On June 23, 2024, the PCB sector suffered an unusually sharp plunge amid rampant market rumors that triggered panic among investors. News claiming “NVIDIA demands a 10% price cut from PCB manufacturers” spread rapidly, followed by reports that “Sheng Hong Technology’s capacity expansion delays have held back shipments of the Rubin Platform,” pushing market fears to a fever pitch. Meanwhile, brokerage Jefferies released a research note stating that the Kyber backplane PCB solution, originally scheduled for rollout in 2027, might be postponed until 2028, further fueling market worries over the prosperity cycle of AI PCBs. However, in-depth interviews with multiple industrial chain experts, executives of listed companies and market participants reveal that most of these rumors contain obvious exaggerations and misinterpretations. This article debunks the truth behind the rumors from three dimensions—technical context, industrial logic and market landscape—and delivers a thorough analysis of their actual industrial repercussions.

 

 

 

In the era dominated by consumer electronics, PCB manufacturers relied heavily on major clients, wielded weak bargaining power and endured squeezed profit margins over the long run. The arrival of the AI computing cycle has completely rewritten this dynamic. High-end PCB capacity (including HDIs, packaging substrates and high-speed multilayer boards) remains undersupplied, shifting leverage firmly to suppliers. The chairman of a listed PCB firm told reporters: “Pricing power across the industrial chain currently rests largely with upstream material suppliers. Previously, the industry widely adopted capacity-locked quotation models with order cycles stretching six months to a full year. Amid sustained upstream material price hikes, supply chain pricing has shifted to real-time quotes, with order cycles generally shortened to 1–3 months.” Xiong Yiyu, Co-Chief Electronics Analyst at Huachuang Securities, also noted: “Product pricing is now negotiated bilaterally between PCB manufacturers and clients, with PCB suppliers retaining the upper hand at this stage.”

 

 

For cloud service providers (CSPs) pouring capital into AI infrastructure, cost is not the core pain point—capacity and reliable delivery are. A private equity portfolio manager analyzed: “Cloud vendors are not reluctant to spend; they fear supply shortages from upstream suppliers. Therefore, rather than pushing for price reductions, clients prioritize stable delivery capabilities and tend to secure supply via capacity lock-ins and support for capacity expansion.” This stands in stark contrast to the consumer electronics sector. AI servers carry high added value, making downstream clients far more tolerant of price levels and granting PCB manufacturers stronger negotiating clout. Even when clients propose price adjustments, these represent “guidance-oriented cost control” rather than coercive cuts, with both sides negotiating reasonably based on long-term partnerships.

 

 

Regarding the rumor that “slow capacity expansion at Sheng Hong Technology has delayed NVIDIA Rubin shipments,” multiple interviewees emphasized that the Rubin Platform hinges on dozens of critical links including GPUs, HBM, advanced packaging, PCBs, switches, power supplies, liquid cooling and cabinet systems. Verification, certification and system calibration delays at any single stage can disrupt the overall launch timeline. PCBs constitute just one segment, and Sheng Hong Technology only one supplier within the chain, limiting its standalone impact. Research from TrendForce identifies the genuine drivers of Rubin’s delay: lengthier-than-expected HBM4 certification procedures, adaptation work required for network transmission upgrades from CX8 to CX9, challenges to cabinet power management systems posed by drastically increased single-chip power consumption, and full-performance calibration for higher-spec liquid cooling thermal solutions. Small-batch shipments are expected to commence late Q3 2026, with volume ramp-up in Q4 2026. GB300 will remain the mainstream AI server platform across full-year 2026, while Rubin AI Servers will account for merely 10–20% of total shipments.

 

 

 

Jefferies’ research report pointed out that due to the extreme technical complexity of orthogonal backplane PCBs required for intra-cabinet interconnections, the Kyber backplane PCB solution—originally slated for deployment within Rubin Ultra in 2027 to replace partial internal cabinet cabling—may be pushed back to at least 2028. This means Rubin Ultra will most likely retain the Oberon architecture (NVL72) in 2027, which relies on copper cable interconnections instead of backplane PCBs. The Kyber delay will directly downgrade market size forecasts for PCBs and CCL (copper-clad laminates).

 

 

Jefferies’ calculations show that a Kyber delay until 2028 would trim projected 2027 global AI PCB and CCL market sizes by roughly 5% and 8% respectively. Should the delay extend further into 2028 or the solution be scrapped entirely, 2028 PCB and CCL market forecasts would fall by 11% and 16%. Nonetheless, the brokerage stressed that the Kyber postponement does not alter the long-term growth narrative for the PCB industry. Products such as switch boards and mid-boards continue to upgrade toward higher-grade materials including M9, M10 and PTFE, sustaining the broader industry upgrade trend. Post-2027, PCBs will remain indispensable as core scalable interconnection hardware for AI infrastructure.

 

 

The delay creates divergent outcomes across the PCB supply chain. PCB fabrication facilities may face mild pressure from reduced order volumes, yet upstream material segments such as fiberglass fabrics and CCLs will likely maintain robust profitability amid persistent supply constraints. Most material manufacturers can readily pass rising input costs downstream, supporting stronger earnings performance relative to PCB fabrication firms. On the other hand, an extended lifecycle for the Oberon architecture preserves copper cable demand originally slated for replacement by backplane PCBs, creating upside for copper cable manufacturers and bolstering their earnings forecasts. This explains why certain copper cable stocks rallied counter-cyclically amid the broader PCB sector selloff.

 

 

Overall, the June 23 PCB sector slump largely reflected an overemotional market reaction to unsubstantiated rumors. The claim that NVIDIA forced a 10% price cut lacks industrial grounding—three core facts, namely tight high-end capacity, pricing power retained by suppliers, and client prioritization of delivery stability, fully discredit this narrative. Attributing Rubin Platform delays solely to Sheng Hong Technology also oversimplifies the intricate dynamics of the full industrial chain. As for the postponed Kyber backplane solution, while it will dampen near-term PCB and CCL market size forecasts, it does not undermine the long-term investment case for AI infrastructure and may instead generate incremental demand for alternative interconnection solutions such as copper cables.

 

For investors, the critical task is to separate market noise from secular trends. Within the AI computing cycle, the strategic importance of PCBs as foundational interconnection components will only grow alongside rising compute density. Whether switch boards, mid-boards or next-generation high-spec backplanes, these products follow clear technology upgrade trajectories shielded by high material and manufacturing barriers. Firms with abundant high-end capacity and stable access to upstream material resources will keep capturing value from this structural growth story. Short-term market fluctuations may well present attractive long-term entry opportunities.