NVIDIA KYBER DELAYED TO 2028, ASIAN SUPPLIERS HIT

Nvidia's next-generation AI server rack, the Kyber NVL144, will not arrive until 2028 at the earliest, according to analyst firm SemiAnalysis. The delay stems from circuit board manufacturing problems, pushing back the launch by more than a year from initial plans. The announcement hit Asian suppliers hard, with market valuations dropping by double-digit percentages following the news. These suppliers had positioned themselves to support Nvidia's production ramp. The delays extend beyond the base Kyber model. Nvidia has canceled the more powerful Rubin Ultra variant entirely, eliminating a planned premium tier in the product lineup. The postponement creates strategic opportunities for competitors. AMD and Google, both developing competing AI infrastructure solutions, may gain traction in the enterprise market during the extended gap. Data center operators seeking advanced AI capabilities cannot rely solely on Nvidia's timeline and may diversify their supplier base. Circuit board manufacturing, despite appearing commoditized, remains a critical bottleneck in AI chip deployment. The complexity of Nvidia's latest designs—requiring advanced packaging and interconnect technologies—has strained manufacturing capacity across the supply chain. Kyber represents Nvidia's roadmap evolution following its dominant Blackwell architecture. The delay signals that even Nvidia faces constraints in scaling production infrastructure to meet the accelerating demand for AI hardware. The company has not officially commented on the SemiAnalysis report. Nvidia's next earnings call will likely address supply chain questions and revised guidance for future product launches.