On July 6, semiconductor research firm SemiAnalysis released a report: Jensen Huang personally showcased the Kyber NVL144 cabinet at GTC 2026, which encountered a significant setback just three months after its debut, with the landing time delayed by more than 12 months, now pushed to 2028. The accompanying NVL72x2 back-to-back cabinet structure has been directly canceled, compressing the NVLink expansion domain of Rubin Ultra.

This news has become the largest single negative factor for the technology sector today.

Isolated, this is a product delay. Extending the timeline to 30 days reveals a clear retreat curve.

On June 10, SemiAnalysis issued a report to institutional clients, indicating that the large-scale shipment of NVIDIA's native 800VDC power supply architecture has been postponed until after 2028, and mass production of CPO (Co-Packaged Optics) may be delayed until 2028 or even 2029. The U.S. stock market's optical communication sector responded with a sharp decline, with AAOI dropping as much as 17% in a single day, Lumentum falling about 8%, and Himax, Navitas, and Wolfspeed also facing pressure.

On June 30, SemiAnalysis spoke again: the four-chip version of Rubin Ultra, which was prominently released at GTC 2026, was canceled due to manufacturing execution risks and changed to a dual-chip design, with actual computing power and memory bandwidth being about half of the original version. Behind this is TSMC's CoWoS-L advanced packaging reaching physical limits in four-chip and multiple photomask scales, with the successor CoPoS not expected to achieve mass production until the end of 2028.

On July 6, it was Kyber's turn, three reports, three retreats.

What is Kyber, and why is it so difficult?

Kyber is the next-generation cabinet architecture designed by NVIDIA for Rubin Ultra and the subsequent Feynman generation. Its core action is to rotate the computing tray 90 degrees and vertically insert it into the cabinet, like books on a bookshelf, and then replace the thousands of copper cables inside the cabinet with an orthogonal backplane PCB. According to the initial specifications from GTC 2025, a single cabinet has a power consumption of up to 600 kilowatts, supported by a brand new 800VDC power supply system.

This backplane is the most difficult PCB to manufacture in the industry. According to Jefferies' research report, it requires 78 layers of M9-grade material; and according to industry observers' disassembly of the GTC 2026 exhibition machine, the number of connector pins on a single mid-board exceeds 10,000, with the total number of NVLink pins in the entire cabinet exceeding 87,000. Any bending of a single pin could lead to the entire board being scrapped. There are only two or three suppliers globally capable of mass production at this specification.

Jefferies actually issued a warning on June 22: the Kyber backplane PCB solution is highly likely to be delayed until 2028, and in the worst-case scenario, completely canceled, leading to a downward revision of the global AI PCB market size forecasts for 2027 and 2028 by 5% and 11%, respectively, and a downward revision of CCL (Copper Clad Laminate) by 8% and 16% On June 23, following the debunked rumors that "NVIDIA requires PCB manufacturers to reduce prices by 10%", the PCB sector in both the A-share and Hong Kong markets experienced a panic sell-off. Today's report from SemiAnalysis effectively confirmed this warning.

The cancellation of NVL72x2 means that NVIDIA has abandoned the transitional plan of using two racks back-to-back to expand the NVLink domain. Rubin Ultra is likely to retreat to the mature Oberon architecture (i.e., NVL72 form) in 2027, with its expansion capabilities reverting to the previous generation framework.

Jensen Huang has emphasized that NVIDIA is the first technology company in history to announce a four-generation product roadmap all at once. The intention behind the early announcement was to give the supply chain ample preparation time: data center site selection, power supply transformation, and liquid cooling solutions all require upfront investments measured in years.

The side effect has manifested this year: the roadmap itself has become a tradable asset. The market prices the entire industry chain according to the timeline presented in the PPT, modeling optical modules based on CPO penetration rates, valuing PCBs according to the pace of backplane volume release, and scheduling production for power supply manufacturers based on the timing of 800VDC switching. As physical laws gradually retract these commitments, each correction corresponds to a sector-level repricing. Over the past 30 days, the optical communication, PCB, and power supply sectors have each gone through this process in turn.

Winners and Losers in the Industry Chain

The delay has changed the rhythm and rewritten the list of winners.

Copper cable manufacturers have received a "reprieve." The extended lifecycle of the Oberon architecture has allowed the copper cable demand, originally planned to be replaced by backplane PCBs, to be retained. Connector manufacturers like Amphenol are listed as relative beneficiaries within the SemiAnalysis framework, and Vertiv and Legrand have also received a relatively positive assessment.

The logic of upstream materials is the hardest. The supply tightness of fiberglass cloth and CCL is unrelated to Kyber; it is driven by demand across the entire industry, with copper-clad laminates experiencing four consecutive price increases within six months. Kyber's postponement only changes the demand structure, not the total demand, and pricing power remains in the hands of material suppliers.

PCB manufacturers bear the most direct pressure, with those caught in the middle facing the worst situation: high-end players have technical depth and customer stickiness, allowing them to follow specification upgrades; low-end capacities have cost advantages; mid-tier manufacturers are caught in between, and the elimination race is accelerating.

The overall time window for the optical communication and CPO industry chains has shifted backward. Shipments relying on the Rubin Ultra and Kyber platforms for Sidecar have been postponed to the 2028 window. SemiAnalysis maintains a cautious stance on Lumentum, Himax, Navitas, and Wolfspeed, while also indicating that some NPO (Near-Package Optics) projects may accelerate in reverse.

The larger narrative impact is on NVIDIA itself. The halving of Rubin Ultra specifications combined with the delay of Kyber has led some analysts to interpret this as a signal of erosion in NVIDIA's performance moat, with AMD and Google's TPU ecosystem being named as potential marginal beneficiaries This conclusion currently lacks sufficient evidence, but its entry into mainstream discussion is a change in itself.

Subsequent attention should focus on two confirmation signals: whether NVIDIA will positively respond to the timeline of Kyber and Rubin Ultra in the next earnings call; and whether there will be structural adjustments in the order guidance from Taiwanese ODMs and PCB manufacturers.

At the same time, it should be noted that the information regarding delays and cancellations mentioned in this article comes from third-party sources such as SemiAnalysis and Jefferies, and has not yet been confirmed by NVIDIA. Supply chain information has historically been subject to revisions, and NVIDIA's network business executives have previously expressed optimistic views on the CPO timeline that contradict third-party research. Until an official statement is made, consider it a scenario with a higher probability rather than a fait accompli.

Source: Golden Finance

The publisher assumes full responsibility for the content of this article.

Before investing in cryptocurrency, be sure to conduct thorough research, understand the associated risks, and carefully assess your own risk tolerance. Do not overlook the potential for significant losses due to the temptation of short-term high returns