Nvidia's New Blackwell AI Chips Overheat in 72-GPU Servers, Sparking Data Center Delay Fears

Nov 17, 2024, 02:49 PM

Nvidia's latest AI chips, the highly anticipated Blackwell GPUs, are reportedly facing overheating issues in servers, raising concerns about potential delays in data center deployments. According to reports from The Information, the company has asked its suppliers to redesign custom server racks multiple times in recent months to address the overheating problems. The server racks, designed to hold up to 72 GPUs each, are encountering thermal challenges due to the high density of the hardware. This development has made Nvidia's largest customers, including Meta, Google, and Microsoft, anxious about getting their supercomputers operational on time next year.