Nvidia shares swung in premarket after report said next-gen AI rack targets slipped

Nvidia’s stock moved sharply in premarket trading after a market report said the company’s next-generation AI “rack-scale” system is being pushed back by about a year, according to SemiAnalysis. The report centered on Nvidia’s Kyber rack architecture, which is intended to house large numbers of Rubin Ultra chips in a single deployment unit.

The Kyber system, as described in the report, is now expected to reach 2028 rather than the earlier timeline that traders had been working with. The change, if accurate, would shift when customers can begin large-scale installations that rely on the rack’s integrated performance and capacity benefits, rather than on stand-alone GPU shipments.

The same report framed the delay as the result of manufacturing problems encountered along the Kyber ramp. For investors, that distinction matters, because a delay attributed to production constraints can imply a slower pace of system shipments and a longer path to full-scale revenue contribution than a pure product-design change.

In trading, the news fed into broader “AI premium” positioning, with the report’s timing also overlapping with market attention on other chip names. One external market recap said attention around Nvidia’s delay chatter weighed on the perceived pace of next-cycle AI infrastructure build-outs, while other developments provided offsetting momentum in related stocks.

Nvidia has not publicly confirmed the specific Kyber manufacturing timeline in the materials cited by the market post. As a result, the delay figure should be treated as attributed to SemiAnalysis and relayed by news aggregators, not as an official schedule update from Nvidia itself.

Industry context also helps explain the sensitivity. Rack-scale systems are built to turn a data center’s power, cooling, and networking constraints into measurable throughput, and customers plan procurement around when these “system-level” offerings become available. Even when GPU supply is available earlier, a delayed rack-scale platform can slow adoption of the highest-density deployments that customers target for training and large inference workloads.

What to watch next is whether Nvidia or its customers provide any confirmation through product roadmaps, supply updates, or earnings commentary that addresses the Kyber timeline directly. Traders will likely parse subsequent disclosures for language about manufacturing ramp status, customer delivery schedules, and when Rubin Ultra-based configurations can move from planning to deployments at scale.