Multi-layer ceramic capacitors (MLCCs) are very common passive components in various electronic products. Recently, driven by the continuous growth in demand for AI servers, they have once again become the focus of industry attention.

Samsung Electro-Mechanics stated that the MLCC usage per AI server is 10 to 15 times that of an ordinary server. TrendForce also mentioned that as demand for AI servers and self-developed application-specific chip platforms continues to rise, the supply of high-end MLCCs remains tight, the short-term supply-demand gap is difficult to improve quickly, and product prices have fluctuated accordingly. Taiyo Yuden, a leading Japanese MLCC manufacturer, took the lead in raising prices for small-capacitance consumer-grade and some automotive MLCCs, with an increase range of 6% to 13%. This price adjustment has simultaneously spread to the distribution channel market, prompting Samsung Electro-Mechanics to also evaluate adjusting supplier prices.

In the past, MLCCs were generally regarded as passive components with a high degree of standardization and significant industry cyclical characteristics. However, with the explosion of AI server demand, they are no longer just supporting components in the consumer electronics industry chain, but have also become an important indicator for observing the supply-demand prosperity of the AI industry chain.

What is a Capacitor?

The current in electronic products is not always stable. When chips perform high-speed operations and components quickly switch operating states, voltage and current will experience instantaneous fluctuations. If such fluctuations are not effectively suppressed, they will at best lead to unstable signal transmission, interfere with the normal operation of chips, or even cause damage and scrapping of components.

The core function of a capacitor is to temporarily store electric charge in a circuit. When the voltage rises instantaneously, the capacitor can absorb excess charge; when the voltage drops, it can quickly release the stored charge, thereby maintaining stable circuit voltage. We can metaphorically understand a capacitor as a miniature reservoir in a circuit: it automatically stores water to buffer during sudden water surges, and replenishes water to stabilize pressure when water flow is insufficient.

As early as the 18th century, the Leyden jar, as the first-generation electricity storage device, realized charge storage. The real reason for the large-scale popularization and application of capacitors in modern electronic products was the subsequent various chip capacitors that achieved miniaturization and could be batch-mounted on printed circuit boards. Among many capacitor categories, multi-layer ceramic capacitors (MLCCs) are currently one of the most widely used and industrially important core categories.

What is an MLCC?

MLCC stands for Multi-Layer Ceramic Capacitor. From a structural principle, an MLCC is an integrated solid capacitor made by stacking ceramic dielectric layers and metal electrodes alternately layer by layer and then sintering them at high temperatures. This unique multi-layer stacking structure is the key to enabling MLCCs to achieve sufficient capacitance in an extremely small volume.

Simply put, the core value of MLCCs is to achieve thinner, lighter and miniaturized capacitors that can be integrated on printed circuit boards on a large scale. At present, the smallest MLCC size in the industry has reached 0.16mm × 0.08mm. Its extreme miniaturization is crucial for smartphones and wearable devices with extremely scarce internal space.

MLCCs have many excellent properties:

• Small size, suitable for high-density integrated circuit boards;

• Excellent high-frequency performance, meeting the needs of high-speed signal transmission, power supply voltage stabilization and other scenarios;

• Outstanding reliability and stability, adaptable to various complex working environments such as automotive, industrial control, and servers;

• Compatible with automated surface mount technology, supporting large-scale mass production.

Compared with other mainstream capacitor categories, aluminum electrolytic capacitors can achieve ultra-large capacitance but are bulky and have weak high-frequency performance, making it difficult to adapt to highly integrated new-generation electronic products. Tantalum capacitors have excellent stability and are widely used in high-end electronic equipment, but production costs are high due to material and process limitations. MLCCs achieve an excellent balance among the four core dimensions of size, cost, reliability and high-frequency performance, which is why they have become the most widely used and versatile capacitor category in the modern electronics industry.

Why Do AI Servers Require So Many MLCCs?

MLCCs have been around for decades and have long been classified as low-cost, highly standardized passive electronic components. Unlike core devices such as GPUs, HBMs and advanced packaging that directly determine device computing performance, they are basic standard components for circuit boards of all electronic equipment. With the explosive growth of the AI server industry, MLCCs have been revalued, with core driving factors focusing on two major dimensions: a substantial increase in usage and a comprehensive upgrade of product specifications.

According to survey data from TrendForce, a single NVIDIA GB200 motherboard is equipped with approximately 6,500 MLCCs. The new-generation Rubin platform is expected to see MLCC usage per motherboard climb to 12,000 units due to increased thermal design power consumption and a more complex power management architecture. Data from Samsung Electro-Mechanics shows that the MLCC usage per AI server is 10 to 15 times that of an ordinary server. Murata Manufacturing even predicts that capacitor demand in the AI server sector in 2030 will reach 3.3 times that of 2025, indicating vast industry incremental space.

In addition to the surge in usage, MLCC product specifications have also ushered in a comprehensive upgrade. High-end MLCCs dedicated to AI servers are fundamentally different from standard models used in ordinary consumer electronics. As power consumption of AI servers continues to rise, the equipment power supply system is gradually iterating from the traditional 12V architecture to front-end DC voltage architectures of 48V, 54V and even higher. This requires MLCCs in the power conversion link to have higher voltage resistance, greater capacitance and better high-temperature reliability.

Specification upgrades have also significantly raised the manufacturing barriers for high-end MLCCs. To achieve greater power storage capacity in limited PCB space, manufacturers need to ultra-thin ceramic dielectric layers and metal electrodes while stacking more layers accurately. Samsung Electro-Mechanics disclosed that the stacking layers of high-end high-capacitance MLCCs can reach 600 layers. The thinner the dielectric layer and the more stacking layers, the higher the requirements for raw material uniformity, high-temperature sintering process precision and production yield. Any deviation in any link will directly affect product stability and reliability.

This means that the MLCC demand driven by AI servers is not simply a quantitative increase, but a concentration of market demand towards high-end models with high capacitance, high voltage resistance and high reliability, greatly enhancing the industrial value and strategic position of high-end MLCCs.

In addition, under the high-power working conditions of AI servers, the power supply system not only relies on a large number of MLCCs to stabilize circuit voltage, but also needs power semiconductor devices to complete core tasks such as power conversion and energy consumption regulation. The two work together to ensure efficient and stable server operation.

Major MLCC Manufacturers

Industry statistics show that the world's top five MLCC manufacturers account for 77% of the overall market. Murata and Samsung Electro-Mechanics take up more than half of the high-end market. Domestic manufacturers account for about 10% of global total revenue, while Taiwanese manufacturers and European and American companies share the remaining 13% of the market.

Japanese Manufacturers: Core of Industry Technology

Murata Manufacturing has a solid global leading position, with outstanding advantages in automotive, industrial and miniature MLCCs, and continues to increase investment in AI server PDN supporting products. TDK, Taiyo Yuden, Kyocera AVX and Maruwa Electric have sufficient production capacity in high-voltage, high-frequency and high-reliability segments. Japanese companies took the lead in completing the research and development of ultra-thin dielectric and multi-layer stacking processes, mastered a complete reliability database, and continue to monopolize the supply of high-end specifications for automotive and computing power.

Korean Manufacturers: Leaders in High-End Automotive MLCCs

Samsung Electro-Mechanics has consolidated its leading position in high-end MLCCs for AI servers and automotive applications. Its Q4 2025 financial report showed that performance growth was entirely driven by AI computing infrastructure demand, and it simultaneously expanded its product lines of high-capacitance, high-voltage and automotive-grade MLCCs. Samsung also simultaneously deploys silicon capacitors for high-performance chip packaging and AI server power supply, positioning them as a supplementary solution rather than an alternative to MLCC arrays. South Korea masters both traditional MLCC and advanced silicon capacitor technology routes.

Taiwanese Manufacturers: Carrier for High-End and Low-End Production Capacity Connection

After acquiring Kemet and Pulse Electronics, Yageo Group has achieved a full-category layout of MLCCs, resistors, inductors and magnetic components, covering automotive, industrial and communication markets. Its new high-end MLCC production line in Kaohsiung is a core supplier of X6S capacitors for AI servers. Taiwanese manufacturers such as Walsin Technology have also benefited from the recovery of AI demand, and raised prices of some resistors and MLCCs after capacity utilization increased. Taiwan supplies both high-end special specifications for AI and massive general-purpose models, making it an important choice for diversified procurement.

Chinese Mainland Manufacturers

Representative enterprises include Fenghua Advanced Technology, Guangdong Sanhuan Group, Microcore Electronics, etc. In 2024, they accounted for about 10% of the global total revenue, with growth relying on domestic demand for computers, home appliances and general consumer electronics. This round of price increase cycle has greatly improved the profits and valuations of domestic leading enterprises, and upstream ceramic material manufacturers have also benefited simultaneously. However, product verification and technical comparison data show that domestic manufacturers still lag behind overseas leading brands in core indicators of high-end MLCCs for AI servers and automotive applications.

European and American Manufacturers

Vishay and Kemet (now merged into Yageo) focus on industrial, aerospace and military segments, with complete product lines of tantalum capacitors, polymer capacitors and film capacitors, supporting a small number of MLCC models. The direct supply volume of MLCCs for AI servers is limited, but they have irreplaceable value in diversified procurement of complete machine passive materials and scenarios with special reliability standards.

Why Are MLCCs in Short Supply and Prices Rising?

The recent focus on MLCCs is not a shortage of all categories, but a clear divergence between high-end products and standard products. AI servers, automotive electronics and industrial fields have significantly boosted demand for high-capacitance, high-voltage and high-reliability MLCCs, while standard products have stopped falling and seen some rebounds due to production capacity allocation, inventory control and channel stocking.

First, the supply elasticity of high-end MLCCs is limited. TrendForce pointed out that cloud service providers are accelerating the adoption of application-specific integrated circuits (ASICs), which has rapidly shifted MLCC demand towards small-size, high-capacitance and high-temperature resistant high-end models.

At the same time, MLCCs used in AI servers need to achieve higher capacitance in a limited board area and withstand higher temperatures, higher voltages and more stringent reliability tests. Therefore, high-end MLCCs cannot be supplemented by simply expanding production on ordinary production lines. Material systems, thin-layer stacking processes, sintering control, mechanical strength and customer certification cycles all directly affect actual supply capacity. Even if leading manufacturers such as Murata, Samsung Electro-Mechanics and Taiyo Yuden increase production capacity for relevant specifications, some high-end MLCCs are still difficult to ramp up quickly due to manufacturing difficulty and yield limitations.

Second, the standard product market is also affected by the adjustment of production capacity structure. Against the backdrop of sustained AI demand, major Japanese and Korean manufacturers have shifted part of their production capacity to high value-added products, resulting in a gradual narrowing of supply elasticity for consumer-grade MLCCs quarter by quarter. TrendForce stated that channel distributors in Taiwan and Chinese Mainland have therefore stocked up on standard products in advance, creating a structural divergence between weakening actual ODM orders and increasing channel orders.

In April 2026, Taiyo Yuden took the lead in raising prices of small-capacitance consumer-grade and some automotive-grade MLCCs by 6% to 13%. TrendForce also pointed out that driven by supply chain price increases, the average price decline of overall MLCCs narrowed to only 0.5%, the smallest decline in nearly three years, indicating that manufacturers' bargaining power has rebounded significantly.

Third, rare earth materials exacerbate supply chain uncertainty. Some rare earth elements are used as doping materials in MLCCs to improve capacitance temperature characteristics, service life and reliability. Therefore, after China implemented export controls on rare earths, the high-end MLCC supply chain also faced raw material supply risks.

Therefore, this round of MLCC price increases cannot be simply attributed to "AI servers buying up all MLCCs". The real core change is that AI servers and self-developed application-specific chip platforms concentrate demand on a few high-end specifications, resulting in structural supply shortages in high-end production capacity. At the same time, affected by production capacity allocation and channel stocking, the standard product market has gradually stabilized from the previous downward cycle.