Amazon Web Services is raising the price it charges customers to rent Nvidia-powered AI computing, according to a report from Catherine Perloff in The Information.
The increase applies to EC2 Capacity Blocks, an AWS service that lets businesses reserve and rent AI compute in advance rather than scrambling for it when demand spikes. Per The Information, the company said prices for that service are going up by 20%.
Notably, the increase is targeted. The Information reports that pricing for Trainium — Amazon's own line of in-house AI chips — is unchanged. That gap matters: it makes renting Nvidia hardware on AWS meaningfully more expensive while keeping Amazon's homegrown silicon at the same price.
The move was also flagged by Techmeme and surfaced through Google News, both pointing back to The Information's reporting as the original source.
For a general reader, the takeaway is about the economics of the AI boom. Nvidia's chips remain the default engine for training and running large AI models, and access to them is scarce and costly. When the biggest cloud provider raises the price of renting that hardware, the added cost tends to flow downstream to the startups and enterprises building AI products.
Leaving Trainium prices flat also reads as a nudge. By widening the cost difference, AWS gives customers a clearer financial reason to consider Amazon's own chips instead of Nvidia's — a dynamic worth watching as cloud giants push to reduce their dependence on a single dominant supplier.
Why it matters: a 20% jump in the cost of renting Nvidia AI compute from the world's largest cloud provider raises the price of building AI and sharpens the competition between Nvidia's chips and the cloud providers' own homegrown alternatives.