According to the Information, new Nvidia AI processors are overheating servers

Nvidia's new Blackwell AI chips, which have already experienced delays, are now facing issues with server overheating, raising concerns among some customers about the ability to prepare new data centers in time, according to a report by The Information on Sunday.

The Blackwell GPUs reportedly overheat when connected in server racks designed to house up to 72 chips, as per sources familiar with the matter. To address this, Nvidia has requested several design modifications from its suppliers to mitigate the overheating problem. This information was provided by Nvidia employees involved in the project, as well as customers and suppliers with insight into the situation, though the suppliers were not named.

"Nvidia is collaborating with leading cloud service providers as an essential part of our engineering process. Such engineering iterations are normal and anticipated," stated a company spokesperson in a comment to Reuters.

The Blackwell chips were initially introduced in March, with Nvidia planning to begin shipments in the second quarter. However, delays have arisen, potentially affecting key customers such as Meta Platforms, Google (a subsidiary of Alphabet), and Microsoft.

The new Blackwell chips combine two silicon squares, each the size of Nvidia's previous models, to form a single unit that is 30 times faster at tasks such as generating chatbot responses.

Paraphrasing text from "Reuters" all rights reserved by the original author.