Nvidia has told some of its largest customers that the prices of servers containing its artificial intelligence chips could rise more than 15%, Bloomberg News reported on August 22.
The warning was sent to major buyers of the company’s data center systems. It points to rising component costs rather than changes to the price of Nvidia’s own processors, according to the report.
Memory chips are the main driver behind the increase. Costs for the high-bandwidth memory used alongside AI accelerators have climbed sharply, feeding into the total price of finished server units.
Nvidia does not sell most of these servers directly. Its chips are integrated into complete systems by hardware makers and cloud providers, meaning the added cost is passed along the supply chain to end buyers.
The notification suggests customers should expect the higher figures on future orders. A price move above 15% would affect the budgets of large technology firms building out AI computing capacity.
Nvidia’s data center business has driven its rise to the top of the semiconductor industry, as companies race to expand the infrastructure that trains and runs large AI models. Demand for its chips has consistently outstripped supply.
The company recently deepened its role in funding that expansion, having reached a $500 billion arrangement with Wall Street firms to finance AI infrastructure. Higher server prices add to the mounting cost of the buildout for its customers.
Memory suppliers have redirected production toward high-bandwidth memory as AI demand grows, tightening availability of the chips and lifting prices across the sector.
Nvidia has not publicly confirmed the specific figures cited in the report. The company is scheduled to report quarterly earnings later this month, when executives are expected to address demand and supply conditions.