Nvidia delays next-gen AI rack system Kyber NVL144 by over 12 months to 2028 due to PCB manufacturing issues.
The AMW Read
A 12-month delay on a marquee AI rack product from the dominant infra supplier is a high-significance structural event; novelty is high because it disrupts a stable cadence and opens competitive windows.
Nvidia delays next-gen AI rack system Kyber NVL144 by over 12 months to 2028 due to PCB manufacturing issues.
Nvidia has delayed its next-generation AI rack system, Kyber NVL144, by more than 12 months, pushing the target release to 2028. SemiAnalysis reports that the delay is caused by printed circuit board (PCB) manufacturing issues, impacting Nvidia's data center roadmap. The Kyber NVL144 is a high-density rack system designed for large-scale AI training and inference workloads, and its postponement creates a gap in Nvidia's product cadence.
Why it matters: This delay represents a significant hiccup in Nvidia's data center product cycle, which has been a primary engine of the AI infrastructure build-out. The Kyber line is Nvidia's answer to hyperscaler demand for denser, more power-efficient AI compute clusters. A 12-month slip means hyperscalers and cloud providers may need to extend their reliance on current-generation Hopper and Blackwell-based systems, or accelerate adoption of competing inference silicon from AMD, Groq, or custom ASIC players. The delay also opens a window for alternative rack-scale architectures to gain traction in the enterprise AI segment.
From a substrate perspective, this is a structural supply-side shock in the compute substrate. Nvidia's dominance in AI infrastructure has been predicated on its ability to deliver new hardware generations on a predictable cycle. A PCB manufacturing failure at this scale introduces uncertainty into the capital-compression arc for hyperscalers who have committed billions to Nvidia-based data center builds. The next 12–18 months will test whether the hyperscaler distribution moat Nvidia has built can withstand a multi-quarter product gap, or whether enterprises will begin diversifying their compute procurement strategies.
#Nvidia #KyberNVL144 #AIIinfrastructure #ComputeEconomics #DataCenter #HardwareDelay
