The most numerous component in modern electronics has no logic, no firmware and no brand name printed on it. It is the multilayer ceramic capacitor, or MLCC, and the industry ships it at a scale of more than four trillion units a year, with some tallies closer to five trillion. A premium 5G smartphone carries roughly 1,000 to 1,200 of them. A modern electric vehicle carries somewhere between 8,000 and 15,000. Nearly half of the world's supply of this part runs through a small set of East Asian ceramic specialists, and one company, Murata Manufacturing, holds roughly 40% on its own.
I want to explain why that is, because the answer is less about capital equipment than about materials science and accumulated yield. The MLCC looks like a commodity. It behaves, in supply-chain terms, like a design-locked specialty part. Understanding the difference starts inside the chip.
What an MLCC actually is
An MLCC is hundreds of tiny parallel-plate capacitors stacked on top of one another and wired in parallel inside a single ceramic block. The dielectric between each pair of plates is barium titanate (BaTiO3), a ferroelectric ceramic with a relative permittivity of roughly 2,000 to 10,000, depending on formulation and operating conditions. That high permittivity is what allows meaningful capacitance in a package a fraction of a millimetre across.
Capacitance rises with the number of stacked layers and with the permittivity of the dielectric, and it falls as each ceramic layer gets thicker. That single relationship explains the whole technology race. The industry has never really been competing on outer case size in isolation. It has been competing on layer count and layer thickness: more capacitance in a smaller footprint means printing more layers, each one thinner than the last, while holding sub-micron alignment across a stack that runs from roughly 500 to well over 1,000 layers.
On passives generally, industry tallies put capacitors, resistors and inductors together at somewhere between 70% and more than 80% of total component count on a modern printed circuit board assembly, depending on the source and the product category. The MLCC is the single largest contributor to that count. If you count parts rather than dollars, the board is mostly ceramic.
Four stages, and every one can fail
The manufacturing route has four tightly linked stages. I find it useful to keep them in mind because the failure modes compound instead of averaging out.
Casting and printing
High-purity BaTiO3 nano-powder is mixed with organic binders and solvents into a slurry, which is tape-cast with a doctor blade into ceramic green sheets. These are unfired, flexible films, and for the smallest devices they can run below half a micrometre thick. A fine nickel paste is then screen-printed onto each sheet in the electrode pattern. Silver and palladium pastes are still used for some legacy and high-reliability formulations, but nickel has become the dominant, lower-cost inner electrode across the mainstream market.
Stacking, pressing and dicing
Between 500 and more than 1,000 printed sheets are stacked with sub-micron layer-to-layer alignment. The block is consolidated under warm isostatic pressing at roughly 1,000 bar and diced into individual chip bodies sized to the target case code.
Co-firing
The diced chips are co-fired in a reducing hydrogen and nitrogen atmosphere at 1,200°C to 1,300°C. The reducing atmosphere matters: it sinters the ceramic and the nickel electrodes together without oxidising the nickel. Afterwards the chips get an external copper termination, a nickel barrier layer and a tin plating so they can be soldered to a board.
Here is why I think yield, not any single machine, is the real moat. A green sheet cast unevenly carries its thickness variation straight into the finished dielectric, and it shows up later as a local weak point in breakdown voltage. A print that is misregistered by a fraction of the layer pitch reduces the electrode overlap that sets capacitance, so a chip that looks dimensionally right after dicing can still miss its rated tolerance. And because the ceramic and the metal shrink by different amounts as they sinter, a formulation or stacking error that is invisible in the green state only reveals itself after firing, when the unit is scrap. Every one of the four stages has to be controlled at once.
Layer count is the metric that matters
It is tempting to read MLCC miniaturisation as a shrinking-footprint story: 0603 gives way to 0201, then 01005, then 008004. I think the more accurate reading is that case size is a downstream consequence of a harder problem. You have to print more nickel layers onto thinner barium titanate sheets, keep sub-micron alignment, and co-fire the whole thing without delamination or electrode discontinuity. A manufacturer that cannot reliably print and align a 1,000-layer stack at sub-micron dielectric thickness cannot make a competitive 008004 part, however good its dicing and termination equipment is.
| EIA case code | Dimensions (L x W) | Dielectric layer | Target application |
|---|---|---|---|
| 008004 | 0.25 mm x 0.125 mm | Sub-micron | Ultra-compact 5G RF front-ends, true wireless earbuds |
| 01005 | 0.4 mm x 0.2 mm | Below 1 µm | Mobile application processors, smartwatches |
| 0201 | 0.6 mm x 0.3 mm | 1 µm to 2 µm | General smartphone power decoupling |
| 0603 / 0805 | 1.6 mm x 0.8 mm | Higher-voltage ceramic layers | Automotive EV powertrains, industrial inverters |
The ceiling on case-size miniaturisation has tracked the ceiling on green-sheet and printing precision, not the other way round. That is why a competitor entering at the frontier is not only competing with a product line. It is trying to compress decades of yield learning into a short window against an incumbent that keeps moving the frontier outwards.
Dielectric classes: precision or capacitance, pick one
Not every MLCC uses the same dielectric chemistry, and the choice is not cosmetic. It determines how far capacitance is allowed to drift with temperature and voltage, and therefore what the part can be used for. The EIA system splits dielectrics into two broad classes.
Class 1 dielectrics, typified by C0G/NP0 and built on neodymium titanate rather than pure BaTiO3, hold capacitance drift to roughly plus or minus 30 parts per million per °C across -55°C to +125°C. That is effectively flat. The price is a much lower achievable capacitance density, so these parts go where precision matters more than bulk capacitance: high-Q RF circuits, filters and timing references.
Class 2 dielectrics are X7R, X5R and the automotive-grade X8R. These are doped or high-k formulations of BaTiO3. They trade precision for far higher capacitance density, and the tolerance band widens to around plus or minus 15% over the rated temperature range. That trade explains why Class 2 dominates general-purpose power decoupling and bypass filtering, where the total amount of capacitance matters more than the exact value.
| EIA code | Temperature range | Drift | Class / material | Typical use |
|---|---|---|---|---|
| C0G / NP0 | -55°C to +125°C | about ±30 ppm/°C | Class 1, neodymium titanate | High-Q RF, tuning, timing |
| X7R | -55°C to +125°C | about ±15% | Class 2, doped BaTiO3 | Industrial decoupling, bypass |
| X5R | -55°C to +85°C | about ±15% | Class 2, high-k BaTiO3 | Commercial DC-DC decoupling |
| X8R | -55°C to +150°C | about ±15% | Class 2, automotive-grade | Under-hood automotive ECUs |
Once a designer has chosen a dielectric class, case size, capacitance and voltage rating for a board, that choice is qualified and, for practical purposes, locked in. This detail matters later when we get to sourcing.
Why the demand curve keeps steepening
A capacitor is not counted once per product. It is counted by the thousand. A premium 5G smartphone needs 1,000 to 1,200 MLCCs for power decoupling, noise filtering and signal tuning across its radio, application processor and power-management subsystems. The automotive shift multiplies that by roughly an order of magnitude per vehicle. Estimates for a modern EV run from about 8,000 units, on conservative teardown-based benchmarks, up to 10,000 to 15,000 on Murata's own automotive OEM component audits. A comparable internal-combustion vehicle carries approximately 3,000.
Every additional inverter, onboard charger, battery-management module and infotainment domain that an EV has over an ICE platform is, in practical terms, more MLCC demand. That is a structural tailwind for whoever already runs the largest and highest-yielding lines.
How Murata got to roughly 40%
Murata Manufacturing, headquartered in Japan, is the world's leading passive-component manufacturer and controls roughly 40% of global MLCC share on its own. Global shipments are estimated at more than four trillion units a year, with the market valued at over $14.0 billion annually. Forty per cent of that means Murata alone ships on the order of a trillion or more MLCCs each year.
The book is clear that this did not come from a single breakthrough. It came from the same layer-count and layer-thickness race described above, run continuously for decades. Each step down in case size demanded simultaneous advances in BaTiO3 powder particle-size control, green-sheet casting uniformity, electrode-printing registration and sintering-shrinkage prediction. Each of those is learned largely through cumulative production experience, not through one patentable invention. Scale is itself a barrier: only decades of compounded investment in tape casting, powder chemistry and firing yield make a trillion-unit run economical.
A moat built on tacit process learning is hard to copy for the same reason it is hard to describe. You cannot buy it as equipment, and you cannot reproduce it from a datasheet. You have to run the volume.
TDK and Taiyo Yuden, both Japanese, together with Samsung Electro-Mechanics of South Korea and Yageo of Taiwan, round out the short list of makers that can compete at volume across the full range of case sizes and dielectric classes. None of them individually approaches Murata's share. All five are East Asian, and together they hold the overwhelming majority of global capacity.
| Supplier | Base | Position in the book |
|---|---|---|
| Murata | Japan | Roughly 40% of global MLCC share |
| TDK | Japan | Competes at volume across sizes and classes |
| Taiyo Yuden | Japan | Competes at volume across sizes and classes |
| Samsung Electro-Mechanics | South Korea | Competes at volume across sizes and classes |
| Yageo | Taiwan | Competes at volume across sizes and classes |
What concentration means for a buyer
Bill-of-materials planners often treat MLCCs as the least risky line on the board: cheap, tiny, apparently generic. The concentration data argues the opposite. The parts are needed in the thousands per high-value product, and once a design is qualified around a specific case size, capacitance and dielectric class there is no drop-in substitute. A disruption at any one of the top five suppliers can stall assembly lines that sit far from where the capacitors are made.
Dual-sourcing is heavier than it sounds. Each supplier's dielectric formulation, layer count and sintering profile is proprietary, so a second-source part in the same EIA case size and dielectric class is not automatically interchangeable at the same capacitance and voltage rating. A buyer who switches has to requalify on the actual board, not just match a datasheet number. That cost makes buyers reluctant to diversify even when they understand the risk, and the reluctance reinforces the incumbent's scale advantage. What looks like a spot commodity market behaves more like a small set of long-term, design-locked relationships, with the single-point-of-failure exposure that implies when a plant outage, an export control or a natural disaster hits.
Two tiers below the visible supplier
The exposure runs deeper than the MLCC maker. High-purity BaTiO3 powder synthesis, fine nickel powder production and silver-palladium paste formulation are specialised inputs in their own right. None of these materials is geologically ultra-scarce. The scarce thing is the refined, capacitor-grade powder synthesis, with particle size and purity controlled at the nanometre scale. That processing capability, more than the ore, sets who can supply at scale.
Resistors and inductors are less acute. Thick-film resistor and molded-powder inductor manufacturing is more geographically distributed, and the materials are less exotic. The precision tiers of both families still cluster among a narrow set of specialist suppliers, for the same reason: process know-how compounds with cumulative volume. I cover that side in the companion piece on resistors and inductors.
India: strong on assembly, thin on the ceramic
India's passive-component base has historically leaned on imports from China, Taiwan and Japan, which are exactly the geographies that hold the process know-how above. Under the Scheme for Promotion of Components and Semiconductors (SPECS), Vishay Components India, Syrma SGS and CDIL are named as expanding passive-component production capacity in India. How much of domestic passive demand those expansions will cover is an open data point; the book marks it as evidence still required, pending a clearer MeitY component-survey disclosure.
Dixon Technologies shows the demand side. It reported becoming India's largest smartphone manufacturer by shipments in the quarter ending mid-2025, with over 22% share, and is reported to hold more than 35% of India's LED TV outsourcing market. By the component-count logic above, each of those product lines consumes MLCCs, resistors and inductors in the tens of millions of units annually, sourced predominantly from the East Asian base. Dixon's backward integration, as disclosed, points at camera modules, an optical-transceiver joint venture and a wearables venture, not at MLCC manufacturing. So the assembly scale is real, but the deeper tier is not established: the book does not find confirmed India-based BaTiO3 powder synthesis, which is the layer that most directly explains why Murata and its peers hold their position.
India has some of the highest-volume electronics assemblers in the world and no confirmed capability at the powder-chemistry layer that determines who can make a competitive MLCC.
Reading the numbers with care
A few of the figures in this article deserve a note on how firm they are. The count of MLCCs in an EV is a wide range, 8,000 to 15,000, and the spread is itself informative. The low end comes from conservative teardown-based benchmarks. The high end comes from Murata's own audits of automotive OEM components. Both are credible, and they measure slightly different things. I would quote the range and not pick a midpoint.
Global shipments are similar: the book says more than four trillion units a year, with some tallies nearer five trillion. Murata's roughly 40% share is a share of that supply, and the "nearly half" in the topic framing is a loose way of saying a bit under half. Where I can, I have kept the hedges the book keeps. The same goes for the 70% to more than 80% figure for passives as a share of component count, which depends on the source and the product category.
What a buyer can do about it
The practical implications are less dramatic than the concentration numbers suggest, but they are real.
- Treat MLCCs as a strategic line item, not a rounding error, in any product that uses them by the thousand.
- Qualify a second source early, while the design is still being finalised, because requalification on the actual board is the expensive part.
- Ask suppliers about their own upstream: powder synthesis, nickel powder and paste formulation sit two tiers below the name on the reel.
- Where the circuit allows it, understand whether a Class 1 or Class 2 dielectric is truly required, because that choice narrows the pool of interchangeable parts.
None of this removes the concentration. It changes how much of it a buyer carries into a product's life. The structural point is that a component this small can gate a line that is very large.
Why the moat is hard to attack from below
It is worth asking why the incumbents have not been displaced by a well-funded newcomer. The book's answer is that the advantage is learned, not bought. Every step in miniaturisation demanded simultaneous progress on four things: powder particle-size control, casting uniformity, printing registration and sintering-shrinkage prediction. A newcomer cannot buy that combined knowledge in one piece, and cannot test it without running large volumes at real yield.
There is also a demand-side lock. Once a smartphone or an inverter is designed around a particular qualified part, the buyer's cost of moving is higher than the price difference between suppliers. So scale wins share, share funds the next round of learning, and the requalification cost holds designs in place. That loop is what "decades-compounded process know-how becomes a durable market structure" means in practice. I do not think it is a mystery once you see the loop; it is just slow to build and slow to break.
What to take away
First, the MLCC race is a layer-count and layer-thickness race. Case size is a consequence, not the goal.
Second, dielectric class is a precision-versus-capacitance choice. Class 1 keeps drift near zero for RF and timing; Class 2 accepts around plus or minus 15% for far more capacitance in general decoupling.
Third, Murata's roughly 40% share is a compounded-yield advantage, and it is reinforced by the requalification cost that keeps qualified designs where they are.
Fourth, for anyone planning procurement or industrial policy, an MLCC is a single-region, few-supplier part, and the risk sits two tiers below the visible maker, in powder chemistry. On India specifically, I would treat SPECS-backed capacity as a start and the BaTiO3 gap as the real open question.