NVIDIA's Next-Gen AI Server Rack Reportedly Faces Production Bottleneck, Launch Delayed to 2028

(Central News Agency, San Francisco, July 6, International News) Semiconductor research firm SemiAnalysis has revealed that NVIDIA, the giant in artificial intelligence (AI) chips, has postponed the launch of its new flagship product, the Kyber server rack architecture, originally scheduled for 2027, by 12 months to 2028, due to production bottlenecks in key circuit boards.

According to a report by American financial media CNBC, this is the latest setback for NVIDIA (NVIDIA), raising questions about its product roadmap.

The Kyber NVL144 is a server rack that integrates 144 of NVIDIA's most powerful chips into a single unit for collaborative operation, providing the immense computing power required for AI companies to train and run the most advanced models. The architecture was originally planned to be released in 2027, simultaneously with NVIDIA's next-generation server rack system, Vera Rubin Ultra.

SemiAnalysis stated in a post today that the delay is attributed to bottlenecks in the production of key circuit boards. "The Kyber NVL144 server rack architecture has been postponed to 2028 (launch) due to challenges in the manufacturing of printed circuit board midplanes."

A PCB midplane is a specialized multilayer printed circuit board used to connect electronic modules within a system.

NVIDIA has not yet responded to CNBC's request for comment. (Compiled by Hung Chi-yuan) 1150706

Stand with the facts. Your every sponsorship is a force to protect press freedom.

Download the Central News Agency "One-Hand News" APP to get the latest news in real-time.

This website's text, images, and videos may not be reproduced, publicly broadcast, publicly transmitted, or used without authorization.