Samsung Electro-Mechanics (SEMCO, KRX 009150) crossed KRW 3 trillion in quarterly revenue for the first time. 1Q26 revenue KRW 3.21T, operating profit KRW 280.6B. Package substrate revenue grew +45% year over year, becoming the lead actor of the rerating. The stock rose from KRW 830,000 in late April to KRW 1,024,000 in mid-May — a +23% move. Sell-side target prices jumped from KRW 1.0M to KRW 1.5M within weeks. Yet few investors can clearly state what SEMCO actually makes, why it is an “AI beneficiary,” and what each of its three divisions does. This article dissects SEMCO at the division level.
Key takeaways
- SEMCO is not a chip maker. It makes the power-stability components (MLCC) and the substrates that connect AI chips to motherboards (FC-BGA) that AI silicon needs to function inside a server.
- Three divisions: Component (MLCC, 46% of revenue, 67% of profit), Package Solution (FC-BGA, 20% of revenue, 15% of profit), Optics (camera modules, 34% of revenue, 18% of profit).
- Core shift: previously a “smartphone parts maker.” Today, the rerating thesis is AI-server / xEV high-end MLCC + AI-accelerator FC-BGA in structural supply shortage.
- 1Q26 results: revenue KRW 3.21T (+17% YoY), operating profit KRW 280.6B (+40% YoY). Stripping a one-off severance charge of KRW 71.4B, underlying operating profit was KRW 352.0B.
- The division that matters most: Component (MLCC) defends earnings. Package Solution (FC-BGA) drives the share price. Optics is the optionality bucket.
- At KRW 1.02M today: expensive on 2026 numbers, defensible only if the AI-substrate cycle keeps running through 2027–2028. A patient approach — waiting for next prints — is more reasonable than chasing the move.
1. Start here — why is SEMCO an “AI beneficiary”?
1.1 The invisible infrastructure of AI silicon
In an earlier piece on Samsung Electronics, I called HBM “the workbench of AI silicon.” But even the best workbench (memory) and the best computer (GPU) won’t work if the power supply is unstable (the system crashes) or if the chip isn’t properly wired to the motherboard (data flow breaks).
What SEMCO makes is exactly this kind of “invisible infrastructure.”
What SEMCO parts do inside an AI server:
1. MLCC — the tiny power dam
→ Stores electricity for an instant and releases it smoothly
→ AI chips spike in power draw under compute load;
MLCCs absorb those swings
→ A single AI server uses tens of thousands of MLCCs
2. FC-BGA — the chip's foothold
→ A high-density circuit substrate that connects AI chips
(GPUs, CPUs) to the motherboard
→ Carries both power delivery and signal between chip and board
→ As AI chips get larger and more complex,
substrates get bigger and stack more layers
3. Camera modules — the eyes
→ Cameras for phones, cars, robots
→ Not directly AI-driven, but a real option as the franchise
expands into automotive and robotics
1.2 Why is supply tight right now?
In the Samsung Electronics piece, I noted that HBM is “absorbing fab capacity.” A similar phenomenon is unfolding in MLCC and FC-BGA.
MLCC:
→ AI servers draw a lot of power and swing hard
→ MLCC content per server has jumped
→ At the same time, xEV demand is also exploding
→ Supply can't be added quickly — high-voltage, high-temperature,
high-reliability parts take long qualification cycles
→ Prices are starting to rise
FC-BGA:
→ Bigger AI chips need bigger substrates
→ Bigger substrates → more layers → lower yields
(the percentage of substrates produced without defects falls)
→ Larger panels also fit fewer units per processing line
→ Greenfield substrate fabs take 2–3 years to build
→ Demand is rising now; meaningful new supply is a 2028+ event
→ From now through 2027 is the "supply-short" window
One-line summary: SEMCO isn’t an AI chipmaker; it makes the power and connectivity parts AI chips cannot run without — and those parts are in shortage today.
2. Dissecting the three divisions
2.1 SEMCO’s structure
| Division | Key products | 2025 revenue | Mix | 2025 profit | Profit mix | Margin |
|---|---|---|---|---|---|---|
| Component | MLCC, inductors, chip resistors | KRW 5.20T | 46% | KRW 609.4B | 67% | 11.7% |
| Package Solution | FC-BGA package substrates | KRW 2.30T | 20% | KRW 135.2B | 15% | 5.9% |
| Optics | Camera modules | KRW 3.81T | 34% | KRW 168.7B | 18% | 4.4% |
| Total | KRW 11.31T | 100% | KRW 913.3B | 100% | 8.1% |
Role-split inside SEMCO:
Component (MLCC) = defends profit (67% of operating profit)
Package Solution (FCBGA) = moves the share price (+45% YoY growth)
Optics (cameras) = builds top-line scale + future optionality
3. Component (MLCC) — the profit moat
3.1 What is an MLCC?
MLCC = Multi-Layer Ceramic Capacitor. The name is technical, but the job is simple — it stores electricity for an instant and supplies it smoothly to nearby chips.
Analogy: a water tank in a plumbing system
If tap water surges and dips, the flow is unstable.
A tank stores some water and releases it evenly.
MLCC does the same with electricity:
If current surges and dips, the chip misbehaves.
The MLCC briefly stores charge and releases it smoothly.
Size: smaller than a grain of rice (0.1mm–3mm)
Count: \~1,000 in a smartphone, tens of thousands in an AI server
3.2 Why does SEMCO MLCC matter?
MLCC market structure:
#1 Murata (Japan) \~33% share
#2 Samsung Electro-Mechanics \~23%
#3 Taiyo Yuden (Japan)
#4 Yageo (Taiwan)
→ The top four control most of the market
→ High entry barrier: hundreds of micrometer-thin ceramic layers
stacked uniformly, fired at high temperature without defects
→ SEMCO holds up to 600-layer stacking technology
Why AI-server MLCC is especially hard:
→ Higher voltage, capacitance, and thermal tolerance than phone parts
→ Long qualification cycles; field failure can crash a whole server
→ The list of qualified suppliers is narrow
→ That narrowness translates into pricing power and margin
3.3 1Q26 print and outlook
1Q26 Component revenue was KRW 1.41T (+16% YoY), with utilization at 93% (vs. 70% Package and 66% Optics — the highest in the company).
Management said high-end MLCC demand for AI servers and data centers remains strong into 2Q. Both MLCC shipments and average selling price (ASP) are expected to rise quarter on quarter.
The most important change: selected MLCC products are starting to see price increases. Murata and Taiyo Yuden reported book-to-bill ratios of 1.36 and 1.31 — the highest in five years (post-2021). “BB > 1” simply means orders exceed shipments; the gap is now wide enough to talk about real undersupply.
4. Package Solution (FC-BGA) — the body of the rerating
4.1 What is FC-BGA?
FC-BGA = Flip Chip Ball Grid Array — a high-density circuit substrate that connects AI chips (GPUs, CPUs) to the motherboard.
Analogy: the chip's foothold
AI chips are small and dense, with tens of billions of transistors.
You can't solder them straight onto the server motherboard.
You need an intermediate "foothold" that delivers power and signal
between chip and board. That foothold is the FC-BGA substrate.
Why it's hard:
→ Bigger AI chips → bigger substrates
(server substrates are >4× the area of consumer PC substrates,
with 20+ layers)
→ Bigger substrates warp → chips lose contact
→ More layers must align precisely → yield drops sharply
→ Fine traces must be etched at 5–10 micrometers
(about 1/10th the width of a human hair)
SEMCO became the first Korean company to mass-produce server FC-BGA,
in October 2022. In 2025, it publicly announced a data-center
substrate supply partnership with AMD.
4.2 Why this is “the body of the rerating”
The reason SEMCO went from KRW 830,000 to KRW 1,024,000 is Package Solution.
1Q26 Package Solution:
Revenue: KRW 725B (+45% YoY, +12% QoQ)
Growth drivers: high-end FC-BGA for AI accelerators, server CPUs,
networking
Why the market got excited:
1. Customers want more than current capacity
→ demand > supply → pricing power sits with SEMCO
2. New big-tech customer awarded an AI-data-center networking
substrate program
→ more customers = bigger future revenue
3. Long-term agreement (LTA) discussions are underway
→ not a one-quarter spike — structural demand → multiple expansion
4. Management signaled 2026 capex could double versus 2025
→ ramping investment = a confidence signal that demand is real
4.3 What only SEMCO can do
Typical substrate maker: makes FC-BGA only
SEMCO: makes FC-BGA AND MLCC
Why this matters:
AI servers reduce power loss by placing power-delivery components
directly under the chip. That means stacking — or embedding — MLCCs
on or inside the substrate.
Because SEMCO makes both, it can develop "MLCC-in-substrate"
integration internally.
Murata (MLCC-only) and Ibiden (substrate-only) can't replicate this
in-house.
→ This dual capability is potentially SEMCO's most durable long-run moat.
4.4 What is not yet proven
Package Solution’s 2025 operating margin was just 5.9%. Revenue grew +13% YoY, but profit fell -14%. Why? Raw-material costs (CCL, prepreg) rose +19% while average selling prices rose only +4%.
The key question: in 2026, will revenue growth come with margin recovery? If only the top line grows and margins stagnate, the current price is hard to justify. Whether 2Q Package Solution margin lifts above 12% is the single biggest gate for the rerating to continue.
5. Optics (camera modules) — big revenue, not the protagonist
Optics is SEMCO’s smartphone- and automotive-camera-module business. 2025 revenue KRW 3.81T (34% of company), margin 4.4%.
State of play:
→ Revenue is large (a third of the company)
→ Margin is the lowest in the portfolio (4.4%)
→ Sensitive to the smartphone flagship cycle
→ Not the protagonist of the rerating
Future optionality:
→ ADAS cameras for EVs (number per vehicle is rising)
→ In-cabin driver-monitoring cameras
→ Vision cameras for robots and humanoids
→ Management flagged a robotaxi product ramp starting in 2Q
Current market share: \~9% (down from \~11% in 2024)
→ Real technology, but not the rerating thesis
Investor view: Optics adds revenue stability but does not move the stock. MLCC and FC-BGA do. Automotive and robotics cameras are long-dated options — not something to pay for at today’s multiple.
6. Valuation — already expensive, but for a reason
6.1 Where the stock sits
As of May 15, KRW 1,024,000 / share, market cap ~KRW 76.5T. Up +23% from late April.
| Reference | PER | Reading |
|---|---|---|
| Trailing 12 months | ~110x | Optically very expensive |
| 2026E earnings | ~58x | Once profit recovery is reflected, lower |
| 2027E earnings | ~38x | Only makes sense if the AI-substrate cycle runs into 2027 |
6.2 Sell-side target prices have run
Feb : Samsung Sec KRW 450,000 (on 2027 BPS)
Mar : Hana Sec KRW 550,000 (2026 EPS × 37.5x)
Early-Apr: SK Sec KRW 600,000 (global peer PER 40x)
Mid-Apr: Hana Sec KRW 810,000 (2027 EPS × 35x)
Early-May: Hana Sec KRW 1,000,000 (2027 EPS × 40x)
Mid-May: KB Sec KRW 1,400,000
NH IB / SK Sec KRW 1,500,000
→ Target prices more than tripled in roughly three months.
Why so fast? The valuation frame itself shifted. SEMCO used to trade as a “Korean electronic-parts maker.” Today, the sell side benchmarks it against global AI-substrate and passive-component peers — Japan’s Ibiden, Taiwan’s Unimicron, Japan’s Murata. Apply those multiples and targets jump.
6.3 Scenario-based fair value
| Scenario | 2027E EPS | Applied PER | Fair value | vs. spot |
|---|---|---|---|---|
| Bear (MLCC price hike contained, FC-BGA capex weighs) | KRW 22,000 | 33x | KRW 730,000 | -29% |
| Base (AI cycle persists + Package margin recovers) | KRW 26,800 | 43x | KRW 1,150,000 | +12% |
| Bull (MLCC + FC-BGA price hikes together + LTA) | KRW 30,000 | 50x | KRW 1,500,000 | +47% |
Today’s KRW 1.02M offers +12% upside in the base case — and -29% downside if the bear case prints, +47% upside if the bull case prints. At this price, waiting for the next earnings confirmation is more reasonable than chasing the rally.
6.4 The key framing: not “cheap,” but “expensive without continuous earnings upgrades”
Where SEMCO sits today:
"Is it a good company?" → Yes
"Is the stock cheap?" → No
"So I shouldn't buy?" → Not necessarily — conditionally yes
Conditions:
→ 2Q operating profit above KRW 400B (vs. consensus KRW 378.8B)
→ MLCC price hike spreading from distributors to direct customers
→ Package Solution margin moving above 12%
→ LTA or customer pre-investment formalizing
Two of these = the buy thesis works at today's price.
Zero of these = waiting for a pullback is the right call.
7. What to watch from here
7.1 Earnings triggers
| Trigger | Bull line | Bear line | When |
|---|---|---|---|
| 2Q operating profit | above KRW 400B | below KRW 360B | Late July |
| MLCC price hike | Spreads to direct AI customers | Stays at cost pass-through | 2Q |
| FC-BGA utilization | Full utilization in 2H | New-customer ramp delays | 3Q |
| Package Solution margin | 12%+ in 2Q | Below 10% | Late July |
| MLCC channel inventory | ~4 weeks (tight) | 6+ weeks (loosening) | Quarterly |
7.2 Multiple triggers
| Trigger | Why it matters |
|---|---|
| Sustained AI infrastructure capex | The end demand for both FC-BGA and MLCC |
| Customer long-term agreements (LTAs) | Proves “not one-off” — structural demand |
| Murata / Taiyo Yuden BB ratio | Leading indicator of MLCC cycle (5-year high today) |
| Global substrate capacity plans | How long the supply shortage lasts |
| Glass Core substrate JV | A next-generation packaging option (2027+) |
7.3 Risks to keep in view
| Risk | Description |
|---|---|
| Already expensive | 2027 PER ~38x. If earnings estimates roll over, the stock falls hard |
| FC-BGA capex burden | Doubling capex doubles depreciation; revenue must follow or margin compresses |
| Pass-through failure | If CCL, gold, copper inflation can’t be passed on, margins suffer |
| Smartphone weakness | Camera modules and some MLCC SKUs remain phone-cycle exposed |
| AI capex peak | If AI server investment slows, FC-BGA + MLCC demand bases weaken |
8. How this links to other posts
Samsung Electronics piece: "HBM is the workbench of the AI server"
→ SEMCO: "MLCC is the power stabilizer, FC-BGA is the chip's foothold"
Jeju Semiconductor piece: "Commodity DRAM is short because of AI"
→ SEMCO: "Power components and chip substrates are short because of AI"
Robotics value-chain piece: "Part 1 flagged SEMCO ramping a
humanoid camera in 2H"
→ SEMCO: that ramp lives inside the Optics optionality bucket
All three sit on the same logic:
AI capex → shortage of components, materials, substrates →
pricing power for the bottleneck supplier.
9. The one-line bottom line
SEMCO is the invisible infrastructure layer of AI silicon. MLCC stabilizes power; FC-BGA connects the chip to the board. Both are short of supply. MLCC prices have started to rise; FC-BGA demand has overtaken capacity. If this structure runs through 2027, SEMCO gets rerated from “smartphone parts maker” to “AI infrastructure parts platform.”
But at KRW 1.02M, much of that thesis is already in the price. On 2027 earnings, PER is ~38x — “great company” and “great entry price” are different questions. To chase the move from here, the numbers have to confirm: 2Q operating profit above KRW 400B, MLCC price hikes spreading, FC-BGA margins recovering.
One-line: this is not a cheap stock — it is a stock that is only “not expensive” if earnings upgrades keep coming. The pace of revision matters more than the headline news flow.
This article is research and commentary only and is not investment advice. Divisional revenue and profit figures are based on Samsung Electro-Mechanics’ 2025 annual report and 1Q26 earnings release. The 1Q26 operating profit of KRW 280.6B includes a one-off severance charge of KRW 71.4B. Utilization figures (Component 93%, Package 70%, Optics 66%) are 2025 full-year averages. The ~23% MLCC market share figure is the company’s own estimate. The AMD substrate partnership is per SEMCO’s official disclosure. The new big-tech customer name is not publicly disclosed. The valuation scenarios (Base KRW 1.15M, Bull KRW 1.50M, Bear KRW 730K) are the author’s estimates and may be wrong. Sell-side targets (NH / SK KRW 1.5M, KB KRW 1.4M, Hana KRW 1.0M, etc.) reflect each broker’s report and are subject to change. Glass-core JV mass production is a 2027+ plan and not confirmed. MLCC price hikes are at an early stage in select distributor channels; broader pass-through is not yet confirmed. The analysis may be wrong. Data cut-off: May 15, 2026 KST.
Disclaimer: For research and information purposes only. Not investment advice. Names cited are for analytical illustration; readers should perform their own due diligence and consult licensed advisors before any investment decision.