The component you have almost certainly never thought about is quietly making modern technology possible. A single AI server rack contains up to 450,000 of them. A modern electric vehicle uses between 20,000 and 30,000. Your smartphone carries around 1,200. They cost fractions of a cent each, yet their absence would halt the production of virtually every electronic device on Earth. This is the story of the Multilayer Ceramic Capacitor — the MLCC.
1. What is an MLCC?
A Multilayer Ceramic Capacitor (MLCC) is a surface-mount electronic component that stores and releases electrical charge. It is built by stacking alternating layers of ceramic dielectric material and metal electrodes, co-firing them into a monolithic block, and capping each end with external metal terminations that allow it to be soldered onto a printed circuit board (PCB).
MLCCs belong to the broader family of capacitors — passive devices that do not amplify, switch, or process signals but are absolutely essential to every circuit that does. They serve three primary electrical functions: decoupling (suppressing voltage noise near active components), filtering (blocking or passing specific frequency bands), and bypassing (providing instantaneous local charge reservoirs for ICs whose demand fluctuates faster than a power supply can respond).
They are not glamorous. They ship in plastic tape reels holding thousands of components that look, to the naked eye, like grains of colored sand. The smallest commercially available MLCC — the 008004 package — measures just 0.25 mm × 0.125 mm, making it roughly one-fifth the width of a human hair when viewed from above.
Despite this near-invisibility, MLCCs are among the most consequential components in the electronics supply chain. When they go into shortage — as they dramatically did in 2018 and are doing again in 2026 — production lines from Shenzhen to Stuttgart grind to a halt.
2. The physics of capacitance
To understand why MLCCs are engineered the way they are, you need to understand what capacitance actually is. A capacitor is, at its most fundamental, two conductive plates separated by an insulating material called a dielectric. When voltage is applied, charge accumulates on the plates, creating an electric field through the dielectric. This is stored energy.
Capacitance — measured in Farads — is determined by the following relationship:
C = ε₀ · εᵣ · N · A / d
Where: ε₀ = permittivity of free space (8.854 × 10⁻¹² F/m), εᵣ = relative permittivity (dielectric constant) of the ceramic material, N = number of active electrode layers, A = electrode overlap area, and d = thickness of each dielectric layer. Increasing N, increasing A, increasing εᵣ, and decreasing d all increase capacitance.
This formula is the engineering roadmap for every advancement in MLCC technology over the past five decades. Manufacturers have pushed all four levers simultaneously: developing ceramics with higher εᵣ, stacking ever more layers (N), reducing each dielectric layer to sub-micron thicknesses (reducing d), and increasing electrode area (A) through larger die sizes and improved printing resolution.
Today's most advanced MLCCs achieve dielectric layers as thin as 0.5–1 µm — about 1/100th the diameter of a human hair — and stack over a thousand such layers in a package barely 1 mm long. This is genuinely extraordinary precision manufacturing, comparable in complexity to semiconductor fabrication.
3. Anatomy of an MLCC
An MLCC is structurally simple but materially complex. From the outside in, here are the layers:
External terminations
The silver-gray ends of an MLCC are multi-layer metal coatings. Typically, the innermost layer is silver or copper paste, fired to bond with the ceramic body. Over this sits a nickel barrier layer applied by electroplating, which prevents solder from migrating into the ceramic body during reflow. The outermost finish is tin, which provides the solderability needed for PCB attachment.
The ceramic body (the dielectric)
The bulk of the MLCC is its ceramic dielectric, most commonly based on barium titanate (BaTiO₃) for high-capacitance types or calcium zirconate / modified titanates for precision Class 1 types. The dielectric constant of BaTiO₃ can reach 10,000 or higher with appropriate dopants — compared to about 3–5 for the FR4 glass-epoxy material used in PCBs. This high εᵣ is why ceramic capacitors achieve such remarkable capacitance density.
Internal electrodes
Interleaved with the dielectric layers are thin metal electrodes, typically made of nickel in modern "base metal electrode" (BME) designs. Historically, electrodes used expensive palladium-silver alloys, but the transition to nickel in the 1990s dramatically reduced cost and enabled the mass-market explosion of MLCCs. The electrodes alternate polarity — odd layers connect to one external termination, even layers to the other — creating the parallel-plate capacitor structure described in Section 2.
The protective cover layers
The topmost and bottommost layers of an MLCC are solid ceramic with no electrodes, providing mechanical protection and acting as an insulating buffer between the active stack and the external environment.
4. How MLCCs are made
MLCC manufacturing is one of the most sophisticated high-volume production processes in all of industry. A single Murata factory in Japan produces billions of units per month using a process that requires sub-micron precision at every stage. Here is how it works, step by step:
Ceramic powder — primarily barium titanate with carefully chosen dopants — is milled to a particle size of typically 100–300 nm. This powder is then dispersed in a solvent with organic binders, plasticizers, and dispersants to form a stable ceramic slurry. The chemistry at this stage is a closely guarded trade secret; the exact dopant formulation determines the dielectric's temperature stability, voltage coefficient, and aging characteristics.
The slurry is cast through a precision doctor blade onto a moving polymer carrier film, producing a continuous "green sheet" of unfired ceramic — typically 1–5 µm thick for advanced products. The term "green" here means unfired, not colored. These sheets are dried and then cut or punched into panels for the next stage.
Internal electrode paste — nickel metal mixed with organic binders — is screen printed onto each green sheet in a precise pattern. The electrode is offset slightly from panel to panel so that alternating electrodes connect to opposite ends of the finished component. Print resolution determines how many layers can be stacked and how thin each electrode can be.
Hundreds of printed green sheets are stacked in precise alignment — an operation that requires mechanical registration tolerances of just a few micrometers. The stack is then subjected to isostatic pressing, applying uniform hydraulic pressure from all directions, which bonds the layers into a dense, void-free monolith. This step is critical: any misalignment or delamination creates a defective capacitor.
The compressed block is precision-cut into individual chips using diamond-coated blades or laser cutting. At this stage each "chip" is a raw, unfired block roughly the size of the finished capacitor. The dicing accuracy determines whether electrode layers will properly contact the end terminations after sintering.
The green chips are first heated to ~300–500°C to burn away organic binders, then sintered at 1,200–1,450°C in a controlled-atmosphere kiln. For nickel electrodes, a reducing atmosphere (nitrogen/hydrogen mixture) is required to prevent nickel oxidation — this is one of the key technical challenges that makes high-quality MLCC manufacturing difficult to replicate. The ceramic shrinks by approximately 15–20% during sintering, so dimensional tolerances must account for this predictable but precisely calibrated shrinkage.
Sintered chips are barrel-plated with a silver or copper paste, fired again at ~600–850°C to bond the terminations, then nickel-plated and tin-plated by electroplating. Edge chamfering via tumbling removes sharp corners that could cause cracking during PCB assembly.
Every capacitor is electrically tested — measuring capacitance, dissipation factor (DF), insulation resistance (IR), and often 100% high-voltage testing. Defective parts are rejected. Conforming parts are loaded into tape-and-reel packaging at rates of millions of units per hour for delivery to electronics manufacturers worldwide.
Building a new MLCC kiln facility takes 2–3 years and hundreds of millions of dollars. Lead times for specialized sintering tunnel kilns are 8–12 months. Critically, the process knowledge — especially the precise atmosphere control during sintering and the dopant formulation — is deeply tacit and takes years to master. This is why new entrants and capacity expansions take so long to materialize, and why shortages once begun can persist for years.
5. Dielectric classes explained
Not all MLCCs are the same. The choice of dielectric material has profound implications for performance, and international standards bodies (EIA in the US, IEC internationally) have established a classification system. Understanding this is essential for engineers selecting components.
The three-character code used for Class 2 and 3 ceramics encodes operating temperature range: the first character is the lower temperature limit, the second is the upper limit, and the third is the maximum allowed capacitance change across that range.
Class I — Ultra Stable
Class II — High Capacitance
Class III — Very High Cap
A critical, often-overlooked behavior of Class 2 MLCCs is DC bias derating: the effective capacitance drops significantly as DC voltage across the component increases. A nominally 10 µF X5R capacitor rated at 6.3 V can lose over 50% of its stated capacitance when operating at its rated voltage. Engineers must always consult the DC bias curves and apply generous voltage derating — typically using a part rated at twice the operating voltage.
Automotive-grade MLCCs add further certification requirements. The AEC-Q200 standard mandates additional environmental stress tests (humidity, thermal cycling, mechanical shock and vibration) and requires lifetime qualification at +150°C for 2,000 hours. Qualifying an MLCC to AEC-Q200 adds substantially to cost and lead time — which is partly why automotive-grade parts command significant price premiums over commercial-grade equivalents.
The new X8R, X9R frontier
As electric vehicle powertrain electronics push operating temperatures higher — BEV power modules commonly see 150°C+ junction temperatures — there is growing demand for MLCCs that maintain stable capacitance at these extremes. The emerging X8R (rated to 150°C) and X9R (rated to 200°C) dielectric classes are increasingly available from Tier 1 manufacturers and command heavy pricing premiums due to the more complex material formulations required.
6. Key specifications and parameters
When selecting an MLCC, engineers evaluate a matrix of parameters. Here is a reference guide:
| Parameter | What it measures | Why it matters | Typical range |
|---|---|---|---|
| Capacitance (C) | Charge storage per unit voltage | Core function; determines decoupling efficacy | 100 pF – 100 µF |
| Rated voltage (V) | Max DC voltage across component | Exceeding causes breakdown; always derate Class 2 | 4 V – 10 kV |
| ESR | Equivalent Series Resistance | Determines power loss at high frequency; low ESR is critical for AI server PDN | 1 mΩ – 1 Ω |
| ESL | Equivalent Series Inductance | Limits effectiveness at high frequencies; low ESL needed for GHz decoupling | 0.5 – 5 nH |
| Dissipation Factor (DF) | Ratio of energy lost to energy stored per cycle | Heat generation; important in high-current RF/power circuits | <0.1% (C0G) – 5% (Y5V) |
| Insulation Resistance (IR) | DC resistance across the dielectric | Leakage current; critical for precision and low-power applications | 10 GΩ – 10 TΩ |
| Temperature coefficient / class | Capacitance stability vs. temperature | Sets operating range suitability (see Section 5) | ±30 ppm/°C (C0G) to −82% (Y5V) |
| Case / package size | Physical footprint (EIA code) | PCB real estate, pick-and-place compatibility | 008004 to 2220 (and larger) |
| Capacitance tolerance | Manufacturing accuracy of stated C value | Tighter tolerance = higher cost; RF circuits need ±1–5% | ±1% (C) to ±20% (M) |
Package size nomenclature
MLCC package sizes follow EIA coding: the four-digit code (e.g., 0402, 0603, 1210) represents the length and width in hundredths of an inch. A 0402 is thus 0.04" × 0.02" or approximately 1.0 mm × 0.5 mm. Metric equivalents (EIA 0402 = Metric 1005) are also used. The industry trend has been relentless miniaturization: mass market smartphones and wearables now predominantly use 0201 (0.6 mm × 0.3 mm) and increasingly 01005 (0.4 mm × 0.2 mm) packages.
7. Applications and demand drivers
MLCCs serve in virtually every powered electronic device, but the demand profile has shifted dramatically over the past decade. Consumer electronics remain the volume king, but three structural demand drivers are reshaping the market: AI computing infrastructure, electric vehicles, and 5G telecommunications.
Consumer electronics (the historical bedrock)
A modern flagship smartphone contains 1,200–1,500 MLCCs. A tablet uses approximately 900. A laptop carries around 1,000. These components appear in every circuit: power management ICs, RF front-ends, camera modules, display drivers, audio subsystems, USB controllers, and DRAM decoupling. Consumer electronics accounted for roughly 45% of MLCC demand by volume in recent years, though the segment's pricing power is weak due to intense competition and chronic oversupply of general-purpose parts.
AI servers (the explosive new frontier)
The AI compute boom has fundamentally changed the economics of high-end MLCCs. A conventional enterprise server uses approximately 1,000 MLCCs for power delivery network (PDN) decoupling and signal integrity. But AI training servers are different in kind, not just degree.
"A single Nvidia GB200 NVL72 rack requires approximately 440,000 MLCCs — a quantity 30 times that of a smartphone."
Murata Manufacturing estimates; industry analysis, 2025–2026The reason is the extraordinary power densities involved. Modern AI accelerators can draw thousands of watts at low voltages, requiring hundreds of high-capacitance, ultra-low-ESR MLCCs per device to maintain stable PDN voltage within the microsecond transient windows that matter for compute correctness. These are not commodity parts. They require low ESR (<5 mΩ), high capacitance (≥10 µF) in 0402 or 0603 packages, and must maintain performance at elevated temperatures. Yields on these parts are significantly lower than for general-purpose MLCCs, maintaining a tight supply balance even as overall MLCC shipments grow.
Murata Manufacturing estimates that by 2030, demand for MLCCs from AI servers will be 3.3 times its 2025 level. The hyperscalers collectively committed over $300 billion in AI infrastructure capital expenditure in 2025, and a substantial fraction of that spending translates directly into MLCC demand.
Electric vehicles
An internal combustion vehicle uses roughly 3,000–5,000 MLCCs for body control modules, infotainment, and lighting. A modern battery electric vehicle requires 20,000 to 30,000 MLCCs — a six-to-ten-fold increase — distributed across the powertrain inverter, onboard charger, DC-DC converter, battery management system, ADAS sensors, and the growing electrical architecture of the connected cockpit.
Automotive MLCCs operate under extremely demanding conditions: vibration, thermal cycling between −40°C and +150°C, humidity, and the safety-critical reliability expectations of automotive OEMs. AEC-Q200 qualification is mandatory. High-voltage 800 V BEV architectures (increasingly standard in premium EVs to enable faster charging) require MLCCs rated at 1,000–3,000 V, a segment where only a handful of qualified suppliers exist.
5G telecommunications
5G base stations (particularly millimeter-wave small cells) use more MLCCs per unit than their 4G predecessors, requiring components with excellent high-frequency characteristics for RF filtering and signal routing at frequencies up to 100 GHz. In March 2025, Kyocera-AVX unveiled what it claimed to be the world's first 47 µF MLCC in a 0402 package — tripling capacitance density for space-constrained 5G radio designs. The global 5G infrastructure MLCC market is forecast to reach $1.2 billion by 2027.
8. The vendor landscape
The MLCC industry is a textbook example of an oligopoly rooted in knowledge intensity. Capital requirements for a competitive fab are enormous, but the real barrier is the decades of process know-how embedded in manufacturing teams, kiln recipes, and material formulations. A handful of Japanese and Korean firms dominate the high end; Taiwanese and Chinese players are established in mid-tier and working their way upmarket.
Market share overview
Market share figures are global revenue estimates compiled from multiple analyst sources (Mordor Intelligence, MarketsandMarkets, CECIA); figures vary by source and segment. AI server market share diverges significantly from the overall — see individual profiles.
Tier 1: Japanese and Korean giants
Tier 2: Taiwanese challenger and the US/Japan partnership
Tier 3: Chinese manufacturers — the fast-rising challengers
Chinese MLCC manufacturers have rapidly increased their global market presence. Their collective share surged to approximately 10% of total revenue in the second half of 2024 — a four-percentage-point increase from 2019. Three companies lead this charge:
The strategic significance of Chinese MLCC development is not lost on Japan and Korea. Samsung has responded by accelerating its shift toward premium, high-margin products — AI server and high-voltage automotive MLCCs — that Chinese manufacturers cannot yet replicate. Murata similarly focuses on sub-micron dielectric layer technology, advanced materials, and long-term customer design-in relationships that create switching costs. The competitive dynamics increasingly resemble the memory semiconductor industry, with commodity tiers commoditizing rapidly while high-end tiers remain technically defensible for years.
9. Supply chain and shortage cycles
The MLCC supply chain has been chronically prone to boom-bust cycles, with significant downstream consequences for electronics manufacturing globally. Understanding these cycles — and their structural causes — is essential for procurement professionals and electronics engineers alike.
Building MLCC capacity is slow and capital-intensive. A new tunnel kiln production line requires 8–12 months lead time for equipment procurement alone, plus another 12–18 months of process qualification. Government-backed financing in Japan and Korea accelerates investment for domestic manufacturers, but new entrant capacity takes 3–5 years to become reliably qualified. This structural lag ensures that demand shocks translate into prolonged supply constraints.
Geopolitical risks
The concentration of global MLCC capacity in Japan, South Korea, and Taiwan — three nations at the intersection of US-China strategic competition — creates supply chain risk that has attracted board-level attention at major electronics OEMs. Taiwan's geopolitical exposure is particularly acute given that MLCC production there (primarily Yageo and Walsin Technology) represents meaningful global capacity. The January 2024 Noto earthquake in Japan briefly raised concerns about Murata facility disruptions. Defense and aerospace programs increasingly require domestic or "friend-shored" supply for critical passive components.
10. Where MLCC technology is heading
The MLCC roadmap is driven by five converging pressures: more capacitance in less volume, lower ESR at higher frequencies, thermal stability at higher temperatures, reliability in harsher environments, and integration with active components.
Sub-micron dielectric layers
The frontier of MLCC technology lies in reducing dielectric layer thickness below 0.5 µm while maintaining breakdown voltage and long-term reliability. At these thicknesses, grain boundary effects and defect density become dominant concerns. Murata's proprietary powder synthesis capability — controlling particle size down to approximately 100 nm — is the key enabler, and the capability gap between Murata and its nearest competitors is one of the most significant moats in the electronics components industry.
High-entropy ceramics
Academic and industrial research into high-entropy ceramics — compositions incorporating five or more metal oxides at near-equimolar ratios — has produced remarkable energy storage performance. A BaTiO₃-based high-entropy MLCC published in Science in 2024 achieved recoverable energy density of 20.8 J/cm³ with efficiency above 97%, far exceeding conventional BaTiO₃. The challenge is manufacturing consistency at scale; minor composition errors in production can cause large performance deviations.
Silicon capacitors and thin-film alternatives
Silicon capacitors — capacitors built using semiconductor fabrication techniques — are attracting significant investment, particularly from IDT/Renesas and startups including Murata-backed ventures. Samsung Electro-Mechanics itself launched a silicon capacitor R&D task force in 2022. Silicon capacitors offer lower ESL than MLCCs (because via-based electrical access replaces terminal contacts) and can be integrated directly onto interposers or package substrates adjacent to GPUs and CPUs — a significant PDN advantage at multi-terahertz switching frequencies. They are not MLCCs, but they are competing for the same design slots in next-generation AI compute hardware.
Compositionally graded MLCCs
Research has shown that MLCCs with compositionally graded dielectric layers — where ceramic composition varies systematically across the layer stack — can achieve enhanced dielectric tunability (up to 70%) over wide temperature ranges, enabling adaptive filtering and power conversion circuits that adjust their own characteristics in response to operating conditions. This is an emerging academic concept but points toward future smart passive components.
Integration and packaging
MLCCs are increasingly being embedded directly into PCB substrates or package substrates (as embedded passives), eliminating the solder joint and reducing mounting parasitic inductance. Companies like TDK and Murata offer thin MLCC variants specifically designed for substrate embedding. As chiplet-based AI accelerators mature, co-packaged capacitor technologies — placing decoupling capacitors within the same package as the processor die — are moving from research to product roadmap.
Industry analysts are increasingly drawing a comparison between high-end MLCCs in 2025–2026 and High Bandwidth Memory (HBM) in 2020–2022: a niche, technically demanding component that moves from commodity background noise to strategic bottleneck as a single dominant application — AI training — creates overwhelming concentrated demand. The duration of the current price cycle depends heavily on AI server shipment cadence. But the structural argument — that high-end MLCC capacity cannot be built fast enough to match AI infrastructure investment — appears durable through at least 2028.