Nvidia AI server delay to 2028 hits Asian supplier stocks

Asian technology suppliers fell hard Monday after a report said Nvidia Corp.'s next AI server rack was slipping to 2028, forcing investors to price manufacturing risk more directly into the chipmaker’s release schedule.

The delay centres on Kyber NVL144, a rack system tied to the next Rubin Ultra platform, and was caused by manufacturing difficulties, Bloomberg reported. The first sell-off came in Asia for a reason: printed-circuit-board makers and component suppliers are the listed companies closest to changes in Nvidia’s build calendar, while the US parent stock had not yet opened after the July 4 holiday.

The supplier moves were abrupt. Japan’s Ibiden fell 10 per cent, Hong Kong-listed Kingboard Laminates dropped 18 per cent, Taiwan’s Elite Material lost 10 per cent and South Korea’s Samsung Electro-Mechanics slid 11 per cent, according to Bloomberg. A Bloomberg basket of related shares has now lost 8.5 per cent over two weeks.

In a SemiAnalysis post on X, the research firm wrote: “MASSIVE DELAY: Just 3 months after Jensen demoed Kyber NVL144 at GTC, it has faced major setbacks and has been delayed by more than 12 months, pushing it back to 2028.” That wording left little room for investors to treat the issue as a routine schedule adjustment.

The market question is whether Kyber is a contained timing problem or a warning about the pace at which Nvidia can stack new systems on already strained supply lines. Shawn Oh, head of Korea cash equities at NH Investment & Securities, told Bloomberg the delay was “raising uncertainty around Nvidia’s next-generation scale-out roadmap and creating a wider competitive window for alternative AI platforms.” Suppliers would trade on that risk before many software or cloud names because their orders sit closer to the factory floor.

Gary Tan of Allspring Global Investments took a cooler view of Monday’s price action. He told Bloomberg the selling also reflected profit-taking, sector rotation and caution before the US reopened after the holiday. Many Nvidia-linked Asian hardware names had come into the session after months of AI-driven gains.

The demand case has not gone away. CNBC reported that SemiAnalysis still expects Nvidia’s data-centre compute revenue run rate in the second half of fiscal 2027 to be about 20 per cent above Wall Street consensus, even after factoring in the Kyber delay. That leaves investors with an awkward split screen: AI hardware orders remain strong, while delays deeper in the manufacturing stack can decide which suppliers absorb the first hit.

There is history behind that caution. A Reuters report from November said supply snags were already constraining Nvidia deliveries while demand boomed. The new Kyber timetable pushes that bottleneck debate beyond Blackwell-era shipments and into the next platform cycle, where schedules matter because customers and suppliers plan capacity years ahead.

Asia’s hardware names are, for now, the cleanest real-time read on that risk. A short Kyber delay would make Monday’s drop look like a fast repricing of crowded winners. A longer break in Nvidia’s release rhythm would mean manufacturing risk is starting to matter almost as much as AI demand.