Nvidia's latest AI chips, known as Blackwell, are reportedly experiencing significant overheating problems when deployed in server racks designed for up to 72 chips. This has raised concerns among customers about the feasibility of launching new data centers on schedule.
Nvidia has been working with suppliers to modify the design of the server racks multiple times in order to address these overheating issues. Sources familiar with the matter indicate that Nvidia is actively collaborating with leading cloud service providers to resolve these challenges. A spokesperson for the company emphasized that such engineering iterations are a normal part of the development process.
These ongoing issues add to previous delays in the rollout of the Blackwell chips, which further complicates Nvidia's efforts in the competitive AI hardware market.