You sing your part, and I take the stage.

Investor Network, Wu Wei

After the popularity of "Yi Zhongtian," the main line of the AI computing power sector has once again experienced a rotation. Since June, the A-share market's PCB (printed circuit board) and copper foil/copper-clad laminate sectors have seen a series of "limit-up" or significant price surges. Behind this unusual movement are multiple factors, including the upgrade of AI computing power infrastructure specifications, global price increases due to shortages of key raw material supplies, and the entry of new technologies (such as glass substrates) into their commercialization phase.

In this round of market activity, leading companies with advanced production capacity, such as WUS (002463.SZ), SCC (002916.SZ), and core material supplier Tongguan Copper Foil (301217.SZ), have become the focus of capital attention. However, amidst the high market sentiment, investors must remain vigilant against chasing high prices, price corrections after the recovery of raw material supplies, and potential phase-overcapacity risks that may arise after future concentrated capacity releases while searching for main lines with high performance elasticity.

From GPU chips to optical communications to basic materials, what industrial logic lies behind the construction of AI computing power and industry rotation?

Supply-demand mismatch and specification leap: The underlying logic of sector movement and stock price performance

In the market performance of the first half of 2026, the PCB and materials sectors have shown significant dual increases in performance and valuation. Among them, WUS (002463.SZ) has seen a cumulative increase of over 80% this year, while Shenghong Technology (300476.SZ) has experienced more than a sevenfold increase since 2025, and Tongguan Copper Foil's stock price has tripled within two months due to the price surge.

The core logic behind this strong stock price performance may lie in the leap from "quantitative change to qualitative change" achieved by AI server specifications. With technological advancements, the AI computing power architecture underwent a profound transformation in 2026, with PCBs evolving from mere carriers for connecting electronic components to becoming the core bottleneck determining the signal integrity of computing power systems.

Previously, traditional general-purpose server PCBs typically had 12 to 16 layers, but with the increasing expectations for mass production of next-generation architectures (such as the Rubin platform and GB200 system) from companies like NVIDIA, the PCB layer count for AI servers is generally required to be above 20 to 30 layers, with some core backplanes even reaching 40 to 50 layers The significant increase in layer requirements will undoubtedly enhance the technology premium of related companies. According to research reports from Morgan Stanley and CITIC Construction Investment, with the commercialization of the new generation architecture, the PCB value of a single high-end AI server could reach 8 to 10 times that of traditional servers.

In addition, in the first half of 2026, the upstream PCB industry faced a rare supply shortage. At the end of March, the industrial zone in Saudi Arabia, which supplies about 70% of the world's high-purity polyphenylene ether (PPE) resin, halted production for various reasons, directly causing a shortage of high-end copper-clad laminate substrates, with some high-end PCB product prices soaring by as much as 40% in a single month.

At the same time, leading companies such as Copper Crown Copper Foil raised processing fees, with their HVLP (ultra-low profile) series high-end copper foil prices reaching 10 times that of ordinary copper foil and spot supplies being scarce. Leading copper-clad laminate companies like Shengyi Technology (600183.SH) benefited from this, with the company's market value surpassing 400 billion yuan in mid-June.

Authoritative research institutions both domestically and internationally have also provided expectations for this trend. Prismark predicts that although the overall global PCB market value will maintain a compound growth rate of about 5.4% in the coming years, the PCB sector related to AI servers will maintain an annual compound growth rate of over 20% to 30%.

CITIC Securities pointed out that the demand for high-layer count boards (HLC) with more than 20 layers will experience explosive growth and may become the most certain track in the industry. In addition, the TPCA (Taiwan Printed Circuit Association) believes that due to the high cost of investment in high-precision equipment, the industry threshold has significantly increased, and market share is rapidly concentrating towards leading manufacturers such as WUS and SCC. The business model of the PCB industry is being completely reshaped from "outsourcing" to "deep integration" in early chip research and development.

From computing power chips to key substrates: The rotation and performance realization of the AI industry chain

In fact, since the outbreak of the AI wave at the end of 2022, the capital market has undergone multiple rounds of sector rotation around computing power infrastructure. Its core path may follow the technological development pattern of "from core chips to peripheral infrastructure, from hardware support to material innovation."

Looking back at this round of the AI wave, the first wave of rotation began in the first half of 2023, with funds initially flowing into core computing power chips and model ends, represented by companies like Foxconn Industrial Internet (601138.SH) and Inspur Information (000977.SZ), which are GPU foundry and server manufacturers, gaining market favor; the market logic at that time was to seize the first confirmed computing power shortage dividend.

Entering the second half of 2023, the market began to realize that the data transmission speed of computing power clusters became a bottleneck, and companies like Zhongji Xuchuang (300308.SZ) and Tianfu Communication (300394.SZ), which produce optical modules (CPO/800G), welcomed substantial performance realization. Subsequently, the HBM (high bandwidth memory) and liquid cooling heat dissipation sectors, which solve the "storage wall" and "heat wall" issues, such as Saiteng Co., Ltd. (603283.SH) and Invec (002837.SZ), took over the baton of the upward trend As we enter 2024 and beyond, the main axis of market rotation has officially sunk to the fundamental physical level, namely PCB, copper-clad laminates, and key underlying materials. The performance realization logic during this stage appears to be more brutal, with higher technical barriers.

As data transmission rates approach 112Gbps and even 224Gbps, the signal loss of traditional materials has become intolerable. This forces PCBs to adopt M7 and M8 level ultra-low loss copper-clad laminates. At the same time, due to the "skin effect," high-frequency currents tend to flow on the surface of the copper foil, and rough traditional copper foils can lead to severe signal attenuation. Therefore, HVLP copper foil with a surface roughness of less than 1.5μm has become a mandatory standard for AI computing power.

In this round of material-level market represented by Shengyi Technology and Tongguan Copper Foil, the market's recognition highly depends on the real technical thresholds of enterprises. Research data shows that the capacity utilization rate (operating rate) of leading PCB and copper-clad laminate companies has generally exceeded 100%, operating in an overloaded state.

Companies that can penetrate the supply chains of top-tier computing chip manufacturers and possess mass production capabilities for high-end HDI and ultra-low profile copper foil are enjoying substantial performance dividends; while mid-to-low-end manufacturers that cannot break through the technological gap face the risk of being marginalized by the market.

Crossing Physical Limits: New Bottlenecks in Computing Power Development and Future Market Hotspots

In the current hype surrounding AI computing power, forward-looking industrial capital has begun to seek the next generation of technological paths that can break the physical limits of computing power. As the performance of traditional organic resin materials and ordinary conductive media is pushed to the limit, the future market focus will inevitably shift towards more fundamental technological disruptions.

The market is first focusing on the generational transformation in the packaging field—glass substrates. Traditional organic substrates tend to suffer from severe warping issues when dealing with large-area, high-density AI chip packaging. Glass substrates, due to their extremely low dielectric loss and thermal expansion coefficient (CTE) highly matched with silicon wafers, are seen as the only solution to this pain point.

2026 is regarded as the inaugural year for the substantial commercialization of glass substrates, with TSMC having established relevant pilot lines and giants like Intel clarifying their shipping timelines. In the A-share market, companies like Wogang Optoelectronics (603773.SH), which layout the entire glass substrate industry chain, and Han's Laser (002008.SZ), which involves TGV (Through Glass Via) processing, have seen rapid stock price increases, reflecting the market's optimistic pricing for new technology penetration.

As the power consumption of AI data centers rises exponentially, power supply and distribution efficiency are gradually becoming the biggest physical bottleneck limiting computing power deployment. Market logic is forming a consensus that "the end of computing power is electricity." In the future, ultra-high voltage equipment, high-speed power supply modules (PSU), and intelligent distribution systems may become highly potential incremental segments in the industrial chain. The photovoltaic and energy storage sectors are also becoming increasingly optimistic as a result.

Currently, the high integration of AI chips places extreme demands on packaging materials. For example, the spherical silicon micro-powder filler supplied by Lianrui New Materials (688300.SH) is key to reducing signal loss in high-end AI boards; Wanhua Chemical (600309.SH) specializes in resins, and Dinglong Co., Ltd. (300054.SZ) focuses on advanced packaging materials, both of which have very high technical barriers.

These advanced chemicals, which currently have a low domestic production rate, hold significant potential for domestic substitution driven by supply chain security. The recent stock performance of companies like Lianrui New Materials and Dinglong Co., Ltd. indicates that the market has already made its moves.

Overall, the recent fluctuations in the PCB and copper foil sectors represent a reasonable pricing by capital for the deep restructuring of AI hardware infrastructure. However, against the backdrop of rapidly expanding valuations and leading companies' dynamic price-to-earnings ratios generally returning to historical highs, the market is entering a deep water zone transitioning from "expectation speculation" to "performance verification."

As investors focus on leading companies with "technological advantages, customer orders, and self-developed materials," it may be wise to maintain a degree of restraint and calm, closely tracking the actual implementation of financial report data to avoid the valuation regression risk after the "concept speculation" tide recedes.

What are your thoughts on the extension of the AI industry chain? Feel free to leave a message, comment, or share. (Produced by Siwei Finance)■