Nvidia’s ambitious plans for its next-generation data center infrastructure have hit a snag, with reports indicating a significant delay for the Kyber NVL144 rack, specifically tailored for the Rubin Ultra platform. Originally slated for an earlier release, the deployment of this crucial component has been pushed back by more than 12 months, with analysts now forecasting its arrival in 2028.
Affiliate contentGames up to -90% off
Instant key delivery on Instant Gaming
Browse deals →The primary culprit behind this unforeseen setback, according to the insightful analysis from SemiAnalysis, lies in persistent issues with the PCB midplanes. These components are fundamental to the rack’s structural and electrical integrity, facilitating the high-speed data transfer necessary for Nvidia’s advanced AI and HPC architectures. Technical challenges in perfecting these midplanes have proven more formidable than initially anticipated, leading to the revised timeline.
Furthermore, the delay has reportedly forced Nvidia to scrap an interim solution that was designed to bridge the gap until the Kyber rack was ready. This stopgap measure was reportedly abandoned due to substantial pushback and dissatisfaction from key customers, who likely have stringent performance and reliability requirements that the temporary solution could not meet. This situation underscores the critical importance of robust and reliable hardware in the high-stakes world of AI infrastructure, where any compromise can have significant operational and financial repercussions for clients. The extended delay signals Nvidia's commitment to delivering a polished and fully functional product, even if it means extending the wait for its cutting-edge technology.
