Overnight, Asian tech stocks slumped after a report that Nvidia’s next-generation AI server rack system has been delayed by more than a year due to manufacturing difficulties (we profiled it back in May in "Nvidia's Vera Rubin Rack Will Cost $7.8MM: Here's What's In It".)
Research firm SemiAnalysis said in an X post that Nvidia’s Kyber NVL144 hit setbacks in the construction of printed circuit boards for the platform.
MASSIVE DELAY: Just 3 months after Jensen demoed Kyber NVL144 at GTC, it has faced major setbacks and has been delayed by more than 12 months, pushing it back to 2028. Below, we explain why Kyber has faced massive delays and why NVIDIA’s NVL72x2 back-to-back rack architecture was… pic.twitter.com/VYduxnu01B
— SemiAnalysis (@SemiAnalysis_) July 5, 2026
According to the tweet, the PCB midplane at the center of the Kyber design remains too difficult to produce reliably. As we noted two months ago, the Kyber design consolidates 144 of Nvidia's most powerful chips into a single cabinet, enabling them to operate as one unified system, and was originally expected to arrive alongside Vera Rubin Ultra in 2027.
Meanwhile, cloud providers pushed back against the design over operational complexity, SemiAnalysis said, leaving Nvidia without a tested path to expand the scale-up architecture for Rubin Ultra.
In response, Japan’s Ibiden, a PCB maker that counts Nvidia as its largest client, drop
