Samsung Electro-Mechanics has signed a KRW 1.0722 trillion (approximately USD 780 million) supply agreement with an unnamed global major corporation for multilayer ceramic capacitors used in AI servers.
The one-year contract, running from 1 January through 31 December 2027, is described by the company as its largest MLCC supply contract to date.
Long-term demand visibility for AI-grade MLCCs
The agreement adds to Samsung Electro-Mechanics’ portfolio of long-term supply agreements with more than ten global customers. The company says it is also pursuing further long-term contracts, reinforcing its strategy to expand the proportion of high-value AI-oriented passive components in its business mix.
Since May 2026, Samsung Electro-Mechanics has announced long-term agreements for AI-server MLCCs and silicon capacitors worth approximately KRW 3.32 trillion in total. For the company, the contracts provide a more predictable demand base at a time when AI data-centre infrastructure is driving accelerated investment in high-performance computing hardware.
Why AI servers need specialised MLCCs
Multilayer ceramic capacitors (MLCCs) provide local decoupling, transient-current support and high-frequency noise suppression across processor, accelerator, memory and voltage-regulator power rails. In AI servers, these functions become significantly more demanding because GPUs, AI accelerators and associated memory subsystems create rapid and high-amplitude current transients.
Samsung Electro-Mechanics highlights 24/7 operation, elevated temperature, higher operating voltages, high capacitance requirements and reliability as key differentiators for AI-server MLCCs. These requirements favour suppliers able to combine high capacitance density with controlled electrical performance and robust manufacturing and quality assurance.
The company states that AI servers can employ more than ten times as many MLCCs as general-purpose servers. This reflects the growing number of power domains and the need for dense capacitor arrays positioned close to high-current devices, where low parasitic resistance and inductance are essential to managing voltage droop during fast load steps.
Strategic implication for the supply chain
The size and duration of the deal underline the shift from spot purchasing toward strategic capacity reservation for premium MLCC categories. Unlike broad commodity MLCC demand, AI-server applications concentrate requirements in high-capacitance, high-reliability and temperature-capable parts, for which qualified supply remains limited.
Samsung Electro-Mechanics says it holds more than 40% share of the AI-server MLCC market. The company cited an estimated expansion of the market from USD 1.4 billion in 2025 to USD 5.8 billion in 2030, equivalent to approximately 34% average annual growth.
Design perspective
For power-integrity engineers, capacitor selection should not be based on nominal capacitance alone. Effective AI-server decoupling requires an impedance strategy that considers DC-bias behaviour, temperature performance, ESR, ESL, mounting location, PCB parasitics and the target frequency range of the voltage-regulator module and processor load transient.
High-CV MLCC arrays located close to GPUs, CPUs and ASICs can address fast transients, while other capacitor technologies may provide complementary mid-frequency and bulk-energy support. The practical objective is to maintain a low and controlled power-distribution-network impedance across the relevant frequency range rather than relying on one capacitor type alone.
Further reading
- Nvidia Vera Rubin: Why One AI Rack Needs So Many More MLCC Capacitors
- Samsung Three Pillars MLCC Strategy for AI Hardware Topology
- Murata Publishes Power Delivery Guide for AI Servers
- Temperature, Bias and Ageing Impact to MLCC Ceramic Capacitors
Source
This article is based on a Samsung Electro-Mechanics press release announcing its AI-server MLCC supply agreement and the company’s stated market, contract and product-positioning information.
