Nvidia’s next-generation Kyber rack-scale architecture, designed to house its 2027 Rubin Ultra chips, has been pushed back by more than a year to 2028, according to research firm SemiAnalysis.

The delay underscores growing friction between the chipmaker’s aggressive annual release cycle and the physical limits of global semiconductor manufacturing.

A major Taiwanese assembler of Nvidia AI servers recently flagged emerging component shortages, even as it reported second-quarter sales that beat expectations with a nearly 40% year-on-year increase.

The setback adds to mounting concerns about supply-chain constraints in the AI hardware sector.

A major Taiwanese assembler of Nvidia AI servers recently flagged emerging component shortages, even as it reported second-quarter sales that beat expectations with a nearly 40% year-on-year increase.

This divergence between strong demand and tightening supply suggests that bottlenecks are becoming a structural issue rather than a temporary glitch.

The global semiconductor supply chain faces a structural bottleneck that is unlikely to ease before 2029, as the construction of new fabrication plants requires two to three years to reach operational capacity.