Nvidia's Blackwell AI chips overheat in server racks

Nvidia's Blackwell AI chips are facing another potential delay thanks to overheating problems in server racks designed to hold up to 72 GPUs. The Information (via Reuters) reported that Nvidia has instructed suppliers to make design changes to the high-capacity racks and is working with them to improve thermal performance. 

"Nvidia is working with leading cloud service providers as an integral part of our engineering team and process. The engineering iterations are normal and expected," a Nvidia spokesperson told Reuters. 

This isn't the first time Nvidia has faced a delay with Blackwell. In August of this year, Bloomberg reported that Nvidia had to rework the chip to ensure better compatibility with the Hopper H100 data centers.  

In March, Nvidia had promised that the new chips would be available in the second quarter of the year before it encountered the delays. The Information reports that Nvidia's partners like Microsoft, Google, and Meta now worry that the delay will hinder their timelines for the new chips deployed in their data centers, potentially delaying new iterations of their AI-based products.