The multilayer chip capacitor (MLCC), known as the "rice of the electronics industry", is a necessary basic component for all electronic devices. Over the past few decades, its cycles of ups and downs have been highly correlated with the replacement cycles of consumer electronics such as smartphones. However, the market evolution since 2026 has presented a completely different driving characteristic: The explosive demand for AI computing infrastructure has led to a structural shift in the supply and demand pattern of this basic passive component.
The three major Taiwanese companies have confirmed that they may be facing the longest supply shortage in history.
According to reports from Taiwan's "Business Times" and other media, in recent shareholder meetings and public discussions held by technology companies, the three major MLCC/passive component manufacturers, namely KGI, Wuxinke, and Heshentang, unanimously confirmed that this round of AI-driven shortage cycle will surpass the passive component super cycle of 2017-2018, and may become the longest supply-demand imbalance in history.
Among them, the order-to-shipment ratio (B/B value) of Sunmi has risen above 1.3, and it is reported that it has surpassed most of its competitors. According to the Business Times, it is said that the chairman of Sunmi, Chen Tai-ming, has issued an order to increase production, requiring the production lines to fully reach the MLCC operating rate. The company will adopt a three-pronged approach of increasing operating rate, preparing for inventory, and eliminating capacity bottlenecks to meet the order wave. However, the official communication channel of Sunmi has not confirmed this news yet.
Huaxinke was the most definite "setter" in this round of supply shortage. The general manager of Huaxinke, Zeng Mingcan, directly extended the timeline of the shortage to after 2027, stating that the first wave of demand comes from AI data centers/server infrastructure, and the second wave comes from AIPC and edge devices - the latter potentially has a potential shipment volume of up to 200 million units. Once the memory supply is smooth, the shortage effect will further intensify.
Chairman Tang Jinrong of Heixin Tang disclosed at the shareholders' meeting that the delivery cycle for high-end passive component equipment has now reached 1 to 1.5 years. The equipment that Heixin Tang plans to introduce in the second half of the year is actually the purchase orders that were released two years ago. They will be installed in phases in the third quarter of this year and the first quarter of next year. Based on this, Heixin Tang's production capacity is expected to increase by 20% to 30% this year and by another 30% to 40% in 2027. Tang Jinrong also warned that the supply-demand gap is likely to widen next year.
The institution's assessment: MLCC may become "the next storage technology"
At present, with the explosive growth in AI computing power driving a surge in the demand for MLCCs, the A-share MLCC概念股 have received active attention from institutions. Among them, from April 1st to June 8th, Hongxing Development, Shunlv Electronics, Hongyuan Electronics and Guoqi Materials received a total of 4 or more research visits.
The concept sector of East Finance shows that currently, there are nearly 30 stocks in the A-share market related to the MLCC concept. The combined market value exceeds 1.1 trillion yuan. Sanhuan Group and Xiamen Tungro are at the forefront in terms of scale, while companies such as Xinfu Communication, Fenghua Hi-Tech, Guoqi Materials, and Dongcai Technology all have a market value of over 60 billion yuan.
Furthermore, multiple reports relayed the core conclusion expressed by Goldman Sachs' research report/analysts, referring to MLCC as "the next storage technology":
-
In the current cost composition of AI server material list, MLCC has risen to become the third largest cost item, second only to GPUs and storage chips.
-
Currently, the overall market size of MLCC is approximately 15 billion US dollars, among which the market size in the AI server sector is 1.3 billion US dollars, and it is experiencing a strong growth with a compound annual growth rate of 80%. However, the growth in other key industries such as automobiles and mobile phones has slowed down.
-
The AI-driven MLCC super cycle has just begun. Due to the significant increase in demand for high-end MLCCs driven by AI servers and new energy vehicles, coupled with the long construction cycle of high-end production lines and the difficulty of quickly releasing short-term production capacity, there is a significant gap between supply and demand for high-end models, and MLCC is entering a prosperous cycle
Correspondingly, companies such as Morgan Stanley have conducted a BOM (Bill of Materials) analysis of the Nvidia platform. Compared with the previous generation platform and the Nvidia GB300 platform released in 2025, the cost of MLCC (Multichip Module) materials for the VR200 single-computer cabinet has increased from $1,530 to $4,320, an increase of 182%. This increase is second only to memory (an increase of 435%) and PCB (a 233% increase in circuit boards). At the same time, the usage of MLCC has increased significantly. For the high-density configuration of the GB300 platform (such as NVL72, containing 72 GPUs), the MLCC usage per single-computer cabinet is 440,000 pieces, while for the VR200 single-computer cabinet, it rises to 600,000 pieces, an increase of 36.36%.
Compared with the price increase wave in 2017-2018: What is the same and what is different?
(1) Common points:
The winner structure at the top of the supply chain remains unchanged. According to the 2024 amount-based CR5, it is approximately 77.3%, Murata about 31.8%, Samsung Electro-Mechanics about 22.9%. The high-capacity and ultra-small size/high-temperature-high-pressure specifications are still dominated by the Japanese and Korean systems. The Taiwanese company TAIYO has formed an integrated passive components platform through mergers and acquisitions. Domestic manufacturers are mainly focusing on mid-to-low-end mass production and some high-capacity ramp-up stages to catch up - this is the same as in 2018. The profit pool that benefits the most directly is still at the top of the oligopoly.
The amplification mechanism between the channel market and the spot market, where "extended delivery period → hoarding / advance purchasing → accelerated prices", is also similar. However, in this round, the amplification is on the narrower and more certified channel of AI servers + high-end automotive-grade products.
(2) Differences:
The trigger points and transmission directions of price hikes are different. In 2017, price hikes were across all categories, with the combined demand for smart phones and mining machines, coupled with the active withdrawal of domestic factories from mid-to-low-end production capacity, evolving into a situation of overall shortage and channel speculation. Eventually, it fell back due to overbooking and inventory backlash. The starting point of this round of price hikes was not "general shortage in consumer electronics", but the significant increase in the single-unit usage and value of high-capacity, high-voltage, and high-reliability MLCCs for AI servers, which led to the leading factories filling up their high-end production lines, resulting in a structural shortage pattern of "high-end occupation → mid-to-high-end spillage → some common products also become tight"; price hikes manifested as transmission from AI/automotive-grade high-end to the lower end, rather than spreading out everywhere from the beginning.
The production capacity elasticity varies. For high-end MLCC products, the requirements for dielectric formula, ultra-thin film casting, co-firing process and yield window are higher. Expansion of production capacity not only depends on "facilities + equipment", but is also limited by the delivery time of high-end equipment (about 1-1.5 years) and the certification introduction cycle. This makes it more similar to the characteristics of semiconductors - the shortage is more "sticky".
The downstream demand varies. This time, the customers are more focused on the CSP/AI server supply chain and the 800V high-voltage platform automotive-grade system, rather than just the fluctuations in mobile phone sales. Therefore, the industry's baseline scenario for "how far this can go" often gives a forecast of 2027+ years, which is different from the "fast coming and fast going" channel cycle nature in 2017-2018.
