Samsung Electro-Mechanics Valuation: The Silent Backbone of the AI Boom

Every AI server that ships needs thousands of tiny ceramic capacitors nobody ever sees, plus the substrates that carry the chips everyone talks about. Samsung Electro-Mechanics makes both. It also makes something the AI story leaves out entirely — smartphone camera modules, which are still nearly a third of its revenue. This is what the company’s own Korean-language filings say, as opposed to what the AI narrative says.

🏢 What Is Samsung Electro-Mechanics, and What Does It Actually Sell?

Samsung Electro-Mechanics (KRX: 009150) is a Korean electronic-components maker, 23.69%-owned by Samsung Electronics, that sells passive components, semiconductor package substrates and camera modules — three businesses, not one AI story.

It was founded on August 8, 1973 and listed on the Korea Exchange in February 1979 — the annual report’s own wording is “당사는 1973년 8월 8일에 설립되었으며, 1979년 2월에 한국거래소에 주식을 상장하였습니다” (“the Company was established on August 8, 1973, and listed its shares on the Korea Exchange in February 1979”). The precise listing date recorded elsewhere in the same filing is February 27, 1979. Headquarters is at 150 Maeyeong-ro, Yeongtong-gu, Suwon, Gyeonggi Province. The CEO is Chang Duck-hyun (장덕현), appointed at the March 16, 2022 annual meeting and re-elected on March 19, 2025. The board has seven directors — three executive, four outside. (Source: DART FY2025 annual report, rcpNo 20260310003071, filed Mar 10, 2026.)

Domestic plants are in Suwon (headquarters, R&D centre, marketing), Sejong and Busan. Overseas manufacturing sits in China (Tianjin and Gaoxin), the Philippines and Vietnam, with an R&D subsidiary in India and sales entities across the Americas, Europe and Asia. There were 14 consolidated subsidiaries at the end of 2025, none of them separately listed. For a neutral company overview outside our own reporting, see Samsung Electro-Mechanics on Wikipedia.

The three product lines:

• Components — passive components: MLCCs (multi-layer ceramic capacitors), power inductors, tantalum capacitors, chip resistors. An AI server board consumes them in the tens of thousands, at far higher voltage and temperature grades than a phone.
• Package Solutions — semiconductor package substrates, of which the high-end product is FC-BGA: the precision board that connects a processor to the outside world. The bigger and hotter the chip, the harder the substrate is to build.
• Optical Solutions — camera modules. This is the business the AI framing usually omits, and it is the second-largest of the three.

📊 Where Does the Revenue Actually Come From?

Camera modules are 31.68% of Samsung Electro-Mechanics’ H1 2026 revenue — larger than the substrate business that carries the AI story, and the original version of this article did not mention them at all.

The FY2025 annual report states the split in one sentence: “2025년 매출액은 전년동기 대비 9.9% 증가한 11조 3,145억원으로 각 사업부문별 매출액 및 매출비중은 컴포넌트 사업부문 5조 1,985억원 (45.95%), 패키지솔루션 사업부문 2조 3,018억원 (20.34%), 광학솔루션 사업부문 3조 8,142억원 (33.71%)입니다.” (“2025 revenue rose 9.9% year on year to ₩11,314.5bn; by segment, Components ₩5,198.5bn (45.95%), Package Solutions ₩2,301.8bn (20.34%), Optical Solutions ₩3,814.2bn (33.71%).”)

FY2025 = full year to Dec 31, 2025, consolidated (연결). H1 2026 = the six months to June 30, 2026, consolidated — not annualised, and not comparable line-for-line with the FY column. Sources: DART FY2025 annual report rcpNo 20260310003071 (filed Mar 10, 2026) and DART half-year report rcpNo 20260814003805 (filed Aug 14, 2026). Segment percentages are the filings’ own; USD converted from the filings’ won figures at ₩1,345.99/$1, the settled close of Sep 4, 2026, applied uniformly for comparability — it is not the exchange rate that prevailed in each period.

Two things move in the table. Package Solutions gained 2.11 percentage points of mix between FY2025 and H1 2026; Optical Solutions lost 2.03. That is the AI rotation actually happening, and it is a good deal smaller than the headline version of the story. Components — the MLCC business — has not moved at all: 45.95% then, 45.87% now.

🗓️ How Did Samsung Electro-Mechanics Get Here?

Samsung Electro-Mechanics spent the first half of the 2020s exiting businesses rather than entering them — three separate shutdowns or disposals are recorded in the FY2025 annual report’s five-year history section.

Sources: DART FY2025 annual report rcpNo 20260310003071 §I.2 (five-year history — the filing covers only the most recent five fiscal years, so earlier events are outside its scope); Value-Up disclosure rcpNo 20260319801372; the four supply-contract filings listed later in this article. Won amounts are the filings’ own; USD converted at ₩1,345.99/$1 (Sep 4, 2026) for comparability, not at the rate prevailing on each event date.

Those two suspensions matter for anyone comparing multi-year figures. Because of the discontinued operations, the FY2022 and FY2023 columns in SEMCO’s own R&D tables were restated for comparability. The FY2024 annual report says so explicitly of both years — “제51기, 제50기는 중단사업 발생에 따라 비교표시 목적으로 재작성되었습니다” (“the 51st and 50th fiscal years were restated for comparative presentation following discontinued operations”) (Source: DART rcpNo 20250311001190). Any three-year or five-year growth rate you calculate from stale sources may be measuring a different company.

🌍 Is It Actually a Global Leader in Anything?

Samsung Electro-Mechanics reports a 23% world MLCC share and a 15% world BGA substrate share — and the substrate number fell from 16% over the previous two years.

Source: DART FY2025 annual report, rcpNo 20260310003071, §II.7. The filing labels the MLCC row “시장점유율은 당사의 추정치이므로 조사기관별로 시장점유율이 상이할 수 있습니다” (“market share is the Company’s own estimate, and may differ between research houses”) and the BGA row “시장 점유율은 외부조사 기관인 Prismark의 세계시장점유율 자료(금액기준)를 활용하였습니다” (“market share uses world market share data from the external research firm Prismark, on a value basis”). We are quoting the company’s disclosure; we have not independently verified either number.

Read that table carefully, because it is the opposite of what a bottleneck narrative predicts. If SEMCO were the scarce supplier in a substrate shortage, its share of a growing market should hold or rise. Prismark’s value-based number says it slipped a point in 2025. The honest reading is that SEMCO is a strong number-two or number-three in both categories, not a monopolist — and share data alone cannot tell you whether the slip was capacity, mix or price. We do not know, and the filing does not say.

📈 Is the Business Getting Better or Worse?

It is getting better, and the pace picked up sharply in the second quarter of 2026: Q2 operating profit of $327.2m was 106.7% above the same quarter a year earlier, at a 12.74% operating margin — the highest of the six quarters below.

Consolidated (연결) figures from the DART FY2025 annual report, rcpNo 20260310003071, filed Mar 10, 2026. Note the row label: the net-profit line is profit attributable to owners of the parent (₩706.1bn in FY2025), which is the basis on which SEMCO calculates its own EPS and payout ratio. Total net profit including non-controlling interests was higher — ₩731.0bn ($543.1m) in FY2025. USD converted at ₩1,345.99/$1 (Sep 4, 2026), applied uniformly.

Every column is a three-month period, not a cumulative one. Q1 and Q2 2026 are from the DART quarterly report rcpNo 20260515002842 (filed May 15, 2026) and half-year report rcpNo 20260814003805 (filed Aug 14, 2026). Q4 2025 is not filed as a standalone quarter in Korea: it is derived here as FY2025 minus the nine months to Sep 30, 2025 (rcpNo 20251114002848) — a calculation, not a disclosed figure. USD converted at ₩1,345.99/$1 (Sep 4, 2026).

The Q1 2026 figures in the original version of this article were correct: revenue of ₩3,209.1bn was up 17.2% year on year and 10.6% on the derived Q4 2025 quarter, and operating profit of ₩280.6bn was up 39.9%. It was also, at the time, the first quarter in which SEMCO had passed ₩3tn of revenue — the previous best was the derived Q4 2025 at ₩2,902.1bn.

What the original could not know, because it was written on July 2 and the half-year report landed on August 14, is that Q2 2026 immediately beat it: ₩3,457.2bn of revenue and ₩440.4bn of operating profit. “The first ₩3tn quarter” stopped being the headline six weeks after publication. On a half-year basis, revenue of $4.95bn was up 20.7% and operating profit of $535.6m was up 74.3% against the same six months of 2025.

🔍 Does Demand for SEMCO’s Substrates Really Exceed Its Capacity?

No — and this is the correction that matters most. SEMCO’s own annual report prints an average utilisation rate of 70% for its substrate lines in 2025, meaning roughly three-tenths of that capacity sat idle. The original version of this article claimed demand exceeded capacity “by more than 50%”. That claim was wrong when it was published, and the document contradicting it had been on DART for nearly four months.

Korean annual and half-year reports carry a section — 생산실적 및 가동률 — that most English-language coverage never reads, because it exists only in Korean and only in the filing. It prints capacity, actual output, and the resulting average utilisation rate for each segment. The filing’s own footnote defines the calculation: “가동률은 생산능력 대비 생산실적기준으로 산출하였습니다” (“the utilisation rate is calculated as actual output against production capacity”).

Utilisation rates exactly as printed by the filings — we have not recalculated them. FY2025 covers Jan 1 – Dec 31, 2025 across 294 operating days (DART rcpNo 20260310003071); H1 2026 covers Jan 1 – Jun 30, 2026 across 145 operating days (DART rcpNo 20260814003805). The underlying capacity and output quantities are on different bases — one annual, one half-year — so the quantities are not comparable across the two columns even though the resulting percentages are.

So the substrate business did tighten, and hard: from 70% to 89% in six months. That is a genuine and important change, disclosed by the company, and no English-language wire reported it because it lives in a Korean production table. But 89% is not “demand exceeds capacity by more than 50%”. A company running at 89% still has headroom, and it certainly is not turning away half its order book.

The second half of the original claim — that analysts expected the shortage to persist through 2027 — had no source we could find and none we could verify. We have removed it rather than attribute it to an unnamed consensus.

The evidence that actually supports pricing power

It is in the same filings, two sections earlier, and it is stronger than the claim it replaces. The half-year report states: “MLCC의 평균판매가격은 전년 대비 13.9% 상승하였으며, 반도체패키지기판의 평균판매가격은 전년 대비 18.7% 상승하였습니다. 카메라모듈의 평균판매가격도 전년 대비 2.4% 상승하였습니다.” (“The average selling price of MLCC rose 13.9% year on year; the average selling price of semiconductor package substrates rose 18.7% year on year; the average selling price of camera modules also rose 2.4% year on year.”)

All six rows are the filings’ own year-on-year percentages, quoted as printed. FY2025 column: DART rcpNo 20260310003071, §II.2 and §II.3, comparing full-year 2025 with full-year 2024. H1 2026 column: DART rcpNo 20260814003805, same sections, comparing the six months to Jun 30, 2026 with the same six months of 2025. Coincidence warning: substrate ASP in the H1 2026 column and CCL/PPG input cost in the FY2025 column are both 18.7%. They are unrelated figures from different periods and different tables.

Now the two years read as one story. In 2025, SEMCO’s substrate input cost rose 18.7% while it could only push selling prices up 4.1% — it was absorbing raw-material inflation. In the first half of 2026, input cost inflation barely eased (17.4%) but selling prices went up 18.7%. That is the moment the pricing power arrives, and it lines up exactly with the utilisation jump from 70% to 89% and with the Q2 operating margin of 12.74%.

Note also which way the camera-module line moved: ASP growth collapsed from +14.0% to +2.4% while its input cost turned slightly negative. The AI segments got price; the phone segment did not.

📝 What Long-Term Contracts Has SEMCO Actually Signed?

Four, worth $2,308.3m combined, all disclosed between May 20 and September 1, 2026 — and every one of them is a Components contract. None is for substrates, and none begins before January 1, 2027.

The original article said “long-term supply deals now being signed” without citing any. They exist, they are specific, and they are the single most concrete thing in SEMCO’s disclosure record this year.

The USD figures above are the filings’ own. Each disclosure states its dollar amount and the exact FX rate it used to produce the won headline — for example, the May 20 filing says “상기 2항의 계약금액은 USD 1,035,400,000에 계약일(2026년 5월 20일)의 최초 고시 매매기준율(1 USD = 1,503.80 KRW)을 적용하여 산출함” (“the contract value in item 2 is calculated by applying USD 1,035,400,000 at the first quoted base rate on the contract date, May 20, 2026, of 1 USD = 1,503.80 KRW”). The four filings used four different rates — 1,503.80, 1,544.20, 1,475.60 and 1,376.50 — so we quote the dollars each filing printed rather than re-converting anything. The “% of FY2025 revenue” column is also the filings’ own; note that the first contract runs two years while the other three run one, so the percentages are not additive into a single-year revenue share. Counterparties are withheld in all four filings on grounds of business confidentiality (경영상 비밀유지), with disclosure deferred to Dec 31, 2028 for the first and Dec 31, 2027 for the others.

The half-year report closes the loop on timing, and it does so in one sentence: “상기 공급계약은 계약기간의 시작일이 2027년 1월 1일이기 때문에 당반기에 판매ㆍ공급한 내역은 없습니다.” (“Because the start date of the above supply contracts is January 1, 2027, there were no sales or supplies made under them in the current half-year.”) So $2.3bn of contracted demand exists, and exactly none of it is in any reported number yet. That is a real forward datapoint and a real reason not to extrapolate today’s margins.

The silicon-capacitor contract deserves its own line. The original article called silicon capacitors an option on the 2027+ roadmap that earned no meaningful revenue today. That was right, and the contract tells you why with a date: it is the largest of the four at $1,035.4m, it runs for two years, and it starts in 2027. Glass substrates — the other frontier product — appear in the annual report only as a development programme: “Glass 패키지기판 등 미래를 위한 신제품 개발에도 역량을 집중하고 있습니다” (“we are also concentrating capability on developing new products for the future, such as glass package substrates”). There is no glass revenue, no glass capacity figure and no glass contract in any filing we read.

💰 What Do You Actually Pay for SEMCO Today?

At the September 4, 2026 settled close of ₩1,401,000, Samsung Electro-Mechanics trades on roughly 105 times trailing twelve-month earnings and 10.2 times the book value of its controlling equity — after a 38.3% fall from its June 19, 2026 peak.

Prices and market capitalisation are KRX settled closes for Sep 4, 2026, the last completed trading day before this rewrite — not intraday snapshots. FX ₩1,345.99/$1, the settled rate for the same date (weekend quote rows discarded). Earnings inputs from DART: FY2025 basic EPS ₩9,345 (rcpNo 20260310003071); H1 2025 cumulative basic EPS ₩3,486 and H1 2026 cumulative basic EPS ₩7,476 (rcpNo 20260814003805). The FY2025 dividend per share is the last one declared; SEMCO pays no quarterly or interim dividend, so no 2026 dividend has been set.

The trailing-twelve-month EPS of ₩13,335 appears in no filing. It is our own arithmetic on the three disclosed EPS figures cited immediately above — full-year 2025, less the first half of 2025, plus the first half of 2026 — and we show it only because no twelve-month EPS covering the second quarter of 2026 has been published.

Two of those rows need saying plainly. First, the dividend yield. SEMCO’s own FY2025 filing records a common-share dividend yield of 0.9%, calculated against the share price prevailing around the December 2025 record date. At the September 4, 2026 price the same ₩2,350 dividend yields 0.17%. Nothing about the payout changed; the price did.

Second, the multiple. Trailing earnings have roughly doubled — H1 2026 operating profit was 74.3% above H1 2025 — and the stock is still on 105 times trailing earnings after falling more than a third from its peak. Whatever else is true, the market is not being asked to pay for this year’s earnings. It is being asked to pay for 2027 and 2028, which is also when the $2.3bn of contracted supply begins.

This is one level down the supply chain from the memory names we covered in Korea’s chip supercycle and in our deep dive on what SK hynix actually makes and who buys it. It sits alongside the equipment and materials names we mapped more recently — Hanmi Semiconductor, HPSP and Isu Petasys, whose own filing showed half-year operating cash flow turning negative even as revenue grew. The component layer carries the cycle’s operating leverage in both directions.

🏷️ Which Samsung Electro-Mechanics Ticker Are You Actually Buying?

There are two listed lines — common stock 009150 and preferred stock 009155 — and the preferred traded at ₩501,000 on September 4, 2026, a 64.2% discount to the common, while receiving a slightly larger dividend per share.

Share counts as of Dec 31, 2025 and unchanged since at least Dec 31, 2023, par value ₩5,000 for both lines (DART rcpNo 20260310003071, §I.3). Closes are KRX settled prices for Sep 4, 2026. Dividends per share are the FY2025 declared amounts from the same filing. Of the issued shares, 2,000,000 common (2.68%) and 53,430 preferred are held in treasury and receive no dividend, leaving 72,693,696 common and 2,853,554 preferred outstanding.

SEMCO is a straightforward listing in one respect that matters to foreign buyers: there is no holding company, no separately listed affiliate carrying a similar name, and no US ADR. If you want exposure you buy 009150 or 009155 on the Korea Exchange. The preferred is the cheaper line on both price and yield, and the trade-off is the usual Korean one — no vote, and materially thinner liquidity.

🏛️ Who Controls Samsung Electro-Mechanics, and What Do Minority Holders Get?

Samsung Electronics owns 23.69% and has been the largest shareholder without interruption for at least five fiscal years; the National Pension Service holds 11.00%, and small shareholders — 195,857 of them — hold 52.97%.

Register basis December 31, 2025, common shares only, as disclosed in both the FY2025 annual report (rcpNo 20260310003071) and the H1 2026 half-year report (rcpNo 20260814003805) — the filings warn that holdings may have changed since the register date. Percentages are against 74,693,696 voting shares. Rows do not sum to 100% because these are the disclosed categories, not an exhaustive breakdown.

The ownership chain above SEMCO runs Samsung Electro-Mechanics ← Samsung Electronics ← Samsung Life Insurance (Samsung Electronics’ largest shareholder) ← Samsung C&T, which holds 19.34% of Samsung Life. SEMCO was designated part of a cross-shareholding-restricted business group on May 1, 2026. We cover how these chains price in our work on SK Hynix’s Nasdaq listing.

Shareholder returns

SEMCO filed a Corporate Value-Up plan on March 19, 2026 (rcpNo 20260319801372), Korea’s voluntary programme for laying out capital-return policy. The operative commitment is one line: “배당성향 20% 이상으로 주주환원 정책 지속 유지” (“maintain the shareholder-return policy with a payout ratio of 20% or more”). The rest of the plan is a business-portfolio statement about AI, mobility and humanoid applications, and it is explicitly labelled forward-looking information.

Source: DART FY2025 annual report, rcpNo 20260310003071, §III.6, which prints the dividend totals in millions of won (₩87,022m, ₩136,128m and ₩177,679m). The payout ratios and yields are that filing’s own printed values and have not been recalculated here. It defines the payout ratio as total cash dividend divided by consolidated net profit attributable to owners of the parent. Dividend totals converted to USD at ₩1,345.99/$1 (Sep 4, 2026) for comparability.

A second, independent filing corroborates the FY2025 figures to the won: the Value-Up disclosure (Source: DART rcpNo 20260319801372, filed Mar 19, 2026) prints the same 25.2% payout ratio, an FY2025 dividend of ₩177,678,715,200 against an FY2024 dividend of ₩136,127,727,700, and states the increase as 30.5%.

Three observations a foreign investor should not have to dig for. SEMCO has paid a dividend in each of the 20 fiscal years from 2006 through 2025, and has never paid a quarterly or interim one — the entire distribution arrives once a year on a December 31 record date, with the amount fixed in March. Second, the payout ratio has cleared the stated 20% floor in all three years shown, and FY2025 exceeded it. Third, and less comfortably: SEMCO holds 2,000,000 treasury common shares, and we found no disclosure of any cancellation. Treasury stock that is bought but not cancelled does not reduce the share count and can be re-issued; in Korea this is a live governance distinction, not a technicality.

Two regulatory items are on the record and are small but disclosed: a written warning from the Financial Supervisory Service dated March 11, 2024 for breaching large-holding and ownership reporting obligations, and a ₩3.2m fine from the Ministry of Employment and Labour dated October 21, 2025 for failing to post material safety data sheets, which SEMCO paid voluntarily.

🏗️ Where Is the Money Going?

Capital expenditure in the first half of 2026 was $626.5m — 89.6% above the same six months of 2025, and already 70.7% of everything SEMCO spent on property, plant and equipment in the whole of 2025.

H1 columns are cumulative six-month cash flows (Jan 1 – Jun 30), which is how the Korean semi-annual cash-flow statement is presented — not three-month figures. Sources: DART rcpNo 20260814003805 (H1 2025 and H1 2026 columns of the consolidated statement of cash flows) and rcpNo 20260310003071 (FY2025). The FY2025 cash balance is the December 31, 2025 position. USD converted at ₩1,345.99/$1 (Sep 4, 2026).

R&D corroborates the same direction, and the original article’s unsourced “36% R&D increase since 2022” turns out to have been correct. The two endpoints sit in two different filings, which is why it needs two citations.

The FY2022 base of ₩577,140m is printed in the FY2024 annual report (Source: DART rcpNo 20250311001190, filed Mar 11, 2025, §II.6). The FY2023, FY2024 and FY2025 figures of ₩555,756m, ₩666,310m and ₩786,951m are printed in the FY2025 annual report (Source: DART rcpNo 20260310003071, §II.6).

The four-year change is therefore +36.4%, which is our arithmetic on those two filings rather than a figure either one prints. Both filings note that the FY2022 and FY2023 columns were restated for comparability following discontinued operations, so the two ends of the comparison are on the same basis.

📚 Lingo Check

🎯 Why It Matters for K-Export Stars

The claim that killed the first version of this article — “demand exceeds capacity by more than 50%” — is the kind of number that circulates in English-language AI coverage without a document behind it. The document that refutes it was sitting on DART, in Korean, in a production table, four months before we published. That is the whole thesis of this site in one example: the gap is not analytical sophistication, it is that almost nobody reading in English opens the filing.

The replacement facts are better than the claim they replace. A utilisation rate that went from 70% to 89% in six months, substrate selling prices that flipped from +4.1% to +18.7%, and $2.3bn of contracted supply that starts on a specific date in 2027 are all more useful to an investor than an unsourced shortage percentage — and all three came out of the same filings, in the same sections, that the original never opened.

Disclaimer: This article is for informational and educational purposes only and does not constitute investment advice or a recommendation to buy or sell any security. Figures are point-in-time and were accurate as of the dates stated; prices and valuations change. Always do your own research and consult a licensed financial professional before investing.