Nvidia warns AI servers will cost 15% more

Nvidia’s largest customers have been told that servers containing its AI chips may cost more than 15% extra from early next year, with the rise reaching systems built around Vera Rubin and Grace Blackwell, according to The Next Web’s report on the memory-driven increase. Contract manufacturers are passing the warning to operators including Microsoft, Google, and Oracle. Builders budgeting new clusters should treat memory as a live line item rather than a fixed server assumption, alongside the site’s analysis of AI infrastructure financing.