Analysis: PCB emerges as new critical bottleneck in AI infrastructure supply chain; high-end HVLP copper foil demand seen surging

Following the chip boom, the AI supply chain is now searching for its next critical enabler.

Over the past two years, the most prominent position in the AI industry chain belonged to chips. NVIDIA’s GPUs became the core asset of global capital markets, while servers, optical modules, and data centers—all tied to computing power—became recurring themes in A-share market trading.

But as 2026 approaches, a new shift is emerging. Market attention is moving from the upstream chip sector toward manufacturing-stage components. One such component, long considered part of traditional electronics manufacturing, is the PCB (Printed Circuit Board)—now becoming a key enabler in AI infrastructure upgrades.

Since September, the PCB supply chain has been consistently active. From high-speed PCBs to copper-clad laminates and copper foil materials, multiple listed companies have drawn market interest.

As of the close on September 23, Aohong Electronics maintained strong momentum, hitting the daily涨停 limit seven times in nine trading days. Daya Shengxiang, despite clarifying that its ultra-thin PCB fiberboard used in wood-based panels is not an electronic component and does not form part of a PCB—serving only as a protective pad beneath copper-clad laminates—still saw its stock price rise for four consecutive涨停 days. Other stocks, including Zhongyi Technology, Jitai Shares, Jiangnan New Materials, and Yidun Electronics, also hit their daily涨停 limits.

The market is no longer trading purely on “AI concepts.” It is now focused on a more tangible question: As AI computing power continues to expand, who can support the next generation of computing infrastructure?

The answer is pointing to a single circuit board.

01 The Underestimated PCB: A Bottleneck for AI Computing Power

The AI competition over the past two years has essentially been a race for computing power. Those with more GPUs and larger data centers possess stronger large-model training capabilities. Consequently, capital markets first focused on chips—AI chips, servers, and optical modules became the most sought-after segments of the supply chain.

However, as AI infrastructure expands, a new question arises: If GPU numbers keep increasing, where will the real bottleneck for AI development lie?

The answer is shifting from chips to the entire system. The reason is simple: AI servers are fundamentally different from traditional servers. Traditional servers rely primarily on CPUs for computation, while AI servers require large numbers of GPUs to work in parallel. This demands higher data transmission volumes, higher signal frequencies, greater power consumption, and more complex connectivity.

The more powerful the chip, the higher the demands on the infrastructure that supports it. This is where the PCB’s importance is being rediscovered.

Known as the “mother of electronic products,” the PCB traditionally served a connecting function. Smartphones need PCBs, computers need PCBs, and automotive electronics need PCBs. But AI servers are changing this logic. AI servers do not require ordinary PCBs; they need high-speed, high-frequency, low-loss PCBs capable of handling data transmission between GPUs, switching chips, high-speed memory, and power systems.

A single PCB now determines whether an entire computing system can operate efficiently. This is why value in the AI supply chain is migrating from chips to infrastructure.

02 Beyond Chips: The “Copper Foil War” in AI Hardware

If we further dissect a high-end PCB, we find that the real competition is not just in PCB manufacturing but also in upstream materials, which are becoming a new battleground.

Behind every high-performance PCB lie copper foil, fiberglass cloth, resin, and copper-clad laminates. Among these, copper-clad laminates determine PCB performance, and AI servers require high-frequency, high-speed materials. During high-speed data transmission, signal loss becomes a limiting factor, making low-loss materials increasingly important.

HVLP (High-Voltage Low-Profile) copper foil is a prime example. Compared to ordinary copper foil, HVLP copper foil reduces signal loss during high-speed transmission and is a key material for AI server high-speed PCBs.

According to a research report from Dongwu Securities, demand for high-end HVLP copper foil specifically for AI servers is expected to reach approximately 24,000 tons in 2026, a year-on-year increase of about 260%. Demand is projected to further rise to 50,000 tons in 2027.

However, high-end materials cannot be produced simply by expanding capacity. They require equipment upgrades, process accumulation, and customer certification. This means a new supply-demand imbalance is forming in the AI supply chain: not all PCBs are scarce; what is truly scarce are the high-end PCBs and materials that meet AI server requirements.

This is why capital markets are re-evaluating PCB companies. Wus Printed Circuit (Kunshan) Co., Ltd. is a typical example. In the past, investors viewed Wus as a PCB manufacturer. But in the AI era, the market is reinterpreting its value: the company’s core value lies not just in manufacturing PCBs, but in producing high-end PCBs that meet the demands of high-speed servers.

This shift means that PCB companies are transforming from traditional electronics manufacturers into AI infrastructure suppliers.

A similar transformation is happening among material companies. Shengyi Technology represents the copper-clad laminate direction. In the past, material companies often occupied a lower position in the supply chain. But in the AI era, a new logic is forming: materials determine performance ceilings. Without better materials, chip performance cannot be fully unleashed, and server speeds cannot be further improved.

Thus, AI competition is not only about chips but also about materials.

Meanwhile, companies like Dongshan Precision and Shenghong Technology are also making continuous investments in the high-end PCB space. They all face the same industry trend: growing demand for AI servers is pushing the PCB industry from scale-based competition to technology-based competition. In the past, lower costs gave an advantage; in the future, the ability to produce higher-end products will be the key differentiator.

03 AI’s “Shovel Seller” Now Belongs to PCB

Every industrial revolution follows a similar pattern. Initially, capital markets focus on the most dazzling technology. The internet era focused on platforms; the new energy era focused on vehicles and batteries; the AI era initially focused on GPUs and large models.

But as industries enter the scaling phase, markets begin to seek out infrastructure. The PCB is at this stage now. It may not be the most glamorous part of the AI industry, but it is an unavoidable link for AI’s large-scale deployment. After all, AI does not run on concepts—it runs on servers, servers run in data centers, and data centers run on a vast array of basic components.

Therefore, the AI supply chain will inevitably return to manufacturing.

However, capital markets must also recognize the other side. The PCB industry has experienced multiple cycles in the past: rising demand, capacity expansion, capacity release, and price declines. The biggest difference this time is that AI-driven demand is more concentrated in the high-end segment.

Companies that will truly benefit in the future need to meet three conditions:

1. High-end product capabilities
2. Access to core customer supply chains
3. Ability to convert demand into profits

Capital markets can trade on industry trends in advance, but ultimately, a company’s value is determined by orders and financial data. This is why the AI PCB rally will not simply replicate past electronics cycles. The future competition is not about who has more capacity, but about who possesses the most scarce manufacturing capabilities in the AI era.

Over the past two years, people searched for those who build the AI brain. Then they looked for those who connect the AI brain. Now, they are looking for those who support the AI brain’s operation.

A PCB may seem like just a circuit board inside an electronic device. But in the AI era, it carries more than circuits—it carries the entire computing infrastructure: from GPUs to optical modules, to PCBs, and further to copper foil, copper-clad laminates, and testing equipment.

Artificial intelligence will ultimately return to manufacturing. And a single PCB is becoming the new “shovel-selling business” of the AI era.

Special Notice: The content above (including any images or videos) is uploaded and published by a user of NetEase Hao, a social media platform that only provides information storage services.