BEIJING, Aug. 5 (Xinhua) -- The large AI model boom in China is fueling soaring demand across upstream sectors from industrial metals to high-end boards, and ushering in a new wave of opportunities and value reassessment.
As AI computing power demands high-frequency, high-speed, and low-loss signal transmission, the printed circuit boards (PCBs) sector has seen a massive surge in orders, according to Wang Chen, an executive with the Greater Bay Area National Technology Innovation Center.
Wang noted that top-tier PCBs are in short supply and orders to some leading manufacturers already booked through 2027.
Simply put, PCB is the "skeleton" and "nervous system" of electronic devices. The more advanced the PCB, the faster and more reliably AI computing power can perform.
The PCB industry is now grappling with widespread material shortages, with multiple upstream suppliers issuing price-hike notices. Prices for high-speed PCBs have more than quadrupled, forcing many downstream manufacturers to raise their own prices accordingly or even delay deliveries, industry insiders said.
"This signals that the AI computing power boom has tangibly reached the underlying hardware layer," said Gong Chao, a researcher at Tongji University's National Institute of Innovation and Development.
A fresh wave of industrial chain reactions in AI computing power infrastructure has begun, flowing from AI algorithms and large models to GPUs and optical modules, and now reaching foundational hardware like high-end circuit boards and other components, according to Gong.
He noted that the surge in orders for premium circuit boards sends three major signals. "First, AI-driven demand has transmitted from the chips to physical hardware construction, indicating cloud vendors and computing power operators are launching large-scale procurement of complete AI servers, as well as construction and equipment installation for intelligent computing centers."
The order boom also highlights a clear distinction in the industry chain. Price hikes mostly feature low-loss and high-speed products rather than the entire industry line.
Moreover, constrained by the current supply shortages and new supply cycle limitations, the surge of orders for high-end circuit boards is expected to persist for an extended period, according to Gong.
China's AI industry was valued at over 1.2 trillion yuan (about 176.7 billion U.S. dollars) in 2025, up 40 percent year on year, according to the China Academy of Information and Communications Technology (CAICT) under the Ministry of Industry and Information Technology.
As of June 2026, China was home to more than 6,600 AI companies, accounting for 15 percent of the global total. The country has developed a complete industrial system spanning foundational infrastructure, models and frameworks, and sector-specific applications, CAICT said.
Against the backdrop of rapidly growing AI inference demand, global AI server shipments in 2026 are expected to increase by 28.3 percent year on year, according to data from the global consultation firm TrendForce.
PCBs for AI servers are being upgraded to 20-30 layers or more in 2026, with stricter requirements for impedance control and crosstalk suppression. The product mix is shifting from mid- and low-end general-purpose boards toward high-layer-count, high-order HDI (High Density Interconnect) boards, and high-speed high-frequency boards.
"The ripple effect has now reached most segments of the AI industrial chain, all of which are experiencing robust demand and strong growth, and upstream metals will be the next to benefit," Gong said.
Data from the China Nonferrous Metals Industry Association showed that in the first half of the year, prices of nonferrous metals including copper, aluminum, tin, tantalum and indium climbed sharply, driven by AI computing infrastructure construction. Tin prices jumped 40 percent within six months, indium rose 60 percent, and tantalum soared 158 percent.
Amid the AI frenzy, however, experts also caution against potential risks. "Many upstream players in the AI computing chain, such as high-end PCB and optical fiber manufacturers, are now aggressively expanding capacity. Should large-scale AI model deployment fall short of expectations over the next two to three years, coinciding with the release of this new capacity, related sectors could face significant downside pressure," Gong warned. ■
