
The revenue-sharing model, formally introduced on July 1, 2026, was essentially Nvidia playing venture capitalist and hardware supplier at the same time. AI cloud providers could procure Nvidia’s GPUs with financial support from Nvidia itself. In return, Nvidia would receive both traditional hardware sales revenue and a percentage of the cloud revenue those GPUs generated.
Initial participants included Australia’s Sharon AI, which was expected to deploy up to 40,000 Grace Blackwell GB300 GPUs over six years, and Indonesia’s Firmus Technologies in Batam, where deployment could scale to 170,000 GPUs with anticipated revenue in the range of $25 to $30 billion over six years.
The backstop business was already massive
