Nvidia‘s latest Blackwell AI chips, known for their groundbreaking performance, are facing significant challenges due to overheating issues in server setups.
Reports reveal that the chips overheat when installed in server racks designed to hold up to 72 units, causing concern among major customers like Meta, Google, and Microsoft. The delays and technical issues raise questions about data center readiness for these high-demand chips.
To address the problem, Nvidia has requested several redesigns of the server racks from its suppliers. Sources indicate that the company is working closely with its engineering teams and cloud service providers to resolve the overheating concerns. A spokesperson for Nvidia described the process as “normal and expected engineering iterations,” emphasizing their commitment to delivering a reliable product.
The Blackwell chips, unveiled in March 2024, promise to deliver 30 times the performance of Nvidia’s previous offerings, making them a key component for AI applications such as chatbot response generation. Originally slated for shipment in Q2 2024, the chips have already faced delays, and the overheating issues may further impact deployment timelines.
Despite these setbacks, Nvidia remains focused on addressing the challenges, aiming to meet the expectations of its customers and maintain its leadership in the AI hardware market. The outcome of these efforts will be crucial for companies relying on Blackwell GPUs to power their next-generation AI infrastructure.
