Samsung Electro-Mechanics has spent most of 2026 making two kinds of announcements that normally appear in the opposite order.
Customers are committing first. Capacity is following.
On September 1, the company disclosed its largest-ever MLCC supply agreement: KRW 1.07 trillion of AI-server capacitors to be delivered during 2027. Samsung Electro-Mechanics says it now has long-term MLCC supply agreements with more than ten global customers and holds more than 40% of the AI-server MLCC market.
Less than a month later, it announced KRW 6.78 trillion of investment in high-performance semiconductor substrates in Korea and Vietnam. In that case, customers are not merely promising future demand. Samsung Electro-Mechanics says global customers are also supporting the investment financially and committing enough volume to support future utilization.
The two transactions cover different products and different contract structures. Together, however, they reveal the same strategic change: Samsung Electro-Mechanics is increasingly adding expensive, technically difficult capacity after customer demand has already become visible.
That matters more than the headline size of the orders. It changes how capacity risk, customer qualification and market-share gains interact.
The important change is not the order book itself
Electronic-component companies have traditionally lived with an uncomfortable sequence. Demand strengthens, suppliers decide whether the cycle is durable, factories expand, capacity eventually arrives, and the market may have changed by the time the equipment is ready.
AI infrastructure is altering that sequence for the most constrained components.
Hyperscalers and accelerator vendors cannot easily accept a server shipment being delayed because a relatively inexpensive capacitor or package substrate is unavailable. The cost of component redundancy is small compared with the value of an accelerator system and the cost of delaying an AI cluster.
That changes customer behavior. Instead of treating passive components and substrates as interchangeable spot purchases, customers are increasingly reserving future qualified capacity.
Samsung Electro-Mechanics’ contracts are evidence of that transition.
MLCC: the customer commitment comes before the capacity allocation
The company’s AI-server MLCC position is already materially stronger than its position in the broad MLCC market.
Samsung Electro-Mechanics states that it holds more than 40% of the AI-server MLCC segment. The distinction is important. Aggregate MLCC market share includes enormous volumes of commodity capacitors used in smartphones, PCs and consumer electronics. AI-server MLCCs are a narrower market with tougher requirements for capacitance, reliability, temperature, voltage and size.
That makes revenue share increasingly different from unit share.
An MLCC manufacturer does not have to win every commodity socket to gain economic share. It can allocate more manufacturing capacity to higher-value specifications where the number of qualified suppliers is smaller and the selling price per component is higher.
Samsung Electro-Mechanics is doing exactly that. Its September contract covers KRW 1.07 trillion of 2027 AI-server MLCC supply. The company says more than ten global customers have already signed long-term agreements, while additional customers are under discussion.
The contract does not remove manufacturing risk. A June MLCC disclosure explicitly stated that there was no deposit or advance payment attached to that agreement. The customer is securing future product; it is not necessarily financing the factory.
But the LTA changes a different risk: demand visibility.
A manufacturer deciding whether to devote advanced lines, engineering resources and scarce materials to a difficult high-capacitance specification can do so with greater confidence when a portion of future output is contractually committed.
The result can become self-reinforcing. More committed high-end demand justifies more high-end production. Greater production scale improves the supplier’s ability to qualify next-generation products. Successful qualification then creates access to the next round of customer commitments.
Market share should be measured in value, not just pieces
This mechanism changes how Samsung Electro-Mechanics’ potential market-share gains should be evaluated.
It would be too aggressive to assume that the company must take a steadily larger percentage of every MLCC shipped worldwide. Some of the opposite may occur. Japanese and Korean suppliers are directing more resources toward AI, data-center and automotive products, creating room for Taiwanese and Chinese manufacturers to absorb lower-end specifications.
Samsung Electro-Mechanics could therefore gain value share in advanced MLCCs while maintaining, or even relinquishing, some share in less profitable unit categories.
For the economics of the business, that can be the more important outcome.
AI-server demand is concentrated in specifications that are harder to manufacture. Thin dielectric layers, high layer counts, compact form factors and stringent reliability requirements place greater pressure on process control and yield. TrendForce reported in 2026 that capacity additions for several advanced MLCC specifications were being constrained by manufacturing difficulty even as major suppliers expanded production.
Capacity on paper is therefore not the same as qualified capacity.
A competitor can build a factory. Winning an AI-server socket still requires product development, stable yield, customer qualification and confidence that the supplier can deliver consistently through multiple accelerator generations.
The more customers use LTAs to reserve Samsung Electro-Mechanics’ qualified output, the more difficult it becomes to describe MLCC simply as a commodity capacity race.
FC-BGA makes the contracted-capacity mechanism even clearer
The September 29 substrate investment is structurally different from the MLCC agreements.
Samsung Electro-Mechanics announced KRW 4.27 trillion of investment at its Sejong site and KRW 2.51 trillion at its Vietnam subsidiary for high-performance FCBGA capacity. The company said global customers would provide investment funding while also securing mid- to long-term volumes to support future utilization.
This changes both sides of the capacity equation.
Customer funding reduces part of the financial burden of building new equipment. Volume commitments reduce the risk that an expensive production line enters operation without enough orders to absorb its fixed costs.
The manufacturer still carries execution risk. Substrate size is increasing, layer counts are rising, materials remain constrained and yield must be stabilized. Customer funding cannot make those engineering problems disappear.
But the economics of an expansion backed by customer capital and committed utilization are very different from those of a speculative capacity build.
| Mechanism | AI-server MLCC | FC-BGA expansion |
|---|---|---|
| Customer commitment | Long-term supply agreements | Mid- to long-term volume commitments |
| Customer financing | Not inherent to the LTA structure | Customer investment support disclosed |
| Main risk reduced | Demand and allocation uncertainty | Capex and utilization risk |
| Competitive effect | Locks qualified high-end capacity to customers | Enables large capacity additions with stronger demand visibility |
For Samsung Electro-Mechanics, the significance extends beyond substrates. It suggests that large AI customers increasingly view certain components as capacity that needs to be reserved rather than inventory that can be purchased later.
Investment is following the same high-value migration
The company’s longer-term investment program reinforces that direction.
Samsung Electro-Mechanics disclosed in July that it plans approximately KRW 15 trillion of investment in Busan through 2040 to build a core production and R&D base for AI data-center package substrates and high-value MLCCs.
The disclosure does not provide a standalone MLCC investment figure. Treating the entire KRW 15 trillion as MLCC capex would therefore be misleading.
The strategic allocation is still clear. The company is directing manufacturing resources toward the parts of its portfolio where AI raises technical requirements and where qualification matters more than commodity scale.
This is also visible in current operating results. In the second quarter of 2026, Samsung Electro-Mechanics’ component division generated KRW 1.65 trillion of revenue, up 29% from a year earlier. The company attributed part of the improvement to growing MLCC sales into AI servers, networks and other data-center applications.
The revenue mix is moving before all of the announced capacity has arrived.
Why share gains can persist beyond one shortage cycle
Shortages alone do not create durable market share.
A temporary supply squeeze can disappear when competitors add capacity, inventories normalize or customers qualify alternative vendors. The stronger argument for Samsung Electro-Mechanics depends on the sequence of qualification and reinvestment.
First, AI accelerators are moving toward higher power density. That raises demand for high-capacitance and high-reliability MLCCs.
Second, manufacturing those specifications at scale requires tighter process control and higher yields. The addressable supplier set narrows.
Third, customers respond by signing LTAs with suppliers that already have qualified products and a credible roadmap.
Fourth, those contracts give the supplier greater confidence to allocate equipment, engineering and capital to the next generation of products.
That investment improves the probability of winning the next qualification cycle.
The competitive advantage is therefore less about the MLCC itself than about the loop connecting qualification, contracted demand and reinvestment.
Murata remains the constraint on the share-gain thesis
None of this implies an uncontested market.
Murata remains one of the industry’s largest and most technically capable MLCC suppliers. TrendForce’s mid-2026 data showed strong order momentum across Murata, Samsung Electro-Mechanics and Taiyo Yuden rather than at Samsung alone.
Murata is also adding advanced capacity. If its new lines ramp quickly, yield improves and customers deliberately maintain dual sourcing, Samsung Electro-Mechanics’ share gains could be limited even in a strongly growing market.
There is another counterforce. Chinese and Taiwanese manufacturers are gaining experience as leading Japanese and Korean suppliers reduce exposure to some commodity products. Over time, technology acquired in lower-end products can migrate upward.
The defensible part of the Samsung thesis is therefore not permanent scarcity. It is the time required for competing suppliers to create equivalent qualified capacity while accelerator specifications continue to move.
The stronger signal is the contract structure
Samsung Electro-Mechanics used to be easy to describe as a cyclical electronic-component manufacturer. Demand moved, utilization followed, pricing changed and earnings amplified the cycle.
AI infrastructure is introducing a different operating model at the high end.
MLCC customers are reserving future production through LTAs. Substrate customers are going further by helping fund capacity and securing utilization before factories begin mass production.
That does not eliminate cyclicality. Nor does it guarantee that every announced investment will produce the expected return.
But it reduces the amount of capacity being built purely on a forecast.
If that structure persists, Samsung Electro-Mechanics’ competitive position will increasingly depend on how much qualified AI capacity it can repeatedly convert from customer commitment into production—not simply on how many capacitors it can manufacture.
That is the mechanism through which today’s order book could become tomorrow’s market share.
Sources Used
Research Cut-off: 2026-10-01 22:16
Public primary sources
Samsung Electro-Mechanics — AI-server MLCC supply contract, Sep. 1, 2026.
KRW 1.0722 trillion contract for 2027, more than ten global MLCC LTA customers, and company-stated AI-server MLCC market share above 40%. Samsung SEM
Samsung Electro-Mechanics — FCBGA capacity expansion, Sep. 29, 2026.
KRW 4.27 trillion at Sejong plus KRW 2.51 trillion in Vietnam. The company explicitly states that global customer funding and mid- to long-term volume commitments are intended to reduce CAPEX and utilization risk. SEC
Samsung Electro-Mechanics — Q2 2026 results, Jul. 30, 2026.
Q2 revenue of KRW 3.4572 trillion and operating profit of KRW 440.4 billion; component-unit growth was supported by data-center and automotive MLCC sales, while the company confirmed more than ten MLCC LTAs. SEC
Samsung Electro-Mechanics — Mid- to Long-Term Investment Strategy, Jul. 3, 2026.
Approximately KRW 15 trillion through 2040 for the Busan AI data-center package substrate / high-value MLCC production and R&D base. The disclosure does not provide a separate MLCC-only amount. SEC
Samsung Electro-Mechanics — MLCC supply disclosure, Jun. 30, 2026.
KRW 453.99 billion contract for 2027 and, importantly for the mechanism analysis, no deposit or advance payment. Samsung Semiconductor
Samsung Electro-Mechanics — Silicon capacitor contract, May 20, 2026.
KRW 1.557 trillion supply agreement for 2027–2028, evidence that the company’s AI component strategy is broadening beyond conventional MLCCs. SEC
Public secondary / industry sources
TrendForce — Jul. 6, 2026.
Murata, Samsung Electro-Mechanics and Taiyo Yuden all recorded sharply elevated book-to-bill ratios as AI-related high-end MLCC demand increased. TrendForce
TrendForce — Jun. 17, 2026.
Advanced X6S specifications remained difficult to ramp because of manufacturing complexity and yield constraints, making nominal capacity expansion different from effective qualified supply. TrendForce
TrendForce — Jul. 28, 2026.
Shipments of the three leading Japanese/Korean MLCC suppliers reached a five-year monthly high while production resources shifted toward higher-end AI specifications.