MLCC shortages expand
Shortages that began with memory and GPUs are now spreading to multi-layer ceramic capacitors, the small passives used across power delivery on boards. Digitimes reports AI infrastructure buildouts are tightening MLCC supply, which makes these otherwise low-cost parts potential schedule blockers and requalification triggers. That shift means manufacturing teams need to treat passives as yield‑critical items rather than easy substitutions. (digitimes.com)
A multi-layer ceramic capacitor is a tiny charge reservoir made by stacking many ceramic and metal layers, and modern boards scatter them around chips the way a city scatters water towers around neighborhoods. Murata says the stacked structure lets the part store more charge, and its low-inductance versions are used for power-supply decoupling near high-speed integrated circuits. (murata.com 1) (murata.com 2) That “power-supply decoupling” job means smoothing out tiny voltage dips and spikes before a processor notices them. KYOCERA AVX says these ceramic capacitors have to sit physically close to the processor because the fastest current bursts happen right at the chip package. (kyocera-avx.com) Artificial-intelligence servers are unusually hard on that power system because the processors run below 1 volt while their current can jump by tens to hundreds of amperes almost instantly. Samsung Electro-Mechanics says that combination is pushing a shift toward higher-capacitance ceramic capacitors in artificial-intelligence server power designs. (samsungsem.com) That is why a part that used to be treated like cheap background hardware is suddenly turning into a bottleneck. DigiTimes reported on April 10, 2026 that supply tightening, which started with memory chips, is now showing up in multi-layer ceramic capacitors as global companies keep expanding artificial-intelligence infrastructure. (digitimes.com) The volumes are big enough to change factory math. Power Electronics News, citing European Passive Components Institute remarks from 2025, said one high-end artificial-intelligence board can use 10,000 to 20,000 ceramic capacitors, or roughly 10 to 15 times the count in conventional systems. (powerelectronicsnews.com) The squeeze is concentrated at the high end, not across every capacitor on earth. TrendForce said in February 2026 that demand for high-end multi-layer ceramic capacitors tied to Nvidia GB200 and GB300 servers and custom chips from Amazon Web Services and Google had pushed Murata, Samsung Electro-Mechanics, and Taiyo Yuden above 80% utilization. (trendforce.com) The supplier mix also makes substitutions harder than buyers expect. DigiTimes said Japanese and Korean suppliers, especially Murata and Samsung Electro-Mechanics, are the main sources for the high-spec parts going into artificial-intelligence infrastructure, which means a shortage in one narrow grade can block an entire board build. (digitimes.com) Even when a replacement part looks similar on paper, manufacturers often cannot just swap it in the way they might swap a screw. Texas Instruments notes that ceramic capacitors differ in electrical behavior and even mechanical side effects like audible vibration, so changing package, dielectric, or structure can force new validation work on the board. (ti.com) Murata is already adding capacity, which tells you suppliers think this is not a one-quarter blip. A new Murata production building for multi-layer ceramic capacitors in Japan was completed this week with nearly 70,000 square meters of floor area and 10 floors above ground. (evertiq.com) So the story is not that servers are running out of glamorous chips. The story is that racks full of expensive graphics processors can now be delayed by millimeter-scale ceramic parts that cost little individually, appear by the thousands on each board, and cannot always be replaced without redesign and requalification. (digitimes.com) (trendforce.com)
