Bank of America responds to "Nvidia Kyber rack delays": The concerns are reasonable, but demand for PCB/CCL/substrates remains strong and sees the adjustment as a buying opportunity.

Bank of America responds to "Nvidia Kyber rack delays": The concerns are reasonable, but demand for PCB/CCL/substrates remains strong and sees the adjustment as a buying opportunity.

Before the market opened on July 6th, semiconductor industry research agency SemiAnalysis posted six consecutive tweets on the X platform, directly naming Nvidia Kyber NVL144 rack as facing a “major delay”, while Rubin Ultra was reduced from a four computing chip version to a two chip version, and the timeline for CPO (co-packaged optics) adoption is also facing uncertainty. The news quickly spread, and the stock prices of related sectors such as CCL (copper clad laminate), PCB (printed circuit board), substrate, and testing interface in Taiwan fell by an average of 8-9% on the day.

Bank of America Securities Taiwan technology team (analyst Mike Yang and others) promptly released a report stating: The technical concerns are reasonable, but the drop in stock prices is actually a good entry opportunity.

Concerns are reasonable, but different logical frameworks: demand scale-down ≠ demand disappearance

Analysts believe that potential delays or cancellations in the CCL/PCB/substrate sectors are essentially “demand scale-down in response to supply constraint”, rather than a reversal of demand trends. They compare this to previous market concerns about DRAM demand for Vera CPUs—a similar concern at that time also led to stock price fluctuations, but did not change the demand trend.

Why might this drop be an overreaction? Analysts provide a key logic: The cost proportion of CCL/PCB/substrate in the total AI server BOM (bill of materials) is low, only low to mid single-digit percentage. The low cost proportion means lower price elasticity—when customers cut orders or delay purchasing, it’s usually the high-value, high-cost components (such as memory) that are cut first, not basic materials like these.

This logic is supported by the supply side. Analysts state that high-end CCL (M8 grade and above) and ABF substrates are expected to remain in short supply at least until the end of 2027. The root cause is the long equipment delivery cycle, while demand driven by server CPUs and accelerators is structurally growing. In other words, even if short-term demand rhythm is disrupted by Kyber delays, the speed of capacity expansion on the supply side cannot catch up with demand in the short term.

Therefore, Bank of America categorizes the current share price drop as an “enhanced entry point”, maintaining a buy rating on related sectors.

Q glass substrate M9 will outperform PTFE; sub-sector resilience rank varies

The manufacturing challenges of PCB mid-board for Kyber rack directly point to the technological route competition for high-end CCL materials. The backplane involves M9 grade CCL, and there are currently two industry routes: M9 based on Q-glass (quartz cloth, hereafter Q glass) and based on PTFE (polytetrafluoroethylene), both schemes can improve dielectric constant (Dk) and loss factor (Df), but have different manufacturing bottlenecks.

Analysts note that the production challenges of these two schemes “lie in different links”—PTFE’s difficulty is on the CCL production side, and Q glass is difficult on the PCB processing side (yield improvement). This difference leads Bank of America to judge that Q-glass-based M9 grade CCL will penetrate AI servers faster.

If PCB-side yield improvement in the Q glass scheme proceeds moderately, the upside of high-end CCL demand (i.e., supply-demand gap) will actually be greater. This logic constitutes a potential benefit for CCL suppliers—the slower the yield climb, the harder effective demand is met, the higher the scarcity premium.

But not all sub-sectors are equally exposed to the impact of Kyber delay. Analysts provide an explicit order for resilience: Test interface/substrate > CCL > PCB, because solutions closer to the chip end are more directly driven by platform specification change and more closely tied to the AI server upcycle.

For CCL, differentiation between TUC and EMC advantages:

  • TUC: has more upside for market share expansion and pricing in the short to medium term
  • EMC: has a long-term advantage due to substrate material and M9-Q technology barriers

For substrate and test interface, Bank of America believes the negative impact of the “2 chip solution” on single substrate content can be offset by the overall increase in shipment volume. Additionally, the report especially points out, the logic of market concerns that memory customers will pressure substrate prices “does not hold” (unwarranted)—because BT substrate (bismaleimide triazine substrate) supply is continually tight, glass cloth costs are high, and memory demand remains strong, these three factors together support pricing resilience for substrates.

 

 

 

~~~~~~~~~~~~~~~~~~~~~~~~

The above highlights are from Zhuifeng Trading Desk.

For more detailed interpretations, including real-time insights and frontline research, please join Zhuifeng Trading Desk ▪ Annual Membership

Risk Disclaimer and TermsThe market has risks; investments need to be made cautiously. This article does not constitute personal investment advice, nor does it take into account individual users' specific investment goals, financial situations, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article are suitable for their particular situations. Invest at your own risk.