Chinese Firms Supply 30% of US AI Imports as Shenzhen Data-Center Stocks Double

Q: Beyond chips, what overlooked component shortage could be the next bottleneck for the global AI boom?

The next likely bottleneck is **multilayer ceramic capacitors, or MLCCs**. They are small, cheap-looking parts, but AI servers need them in huge volumes to stabilize power and filter noise around GPUs, memory and networking gear. Research cited in the reporting says AI servers use roughly 10 to 15 times more MLCCs than general-purpose servers, with some systems needing up to 28,000 per server. Nvidia’s newer AI platforms have sharply increased MLCC content as board complexity rises. The strain is already visible. Lead times for high-capacitance MLCCs have stretched beyond 20 weeks in parts of the market, and suppliers have raised prices. Analysts describe this as one of the longest MLCC upcycles on record. The issue is not mainly raw-material scarcity; it is that major producers in Japan, South Korea and Taiwan are shifting capacity toward higher-end AI and automotive parts, while new capacity takes time to add. This matters because chips cannot be deployed without the surrounding power-delivery hardware. A shortage of MLCCs can delay server boards, racks and networking equipment even if GPUs are available. It also fits the broader pattern highlighted in the China supply-chain story: the AI boom increasingly depends on “less sexy” components such as PCBs, optical modules, transformers and power systems, many of them supplied by Chinese manufacturers. If MLCC shortages deepen, they could become the clearest non-chip choke point in the global AI build-out.

3 articles · Updated · CNBC · Jun 28

Chinese companies now account for nearly 30% of U.S. imports of AI-related products, supplying data-center essentials from transformers and batteries to chemicals, optics and circuit boards rather than advanced chips.
22V Research said those AI-linked exports generated about half of China’s export growth this year, helping turn hardware makers into a major driver of mainland stock gains.
The ChiNext index has doubled over 12 months, and nine of the 10 biggest AI supply-chain names in the CSI 300—excluding CATL—at least doubled in value by May.
InnoLight and Foxconn Industrial Internet have reached roughly 1 trillion yuan market caps, while Foxconn said its cloud-computing business jumped 88.7% in 2025 to 602.68 billion yuan.
JPMorgan said optics is the key 'picks and shovels' segment for AI data centers, with some overseas orders stretching into 2027-28 and China’s fiber and cable export value running about four times year-earlier levels.

Sources

Left100%

CNBC3h ago

Chinese Firms Power US Data Centers, Drive Stock Gains

Bloomberg3h ago

Wall Street Bets Billions on Power Firms as AI Boom Drives IPO Rush - Bloomberg

Instagram3h ago

Data centers are becoming a lot more common in everyday American life. According to Pew Research Center, 38% of Americans now live within five miles of at least one operational data center.

5 Sources

Is the US trading chip supremacy for a critical dependency on China's foundational AI hardware?

Beyond chips, what overlooked component shortage could be the next bottleneck for the global AI boom?

How US AI Ambitions in 2026 Are Threatened by Chinese Supply Chain Dominance and Market Leverage

Overview

As of June 2026, the United States is increasingly reliant on Chinese firms for critical AI infrastructure components, not just finished products but also foundational elements like high-speed data transmission and raw materials. This dependence is highlighted by the rise of Zhongji Innolight, now the largest component of China’s CSI 300 Index, driven by soaring global demand. Zhongji Innolight supplies major US tech companies, including Alphabet, which alone accounts for a significant share of its sales. These market dynamics reflect a broader shift in both the US and China toward AI-centric industries, deepening supply chain interconnections and strategic vulnerabilities.

...