Nvidia‘s new Blackwell AI chips, which have already experienced delays, are facing additional challenges with overheating issues in the server racks that house them, raising concerns among customers about the timeline for launching new data centers, The Information reported on Sunday.
The overheating occurs when the Blackwell graphics processing units (GPUs) are stacked together in server racks designed to hold up to 72 chips, according to sources familiar with the matter.
To address the issue, Nvidia has reportedly asked its suppliers to alter the design of the racks multiple times, as stated by Nvidia employees, customers, and suppliers with knowledge of the situation. The suppliers, however, have not been named in the report.
Also read | Smart ring leader Oura plans international push as CEO touts new features and thinking on hardware
In response, a Nvidia spokesperson told Reuters, “Nvidia is working with leading cloud service providers as an integral part of our engineering team and process. Engineering iterations are normal and expected.”
First unveiled in March, the Blackwell chips were initially slated for release in the second quarter but have since faced delays, potentially impacting major clients such as Meta Platforms, Google, and Microsoft.
The Blackwell chip combines two large silicon squares from Nvidia’s previous model into a single component, offering up to 30 times the speed for tasks like generating chatbot responses.
For more such updates, click here.
