Hanmi Semiconductor (042700): HBM Bonder Maker’s Q2 Operating Profit Rose 51%, Its Half-Year Fell 11%

Every high-bandwidth memory stack in an AI server was pressed together by a machine. For most of the TC bonders sold into HBM production last year, that machine came from a 45-year-old company in Incheon that almost no English-language investor can name — and whose latest filing tells two completely opposite stories depending on which column of the same table you read.

🏢 What Is Hanmi Semiconductor, and What Does It Actually Sell?

Hanmi Semiconductor is a Korean semiconductor back-end equipment maker whose flagship product, the HBM thermal-compression (TC) bonder, is the machine that presses memory dies into the 12-high and 16-high stacks inside AI accelerators. It sells capital equipment to chipmakers and packaging houses — not chips, not services.

The company (Hanmi Semiconductor Co., Ltd., KOSPI: 042700) was founded on December 24, 1980 and is headquartered at 14 Gajwa-ro 30beon-gil, Seo-gu, Incheon. It listed on the KOSPI on July 22, 2005. Kwak Dong-shin has been CEO since August 2007 and is also the largest shareholder. As of the FY2025 business report the parent company employed 737 people (689 men, 48 women), with average male tenure of nearly ten years; management states roughly 30% of headcount is R&D. (Source: DART half-year report, filed Aug 14, 2026; DART FY2025 business report, filed Mar 12, 2026)

The product line, in the filing’s own order:

Source: DART half-year report, H1 2026, “II. Business Overview” (rcpNo 20260814003665). Product descriptions are the company’s own.

One line in that section is worth flagging because it is a claim about the whole industry, not just about Hanmi: management writes that hybrid bonder development across the memory industry has been delayed by technical difficulties, which is why wider-die TC bonding matters more. That is the company’s assertion, and it happens to be the assertion most favourable to its existing product line. Treat it as a management view, not a settled industry fact.

📊 Where Does Hanmi Semiconductor’s Revenue Actually Come From?

Almost all of it — 89.3% in the first half of 2026 — is equipment sales, and 96.0% of it is billed into Asia. Hanmi reports a single operating segment, so the only splits it discloses are by product type and by region.

Source: DART H1 2026 half-year report, consolidated note 34 “Operating Segments” (rcpNo 20260814003665). Figures are cumulative six-month, not three-month. FX: ₩1,346/$1 as of Sep 4, 2026, applied to every won figure in this article.

A separate table in the business section splits the same revenue a different way — exports vs. domestic, which is not the same basis as the segment note’s regions, because “Asia” includes Korea. On that export/domestic basis the story is a violent reversal:

Source: DART H1 2026 half-year report, “II-4. Sales and Order Status” (rcpNo 20260814003665). Korean row labels reproduced as printed. Note this is a different basis from the segment note above — do not add the two tables together.

Between FY2024 and FY2025 total revenue barely moved (+3.2%), but exports doubled while domestic sales fell by two-thirds. The customer base was swapped out underneath a flat top line. That is the kind of change a single-year revenue chart cannot show you.

Who buys the machines?

The filing names customers but discloses no revenue share for any of them: SK hynix, Micron Technology, ASE, Amkor, JCET, Huatian Technology, TFME, Infineon, ST Micro, PTI, Skyworks and Luxshare abroad; JCET StatsChipPAC Korea, ASE Korea, Amkor Korea, Samsung Electro-Mechanics, LG Innotek, Korea Circuit, SFA Semicon and Signetics domestically. We did not find a major-customer concentration disclosure — the consolidated segment note discloses revenue only by region and by product, with no “customers representing 10% or more” line. That is an absence in the filing, not a confirmation that no such concentration exists.

What the disclosures do show is that every individually announced order in the last year went to one buyer:

Sources: DART single supply-contract disclosures 2025-11-17, 2026-01-14, 2026-06-08. Won values converted at ₩1,346/$1; the filings state ₩1,567,500,000, ₩9,650,000,000 and ₩44,200,000,000 respectively. Note the first is VAT-inclusive and the other two are not — the disclosures say so.

The half-year report also carries a formal order-status table, and it is worth describing exactly as printed, because the empty cells carry information. The table has one row — the June 8 GRIFFIN order, order value ₩44,200,000,000 — and the columns for 기납품액 (delivered to date) and 수주잔고 (order backlog) are left blank. Hanmi therefore publishes no backlog figure at all. In section III-8 of the same filing (“Other financial matters — 3) Order contract status”) it explains why: “For some order contracts we do not disclose or publicise the contract at the customer’s request,” with mandatory disclosure applying only above 5% of prior-year revenue. We do not know Hanmi’s order backlog, and neither does anyone reading only the filings. (Source: DART H1 2026 half-year report, “II-4-3) Order Status” and the internal-control section, rcpNo 20260814003665)

🗓️ How Did Hanmi Semiconductor Get Here?

Hanmi spent four decades as a back-end tool maker for conventional packaging, and was re-rated in three years by one product: the HBM TC bonder. The company’s own history table in DART is short, so the timeline below is assembled from dated primary filings.

Sources: DART FY2025 business report; H1 2026 half-year report; Other Management Matters disclosure, Aug 10, 2026; Tangible Asset Acquisition Decision, Aug 18, 2026. Plant 8 cost ₩56,000,000,000; Plant 7 ₩57,824,000,000.

Two of those rows post-date the half-year numbers entirely. Plant 8 and HANMI USA were both decided in August 2026 — after the June 30 balance-sheet date — and neither appears in any figure elsewhere in this article. Combined, Plants 7 and 8 represent about $84.6m of committed spending against $604.3m of total assets at the end of FY2025.

🌍 Is Hanmi Semiconductor Actually a Global Leader in TC Bonders?

The company reports a 71.2% global share of TC bonders for HBM in 2025, citing TechInsights — but that number reaches investors through Hanmi’s own disclosure, not through a published third-party study we could inspect. The distinction matters on a filing this consequential, so here is exactly what was said and where.

Sources: the TechInsights sentence quoted in row 1 appears in the FY2025 business report (“시장조사기관 테크인사이츠가 발표한 글로벌 TC본더 시장에서 71.2% 점유율로 1위를 기록했습니다”) — the word “테크인사이츠” does not appear in the Value-Up filing at all. The 90% HBM3e figure and the 95% target are from the Corporate Value-Up Plan disclosure of Dec 26, 2025, which states the 71.2% without naming a source. MSVP and EMI Shield: H1 2026 half-year report. Translations of the Korean originals are ours.

Three of these five leadership claims carry no number at all. That is normal in Korean filings and it is why the phrasing here is “the company reports” rather than “Hanmi leads.” The 71.2% figure is the strongest of them because it names an outside research house and a year. It is still a company disclosure.

The filing never discusses competitors in its business section — but it does name one, in the contingent-liabilities note. Hanmi and Hanwha Semitech are suing each other over patents, in both directions, in the Seoul Central District Court.

Source: DART H1 2026 half-year report, “XI-2. Contingent Liabilities — 1) Significant litigation”, base date Jun 30, 2026 (rcpNo 20260814003665). Korean column headers and status reproduced as printed; the “소송 내용” column is our translation of the filing’s one-line description. Amounts are the claim values in the filing’s own unit (천원 = thousand won), converted at ₩1,346/$1. The claim values as printed are 2,500,000 and 1,200,000 (단위: 천원). The consolidated note 32 separately states the group is defendant in one pending suit with a claim value of ₩1,200 million — consistent with the second row.

These are two separate actions, not one action reported twice: different case numbers, opposite plaintiffs, different claim amounts and filing dates eight months apart. The amounts are small next to a $16.29bn market capitalisation — combined they are under $3m — so the money is not the point. What matters is that a company whose investment case rests on a reported 71.2% tool share is in live, two-way patent litigation with a competitor, and that this appears nowhere in its business description. Readers who want the wider frame of who supplies what into Korean memory should start with our chip supercycle overview and the SK hynix business breakdown, since SK hynix is Hanmi’s largest disclosed customer.

📈 Is the Business Getting Better or Worse Right Now?

Both, and the answer depends entirely on whether you read the three-month column or the cumulative column of the same table. This is the single most important thing to understand about Hanmi’s 2026 filings, so here is the whole table with both columns labelled.

Sources: DART Q1 2026 report, H1 2026 half-year report, H1 2025 half-year report, all consolidated. Won figures: Q2 2026 revenue ₩251,163,039,356 and operating profit ₩130,346,798,331 (three-month column); H1 2026 ₩302,065,115,926 and ₩138,803,055,438 (cumulative column). Converted at ₩1,346/$1.

Q2 2026 was the strongest quarter of the ten from Q1 2024 through Q2 2026 on all three measures — revenue, operating profit and operating margin.

Derived from DART quarterly, half-year and annual consolidated filings. Where a filing prints only a cumulative column, the quarter is that cumulative figure less the prior cumulative — e.g. Q4 2025 operating profit = FY2025 ₩251,390,142,142 less nine-month ₩223,762,328,162. Shown in won because the whole row is one currency; at ₩1,346/$1 the Q2 2026 column is $186.6m and $96.8m.

It was also, on its own, 83.2% of half-year revenue and 93.9% of half-year operating profit — because Q1 2026 revenue had fallen 65.5% year-on-year and operating profit 87.9%.

Zoom out one more level and the volatility is structural, not a 2026 accident:

Source: DART consolidated financial statements, FY2024 and FY2025 business reports (rcpNo 20250313001171, 20260312001230). All figures are full-year, converted at ₩1,346/$1.

Look at FY2023: revenue more than halved, operating profit fell 69%, and net profit was the highest of the four years. That is not a rounding artefact. It is the subject of the next section but one.

🔍 Why Did Cash Fall $139m While Operating Profit Hit a Ten-Quarter High?

Because the biggest quarter Hanmi has reported since 2024 has not been collected yet: trade receivables rose from $46.3m to $191.9m over the half — roughly the whole of Q2’s revenue — and consolidated operating cash flow came in at negative $42.1m.

Source: DART H1 2026 half-year report, consolidated statement of financial position and statement of cash flows, “제47기 반기 2026.01.01 부터 2026.06.30 까지” (rcpNo 20260814003665). Won: operating −₩56,605,135,149; investing −₩62,898,968,927; financing −₩76,750,406,539; cash ₩276,235,377,529 → ₩88,575,409,916. Converted at ₩1,346/$1.

Three things are worth separating here, because they pull in different directions.

1. The receivables build is what the payment terms predict. The filing sets out Hanmi’s standard export terms: 30% by bank transfer 30 days after the purchase order, 60% 30 days before delivery, 10% 30 days after inspection. A quarter that ships $186.6m of tools in its final weeks will therefore end the period with a large receivable that collects in the following quarter. Days-sales-outstanding on the half-year revenue base works out at about 155 days, against roughly 39 days on the FY2025 year-end balance — a big move, but from an unusually clean starting point.

2. It is not new, and it is much bigger. H1 2025 operating cash flow was also negative, at −$5.3m. The 2026 half deepened that by roughly eight times.

3. Financing and investing did the rest. Of the $57.0m financing outflow, $56.4m was the FY2025 dividend paid in the period. Of the $46.7m investing outflow, $37.2m went into one financial asset — which is the next section.

🚀 Why Does a Bonder Maker Own SpaceX Stock?

Because Hanmi has been running a securities book alongside its equipment business for years — and in the first half of 2026 it sold the last of a stake in a rival Korean toolmaker and put ₩50.0bn ($37.2m) into SpaceX. The disclosure sits in note 6 of the consolidated half-year accounts, in Korean, with no English equivalent we could find.

Source: DART H1 2026 half-year report, consolidated notes, financial-instruments disclosure (rcpNo 20260814003665). SpaceX: cost ₩50,007,997 thousand, fair value ₩56,784,877 thousand, valuation gain ₩6,777 million recognised in the half — a 13.6% mark-up. Stake column is the percentage printed in the filing.

Hanmi is not the first Korean listed company to route money into Elon Musk’s rocket business through an unlisted-equity line; we wrote about a Korean brokerage doing the same thing at far greater scale in this piece on a $499m SpaceX position. What makes Hanmi’s version worth reading is what it replaced.

The HPSP arc, and why net profit has been a bad guide to this business

Until FY2022 Hanmi held 10.21% of HPSP — a KOSDAQ-listed semiconductor equipment maker — as an equity-method associate carried at $33.1m. During FY2023 Hanmi lost significant influence over HPSP, and IFRS required it to reclassify the holding to fair value through profit or loss. On reclassification it booked a valuation gain of ₩207,081 million ($153.8m). The stake was then sold down over three years and eliminated entirely in the first half of 2026.

Sources: DART FY2023 business report, note 29, and FY2025 business report, note 28 (which prints FY2024 as the comparative). FY2023 components: FVTPL valuation gain ₩207,081,243k, associate disposal gain ₩71,411,764k, FVTPL disposal gain ₩20,775,314k. FY2024: FVTPL valuation loss ₩70,757,540k, FVTPL disposal loss ₩17,254,079k. FY2025: FVTPL valuation gain ₩13,444,913k, FVTPL disposal gain ₩7,536,191k, associate disposal gain ₩850,664k. Converted at ₩1,346/$1.

Read the net-profit column on its own and Hanmi’s best year was 2023 — the year revenue halved. Read the operating column and 2023 was the trough. A minority holding in another company’s stock moved reported profit by more than $220m up, then $65m down, then $16m up, while the equipment business did something entirely different. This is the same earnings-quality problem we unpacked at a much larger Korean chipmaker in why non-operating income matched operating profit at SK hynix: in Korean filings, the line to trust is operating profit, and the line that generates headlines is often net.

The SpaceX position is small by comparison — $42.2m of fair value against $554.9m of equity — but it is the same mechanism. It is a management estimate of the fair value of an unlisted US company, in a half-year report that was reviewed rather than audited, and it flows through profit and loss every period. It had already added $5.0m to the first half’s pre-tax profit before a single machine was shipped.

One further detail is worth reading carefully, because it is the sort of thing only the Korean note contains. The filing’s fair-value hierarchy table places the SpaceX line at Level 1 — the tier reserved for unadjusted quoted prices in active markets — while the two much smaller holdings sit below it. The accompanying valuation-technique note describes only Levels 2 and 3 (market approach and cost approach respectively), so no valuation technique is given for the SpaceX line at all.

Source: DART H1 2026 half-year report, consolidated note 6.3 “Fair value measurement” (rcpNo 20260814003665); the separate-company note prints identical figures. The row as printed at 당반기말 is 56,784,877 / 899,400 / 1,000, totalling 57,685,277; at 전기말 it is 11,032,857 / 1,047,000 / 1,000, totalling 12,080,857. The security names in italics are ours, matched to the amounts in the same filing’s other-investments note — the hierarchy table itself lists amounts only. The Dec 31, 2025 row is the check on the column order: ₩11,032,857 thousand is the KOSDAQ-listed HPSP stake, which can only be Level 1.

💰 What Do You Actually Pay for Hanmi Semiconductor Today?

About 38 times trailing sales and 102 times trailing earnings, for a company with $428m of annual revenue — the most expensive multiple in the Korean semiconductor equipment group by a wide margin.

Market data (KRX only, no DART figures in this note): KRX Open API settled closes — price, shares outstanding and market capitalisation all taken from the same Sep 4, 2026 trading date, so they reconcile by construction. The 2025 year-end close of ₩127,400 is the settled Dec 30, 2025 close from the same source, and is the base for the year-to-date figure. The ₩426,000 intraday print (May 12, 2026) is from KRX daily data; that session closed at ₩377,500. Dividend yield shown here is ₩800 ÷ ₩230,000 at the Sep 4 close.

Filing data: net profit, share counts and the FY2025 dividend per share come from the FY2025 business report. Price-range basis: both ends of the range are closing prices from the monthly price tables the filings publish, each footnoted “최고, 최저가는 종가 기준입니다” (highs and lows are on a closing basis) — the low is September 2025 in the FY2025 business report and the high is May 2026 in the H1 2026 half-year report. The 0.40% figure that appears in the February 2026 dividend filing is a different measure — the 시가배당률, computed on the one-week average price before the board date — which is why it differs from the 0.35% above. Converted at ₩1,346/$1.

How does that compare with Korean peers?

Hanmi earns four to six times the operating margin of three of the five Korean peers below, and trades between 1.4 and 6.1 times their price-to-sales multiples — but it is not the most profitable name in its own peer group. The table deliberately compares on operating profit rather than net, for the reason set out in the previous section.

Sources: DART FY2025 annual financial statements for each issuer; KRX Open API settled closes of Sep 4, 2026 for market capitalisation. Basis matters: HPSP files separate (별도, OFS) statements and the rest are consolidated (연결, CFS) — the two are not interchangeable and are labelled here rather than blended. Converted at ₩1,346/$1. P/S = market cap ÷ FY2025 revenue.

One number in that table earns a second look: HPSP — the company Hanmi used to own a piece of — runs a higher operating margin than Hanmi on separate-company accounts, at a quarter of the market capitalisation.

🏷️ Which “Hanmi” Ticker Are You Actually Buying?

Only 042700 is Hanmi Semiconductor; three other KOSPI-listed companies also trade under a “Hanmi” name and none of them make chip equipment. In Korea, “Hanmi” most often refers to the pharmaceutical group, so a search on the name alone will surface the wrong company more often than the right one.

Sources: DART company registry (corp code lookup) for names, industry codes and incorporation dates; KRX Open API settled closes of Sep 4, 2026 for market capitalisation. Converted at ₩1,346/$1.

Hanmi Semiconductor has no preference-share class — the FY2025 share-count filing shows common stock only. We found no US-listed ADR for Hanmi Semiconductor, which means a foreign investor has to buy the Seoul line: see how to buy Korean stocks as a foreign investor, and how won prices convert to dollars if you are working from a US brokerage screen.

One more distinction that matters for anyone reading the related-party note: Hanmi Semiconductor’s H1 2026 filing lists exactly three related parties — Kwakshin Holdings (an equity-method associate, 40.00% owned at Jun 30, 2026, carried at $6.7m, and described in the filing as a wholesale and retail business), Hanmi International and Doya International. None of the other listed “Hanmi” companies appears anywhere in it.

🌏 Is Foreign Money Buying or Selling Hanmi Semiconductor?

Foreigners have been net buyers in every window we measured; domestic institutions flip sign depending on the window, and that reversal is the story. Reporting a single window here would produce a confident and wrong narrative in either direction.

Source: KRX-derived daily investor flows in our own database, through the settled session of Sep 4, 2026. Flows arrive as share counts; dollar figures are each day’s net shares multiplied by that day’s close and summed, then converted at ₩1,346/$1 — an approximation, not execution prices. The 55-session window is the full history we hold for this ticker, not a round number. Rows do not sum to zero because other investor categories are omitted, and two sessions in the 55-session window (Jun 18 and Jun 19, 2026) carry no retail figure, so the retail column for that window covers 53 sessions. The foreign and institutional columns are complete across all three windows.

Domestic institutions bought roughly $33m of Hanmi over the last twenty sessions and sold roughly $922m over the last fifty-five. Both statements are true, and a report using only one of them would be describing a different stock. Foreign investors, by contrast, were consistent buyers: 2,982,546 shares over 55 sessions is 3.13% of shares issued, against a current foreign holding of 8.04% (Naver Finance, as of Sep 4, 2026). For the mechanics of why these flows matter so much on the KRX, see how foreign and institutional flows move Korean stocks, and our running weekly series at the Foreign Flow Watch hub.

🏛️ Who Controls Hanmi Semiconductor, and What Do Minority Holders Get?

The founding family holds 55.79% directly, with no holding-company layer and no cross-shareholding — an unusually simple control structure by Korean standards. CEO Kwak Dong-shin owns 33.59% personally, and seven relatives hold the rest of the block.

Source: DART H1 2026 half-year report, “Largest shareholder and specially related persons’ shareholdings”, base date Jun 30, 2026 (rcpNo 20260814003665). Sub-totals for sisters and sons are our sums of the individually listed lines; the 계 row is the filing’s own. The National Pension Service line reads 5,175,752 shares at Jun 30, 2026 (5,438,995 at Dec 31, 2025). Every figure in this section is on the same Jun 30, 2026 base date; where a Dec 31, 2025 comparative is used it is labelled and comes from the FY2025 business report.

The only movement in the family block during the half was the CEO adding 80,948 shares to his own holding — a controlling shareholder buying, not selling. The National Pension Service is the only other holder above 5%, at 5,175,752 shares (5.43%) on the same June 30 base date, down from 5,438,995 shares (5.71%) at December 31, 2025.

The float did not grow. The number of people holding it doubled.

Hanmi’s small-shareholder count went from 280,731 to 611,216 in six months — up 2.2 times — while the shares those holders own actually fell. Both figures come from the same standard table in consecutive filings.

Sources: FY2025 business report (base date Dec 31, 2025) and H1 2026 half-year report (base date Jun 30, 2026), “소액주주 현황”. The filings print 280,731 holders and 38,521,819 shares at Dec 31, 2025, and 611,216 holders and 37,496,860 shares at Jun 30, 2026. Average holding is our arithmetic: shares held ÷ holders. 소액주주 is a defined term in Korean filings for holders below 1% of issued shares, not a synonym for retail.

The same float was redistributed among 2.2 times as many people, at an average of 61 shares each — about $10,400 at the September 4 close. That is what a retail crowd arriving looks like in the filings rather than in the flow data, and it sits alongside the flow table above, where retail were net sellers in all three windows. The two are not in conflict: many more people now own a slightly smaller pile, which is what happens when existing holders trim into a rally that keeps drawing new entrants.

What comes back to shareholders

Dividend columns: DART dividend disclosures (alotMatter) in the FY2024 and FY2025 business reports, and the Cash/Stock Dividend Decision of Feb 11, 2026 (DPS ₩800, total ₩75,882,401,600, market-price yield 0.4%).
Cancellation column — a different filing: the FY2025 business report, consolidated statement of changes in equity, line 자기주식 소각 (treasury-share cancellation): ₩130,280,293 thousand for FY2025 and ₩72,602,193 thousand for FY2024. Converted at ₩1,346/$1.

Shares issued fell from 97,339,302 at end-FY2023 to 95,312,200 at end-FY2025 — a 2.08% reduction, funded by $150.7m of cancellations across two years. That is real Value-Up behaviour rather than an announcement, and Hanmi has filed a Corporate Value-Up plan three times in seven months (December 2025, March 2026, July 2026).

Two details in those filings deserve foreign investors’ attention, and one of them we could not fully resolve.

First, the dividend was sized to hit a tax rule. The February 2026 board disclosure states in its own notes that the total was resolved “taking into consideration the satisfaction of the requirements for the special taxation treatment of dividend income from high-dividend companies.” The March and July Value-Up filings then state that Hanmi qualifies under Article 104-27 of Korea’s Restriction of Special Taxation Act, on a 35.5% payout ratio and an 11.1% increase over the FY2024 dividend — just above the 10% increase the filings say the provision requires. Both of those filings carry a standing caveat in their own preamble: whether a company is a “high-dividend company” under Article 104-27 “is determined by the company itself.” That is a self-assessment, not a ruling. The payout was not a policy; it was a threshold.

Second, the July 2026 Value-Up filing states that Hanmi carried out a “비과세 배당” — a non-taxable dividend — for FY2025. The FY2025 statement of changes in equity does show ₩14,926,330 thousand ($11.1m) transferred from capital surplus into retained earnings during FY2025, against nil in FY2024, and the filings quote the Commercial Act provision allowing a capital-reserve reduction where reserves exceed 1.5× paid-in capital. But we could not establish from the filings how much of the FY2025 dividend was treated as a non-taxable return of capital, nor whether that treatment reaches non-resident holders. The word 비과세 (non-taxable) does not appear at all in the FY2025 business report or the half-year report, and the only reference to reducing capital reserves in either document is the boilerplate recital of the Commercial Act power — not a resolution to use it. Korea’s high-dividend-company relief is a domestic income-tax provision; a non-resident’s outcome is governed by withholding and treaty rules instead. Our guide to Korean dividend withholding tax covers what actually gets deducted at source. Do not assume a Korean tax label transfers to your account.

On related-party dealing, the current period is clean: total related-party transactions in H1 2026 were ₩300 thousand of interest income. The comparative period is more interesting — in H1 2025 Hanmi sold part of its stake in associate Kwakshin Holdings to Kwakshin Holdings’ own largest shareholder, receiving ₩3,191,278 thousand for assets carried at ₩2,286,215 thousand and booking a ₩850,664 thousand gain. The filing describes the buyer only as “Kwakshin Holdings’ largest shareholder,” and does not name that person. Separately, the same report’s officer schedule lists Kwak Dong-shin — Hanmi’s CEO and chairman — as concurrently serving as CEO and chairman of Kwakshin Holdings. The filings state both facts; they do not state that the two are the same person, and neither do we. Hanmi’s stake in the associate fell from 49.00% at the end of FY2023 to 40.00% as a result of the sale. The transaction is disclosed, small, and above book value; it is also exactly the kind of item that never appears in English-language coverage.

📚 Lingo Check

🎯 Why It Matters for K-Export Stars

Hanmi Semiconductor is the cleanest example we have found of why reading the Korean filing beats reading the translation. Everything in this article came out of documents that exist only in Korean: the three-month column that says the quarter grew 51%, the cumulative column beside it that says the half shrank 11%, the cash-flow statement that turns the best quarter in two years into negative operating cash flow, the note that reveals a $37.2m SpaceX position, and the board minute that admits the dividend was sized to clear a tax threshold.

None of that is hidden. It is filed, dated and public. It is simply in a language most of the people pricing this $16.29bn company do not read. That is the arbitrage this site exists to close — and it applies just as much to the equipment layer of Korea’s AI memory complex as it does to the memory makers themselves, or to the Korea Discount that hangs over all of them.

This company is one of ten in our K-Semiconductor Filing Map — a supply-chain map where every cell is marked by what the filing itself supports, and each one links to the receipt number it came from.

Disclaimer: This article is for informational and educational purposes only and does not constitute investment advice, a recommendation, or an offer to buy or sell any security. All figures are drawn from official filings (DART) and exchange data (KRX) as of the dates stated inline; prices and market data change. Do your own research and consult a licensed adviser before investing.