Techwing (089030): Why a 173% Profit Jump Ended in a Loss

Somewhere between a finished memory chip and the tester that decides whether it lives, a machine has to pick the chip up, heat it, hold it still, and sort it. A 610-person company in Hwaseong, south of Seoul, has built those machines since 2002 and now sells them to Micron, SK hynix, Samsung, Intel and Infineon. In the first half of 2026 its operating profit rose 172.7%. It still reported a half-year net loss — even though the second quarter, taken on its own, made money. The reason for both facts is a book of currency contracts, some of them signed in 2021, that we could find described in no English-language source.

Updated September 15, 2026 — pledge terms only: Four of the fourteen share-pledge contracts below reached the end of their disclosed term after this article was written: KB Securities’ 29,198-share contract on 7 September, and the IBK (181,052), KB (393,384) and DB (66,778) contracts on 14 September — together 670,412 shares against ₩13.5bn ($10.0M) of loans. Re-running the contract-by-contract arithmetic at ₩43,400 — the regular-session close of 14 September 2026 on the Korea Exchange 09:00–15:30 tape, which is 7.9% below the ₩47,100 used throughout this article — the count is unchanged at four of fourteen below their stated maintenance ratio, and it is the same four contracts (KB Dasomtech 127%, LS 117%, DB 145%, Korea Investment 158%, against required 140/140/200/180%). Aggregate coverage falls from 270% to 248%. Every other figure in this article is stated as of 4 September 2026 and is unchanged.

Which closing price, and why that question is new. From 14 September 2026 a Korean stock has two end-of-day prices, because KRX opened a continuous 16:00–20:00 after-market that day — we cover the mechanics in what actually changed on 14 September. Techwing’s regular-session close was ₩43,400; its after-market then traded lower, and data feeds that sample after 16:00 show roughly ₩43,100. We use the regular-session close, because that is the price the exchange’s own day-on-day change is computed from. The distinction does not change this analysis: at ₩43,100 the four ratios read 126% (KB), 116% (LS), 144% (DB) and 157% (Korea Investment) and the aggregate 247% — still four of fourteen below their stated ratio. If your screen shows ₩43,100 rather than ₩43,400, this is why.

What we cannot tell you: Techwing’s controlling shareholder has filed no updated large-holding report since 24 August 2026, so we cannot say whether those four contracts were repaid, rolled over or replaced. The end of a pledge term is not itself a disclosable event in Korea, which is why the answer is absent rather than hidden. Read against our corrections policy.

🏢 What Is Techwing, and What Does It Actually Sell?

Techwing builds the handling machines that feed finished semiconductors into test equipment — and, increasingly, the consumable kits those machines wear out. It is not a chipmaker and not a tester maker; it is the automation layer in between, in the back end of the semiconductor process.

Techwing, Inc. (주식회사 테크윙) was founded on July 18, 2002 and listed on KOSDAQ on November 10, 2011. Headquarters is at 37 Dongtan Sandan 6-gil, Hwaseong, Gyeonggi Province, with additional plants in Anseong and Asan. It employed 610 people as of June 30, 2026, with average tenure of six years and seven months. It is run as a co-CEO structure by Na Yoon-sung (나윤성, who is also the largest shareholder) and Jang Nam (장 남), an arrangement in place since March 2022. (Source: DART half-year report 2026.06, company overview and employee sections.)

The filing describes the product line in five blocks. The first is the test handler — memory, non-memory and module/SSD — which moves a device to the tester, supplies the test temperature environment, and sorts pass from fail. The second is burn-in equipment (Burn-In Chamber and Sorter), a reliability stress test. The third is the wafer-level Probe Station. The fourth is the Cube Prober, a High Bandwidth Memory test tool. The fifth — and this is the one that pays the bills more reliably than any of the others — is parts and peripherals, chiefly the Change Over Kit (COK).

The COK matters because of what it is: the filing says it “must necessarily be replaced when the device under test changes.” Every time a customer moves to a new package, the kit gets swapped. Techwing supplies those kits directly, so its installed base of handlers generates a recurring consumable stream. In FY2025, parts were 55.2% of consolidated revenue — more than the equipment itself.

The customer list is named in the filing’s own words: “our products are supplied to leading global semiconductor companies including SK hynix, Samsung Electronics, Micron, Intel and Infineon.”

One structural fact that surprises people who assume “equipment maker” means “factory.” Techwing states plainly that it does not run a conventional manufacturing operation: “we design, develop and sell semiconductor test equipment; after design and development, parts fabrication is outsourced, and we perform final assembly and sale. We are therefore not an industry that manufactures using dedicated production facilities.” Because of that, it declines to disclose production capacity or output at all. Its outsourced-processing spend in the first half was $3.3M against $47.4M of raw-material purchases — the parts are bought, not made.

📊 Where Does Techwing’s Revenue Actually Come From?

Roughly half of it comes from consumable parts rather than machines, and 79.3% of it leaves Korea. The equipment gets the headlines; the change-over kits and boards get the repeat orders.

The half-year report prints a five-period product mix. We reproduce all five, because three of them tell a different story than the most recent two do.

Consolidated revenue mix, all five periods as printed. H1 2026 covers six months; the other four columns are full years. Source: DART half-year report 2026.06, rcpNo 20260814001211, “품목별 매출 비중” table.

Two readings, and they point in opposite directions.

The bullish one is memory. Memory test equipment sat in a tight band — 31.2%, 30.3%, 30.8%, 31.3% — for four consecutive years, and then jumped to 47.4% in the first half of 2026. That is the memory capex cycle arriving in Techwing’s order book, and it is the single biggest change in the table.

The bearish one is non-memory. Techwing’s stated strategy since inception has been to widen out from memory handlers into SoC handlers, SSD, burn-in and wafer inspection. The non-memory line went 16.1% → 3.6% → 12.8% → 5.3% → 0.8%. Read only the last three columns and you see a decline; read all five and you see something worse for the diversification thesis — a line that has never sustained a level for two years running. This company is more levered to memory than its product brochure suggests, and the five-year series is how you find that out.

Exports and the domestic swing

Source: DART half-year report 2026.06, “매출실적” table. Percentages are ours, computed from the won figures in the filing. FX: ₩1,345.99/$1, the USD/KRW close our own price history records for September 4, 2026, the last weekday before drafting, applied to every dollar figure in this article except where a filing prints its own rate.

Hold on to the export number. Techwing collects most of its revenue in dollars, and that is not a footnote in this business — it is the reason its income statement looks the way it does. We come back to it below.

Customer concentration: the table Techwing does publish

Korean filers must disclose their top three customers by revenue, anonymised. Techwing does, in one table covering three periods — which means the letters are internally consistent across the columns even though the names are withheld.

Source: DART half-year report 2026.06, “주요 매출처” table (consolidated basis). The filing prints amounts only — every percentage in this table is ours. The four rows sum exactly to consolidated revenue in each of the three periods (FY2024: 83,480 + 17,511 + 16,436 + 68,080 = 185,507 million won), so the percentages are arithmetic, not estimates. H1 2026 is six months; the other two columns are full years, so the account counts are not directly comparable.

Customer A is the only entity in the top three across all three periods. Everyone else rotates: the FY2024 runners-up (C and E) were gone by FY2025, and the FY2025 runners-up (B and D) were gone by mid-2026. Techwing’s business is one very large relationship plus a churning tail.

The filing does not say who A is, and we are not going to guess. What we can do is note what the separate order-backlog table says about which names are currently placing equipment orders — and there, Micron is listed first.

🗓️ How Did Techwing Get Here?

Three things made this company what it is: winning Intel as a customer in 2016, a founder handover in the same year, and its first Samsung HBM Cube Prober order in January 2025. The rest of the timeline is capacity.

Sources: the company-history table (회사의 연혁) in the DART half-year report 2026.06, rcpNo 20260814001211, for every row through Jul 2025; KOSDAQ segment-change disclosure, rcpNo 20250430901394; treasury-share disposal result report, rcpNo 20251020000171; large-holding report, rcpNo 20260824000064.

Four consolidated subsidiaries carry the group beyond the parent: ENC Technology (이엔씨테크놀로지, founded October 1999, display inspection equipment), Techwing Vina (Vietnam, August 2017, semiconductor inspection equipment manufacture), FAS (에프에이에스, June 2019, factory automation), and Techwing Malaysia (August 2022, distribution and service). Techwing Malaysia booked its first revenue in the first half of 2026 — ₩24,700,000, about $18,000, which is a rounding error, but it is the first time a service line has shown up at all.

🌍 Is Techwing Actually a Global Leader, or Just a Korean One?

Techwing’s own annual report says its market share cannot be estimated — and then says the segment it wants to enter is effectively owned by Japanese competitors. Both statements are in the filing; neither has an English equivalent.

Here is the first, verbatim from the annual report’s industry section:

“반도체 장비 제조 산업의 특성상 타 소자업체에 대한 총 발주량 및 업체별 수주량을 파악하기 어려운 관계로 실질적인 시장점유율 추정은 불가능 합니다.”

“Because of the nature of the semiconductor equipment manufacturing industry, it is difficult to ascertain total order volumes to other device makers and order volumes by supplier, and therefore a meaningful market-share estimate is impossible.” — DART annual report 2025.12, rcpNo 20260623000173

This is now the second Korean chip-equipment maker we have found saying the same thing in its own filing — HPSP wrote effectively the same sentence. It is worth treating as a sector norm rather than a company quirk: in back-end equipment, nobody can see the denominator, so a share claim from anyone in this space deserves the question “measured against what?”

The second statement is about the Probe Station, which is Techwing’s stated route into wafer-level test. The temptation is to quote the first half of the sentence and stop. Read to the end:

“도쿄일렉트론(TEL) 등 일본 업체들이 프로브 스테이션 시장을 사실상 점유하고 있습니다. 당사의 제품이 정식 양산공급에 성공한다면 국산화 대체 효과가 커질 것으로 기대됩니다.”

“Japanese companies including Tokyo Electron (TEL) effectively hold the probe station market. If our product succeeds in formal volume supply, the localisation substitution effect is expected to be large.” — same filing

That is a conditional, not an achievement. The company then says it plans to enter volume lines with the TWP300 and lift share with a second-generation vision model, the TWP300H. As of the half-year report, no probe-station revenue is broken out — the “other test equipment” line is 0.9% of sales.

The industry section is similarly candid about the wider picture: “the global semiconductor equipment market is held at high share by overseas companies in the US, Japan and the Netherlands; localisation is rising thanks to domestic equipment makers including us, but the localisation ratio remains low.“ It also notes that equipment orders lag the semiconductor cycle by roughly six months and swing harder than the chip cycle itself — which is worth remembering when reading any single half-year from this sector, and which is the mechanism underneath Korea’s chip supercycle.

What is verifiable: 905 registered patents as of June 30, 2026 (485 in Korea, 420 abroad), up 38 year on year. The overseas split is unusual — Taiwan 193, China 154, US 41, Japan 20, Malaysia 8, Singapore 4. More patents filed in Taiwan than anywhere outside Korea is a fair proxy for where this company expects its machines to end up: OSAT and foundry test floors, not just Korean fabs.

📈 Is the Business Getting Better or Worse?

Operationally, sharply better; on the reported bottom line, worse. Those two sentences describe the same six months, and separating them is the whole point of reading the filing rather than the headline.

Source: DART annual report 2025.12, rcpNo 20260623000173, consolidated summary financial information (연결 요약재무정보) and consolidated income statement. The FY2025 and FY2024 operating margins are the filing’s own printed ratios; the FY2023 margin is ours, computed from the same table. FY2025 was Techwing’s first profitable year in three.

Source: DART half-year report 2026.06, rcpNo 20260814001211, consolidated statement of comprehensive income. The filing prints three-month and cumulative columns side by side; we keep them separate, because they are different periods and must not be read across. Margins are ours, computed from the filing’s won figures.

The operating story is clean and it is not a currency illusion at the operating line. Revenue rose 30.8%, gross margin widened 2.7 points to 42.8%, and SG&A fell 3.3% in absolute won. That combination — more revenue at better margin on lower overhead — is operating leverage, and it produced a 172.7% jump in half-year operating profit. Quarterly, Q2 operating profit rose 96.6% year on year at a 22.0% margin.

Then everything below the operating line went the other way. Other expenses went from $3.7M to $16.2M, other income collapsed from $18.7M to $4.7M, and net finance cost quadrupled. Pre-tax profit turned negative. This is the mirror image of what we found at SK hynix, where non-operating income matched operating profit — same structural point, opposite sign, much smaller company.

🔍 Why Did a 172.7% Jump in Operating Profit End in a Net Loss?

Because Techwing carries $148.0M of contracts to sell US dollars forward — more than its entire FY2025 revenue — and the won moved the wrong way for them. The derivatives line alone swung $31.0M year on year, which is four and a half times the company’s entire FY2025 net profit.

Start with what “other expenses” actually contains. The note breaks it out, and it is almost entirely one thing.

Source: DART half-year report 2026.06, rcpNo 20260814001211, Note 27 (기타수익과 기타비용, consolidated). Cumulative six-month columns only. In won: derivative valuation loss ₩12,629,416,746 (H1 2026) against a valuation gain of ₩17,208,203,114 (H1 2025). The −$31.0M net derivatives swing quoted in the text is our subtraction of those two years, not a figure the filing prints.

Now the instrument. Note 8 identifies every one of these derivatives as a currency forward (통화선도계약) — not an option on the exchangeable bond, not an interest-rate swap. And the business section prints the whole open book, counterparty by counterparty, with direction.

Open forward contracts as at June 30, 2026, exactly as printed. The same filing prints the identical five counterparties and rates as at December 31, 2025, with larger notionals at three of them, for a book totalling $182.5M — so the book shrank by roughly $34.5M over the half. Both totals are our addition of the five notionals the filing prints; the filing itself gives no total. Source: DART half-year report 2026.06, rcpNo 20260814001211, section 5 “위험관리 및 파생거래”. Dollar notionals are the filing’s own units and are not converted. A separate XBRL table in the same filing carries the same rates on a “lower range / upper range” axis and prints ₩1,331.30 rather than ₩1,131.30 for the first Samsung Futures line. We use the narrative table, and the reason is arithmetic rather than preference: the XBRL puts 1,331.30 in the lower-range cell and 1,179.70 in the upper-range cell of the same contract, which cannot both be true. The identical inversion is copy-propagated into the prior-period block and into the separate-entity statements, so it is one typo replicated, not two independent figures disagreeing.

Every position is a sell. That is what an exporter does: it expects dollars, and it locks in a won price for them in advance. When the won strengthens past the contracted rate, the hedge pays and the export receipts shrink. When the won weakens past it, the export receipts swell and the hedge bleeds.

Here is the won’s actual path, which is the missing half of the story:

USD/KRW settled closes on each reporting date itself, from our own price series. Note the direction of travel is not the whole mechanism: the derivative gain or loss in any period is the change in the book’s fair value over that period, not the level of the rate on the last day — which is why the same forward book produced a $12.8M valuation gain in the first half of 2025 and a $9.4M valuation loss in the first half of 2026.

So: the won ran from ₩1,437.91 at the December 31, 2025 reporting date to ₩1,548.61 at June 30, 2026, and Techwing was contracted to hand over $148M at rates as low as ₩1,129.60. That produced a $9.4M mark-to-market loss, plus $6.7M of realised losses on contracts that actually matured — against $2.0M of realised derivative gains in the same period, so the net realised cost was $4.7M. The derivative liability on the balance sheet grew from $10.9M to $16.1M (₩14.71bn to ₩21.69bn), and derivative assets, which stood at ₩34,864,470 at the end of 2025, were at exactly zero on June 30, 2026.

And the half-year frame hides the turn

Q2 2026 was net-profitable; the entire first-half loss was made in Q1. The filing prints three-month and cumulative columns side by side, and they point in opposite directions. Because the headline number every screener carries is the cumulative one, the reversal is invisible unless you read the column next to it.

Source: DART half-year report 2026.06, rcpNo 20260814001211, consolidated statement of comprehensive income and Note 27. The Q2 column is the filing’s own three-month column; the Q1 column is the cumulative column minus the three-month column, which is our arithmetic, not a figure the filing prints. The Q1 EPS is derived the same way and is therefore approximate. Controlling-interest and total-consolidated net income are different lines and are shown separately — they are not interchangeable.

Seventy-eight per cent of the half’s net derivative damage landed in the first quarter, which is where the won’s move against the book was steepest — spot went from ₩1,437.91 at the December 31 reporting date to ₩1,503.33 by March 31, then added only ₩45 more over the whole of Q2. By the second quarter Techwing was back to a $3.86M controlling profit on a 22.0% operating margin. Anyone reading “Korean equipment maker posts first-half loss” is reading a quarter that ended in March.

The detail that changes how you read this is the vintage. Two of the five lines — $37.3M of notional, a quarter of the book — were contracted between March 2021 and May 2022, at rates between ₩1,129.60 and ₩1,266.00, and did not mature until July 31, 2026. Those are hedges struck in the pre-2022 won regime that had to be carried for five years through a currency that never came back. This is not a trading position gone wrong; it is a long-dated export hedge outliving the assumptions it was written under.

Two further things sit in the same neighbourhood and belong in the same frame.

Interest cost is real and rising — on a shrinking loan book. Techwing carried $252.4M of borrowings at June 30 — ₩83.20bn short-term bank debt, ₩56.78bn of long-term debt falling due within a year, ₩125.80bn long-term, and ₩73.99bn of the exchangeable bond net of the exchange-right adjustment. Against ₩61.89bn of cash, that is ₩277.87bn ($206.4M) of net debt on ₩207.24bn ($154.0M) of equity. Total liabilities to equity is 213.9%. Half-year finance costs were $8.5M, up from $5.2M — but total borrowings fell $26.7M over the same six months, from $279.1M at the end of 2025. The cost went up while the principal came down, which points at rates and at the mix rather than at new leverage.

And a large one-off lands in the third quarter, not this one. On May 15, 2026 the board approved the sale of land (13,067.3㎡) and a building (16,714.38㎡) at 47 Dongtan Sandan 8-gil, Hwaseong, to Nine Bell Co., Ltd. (주식회사 나인벨) for $44.2M (₩59,500,000,000) excluding VAT on the building — 9.39% of FY2025 total assets. The stated purpose is “asset efficiency, improvement of the financial structure, and securing liquidity.” Those assets sat on the June 30 balance sheet as non-current assets held for sale at a carrying value of $17.3M (₩23,321,337,565 — land ₩10.74bn, building ₩12.58bn, per Note 34 of the half-year report, rcpNo 20260814001211). The gap between the two filed figures is $26.9M; the accounting gain will differ from that by disposal costs and tax, which the filings do not give. The disposal completed on August 31, 2026 — Techwing amended the filing that day to attach the completion certificate and the payment receipt. (Sources: rcpNo 20260515902262, disposal decision; rcpNo 20260831900368, completion amendment.)

A second property programme belongs next to that one, and it only exists in Korean. The annual report’s capital-investment table shows Techwing bought 25,043 pyeong of industrial land in the Cheonan BIT complex between April 2021 and July 2025 for ₩41,274,417,419 ($30.7M), approved by the board on 28 April 2021. No standalone disclosure was ever filed for it: the report states plainly that the total “fell below the disclosure threshold (10% of total assets)”. The same note says the site will be used, once title transfers, for “future growth product lines (Cube Prober, vision-application equipment and the like)”. Then, on 3 March 2026, the board approved a further ₩15.6bn ($11.6M) of Cheonan BIT land — a single line in the half-year report’s board-minutes table, with no stated purpose and no threshold note attached to it. So the Cheonan programme is at least $42M across five years, none of it separately disclosed, while in the same year Techwing is selling a plant for $44.2M. (Source: annual report 2025.12, rcpNo 20260623000173, business section; board resolution table in the half-year report.)

Put the two together and you get an unusually clean statement of this company’s problem for a foreign investor: in 2026, Techwing’s reported net income has been driven by a currency book in the first half and will carry a property sale in the second. Neither has anything to do with test handlers.

📦 What Is Actually in the Order Book?

$44.8M of equipment backlog, and the customer named first on it is Micron, not a Korean chipmaker. Every dollar-denominated supply contract Techwing has disclosed in 2026 has been with Micron.

Single supply-contract disclosures filed in 2026. Amounts are exactly as the filings state them — each Micron contract prints its own USD figure and the Hana Bank rate used to convert it (₩1,445.20 on Feb 3; ₩1,478.60 on Jul 16; ₩1,412.20 on Aug 10), so we do not re-convert them. The “% of prior-year revenue” column is the filing’s own ratio; the February contract is measured against FY2024 revenue and the rest against FY2025. Sources: rcpNo 20260204900276, 20260414900549, 20260519900243, 20260720900139, 20260810900346.

Three observations that only exist if you read the Korean.

First, the Micron orders total $36.76M across four countries — China, Malaysia twice, and India — in seven months. That is the geography of Micron’s assembly-and-test footprint, and it is where Techwing’s dollar receipts (and therefore its hedging need) come from.

Second, the one Samsung contract is the only one denominated in won, and it is the only one that slipped. The HBM test equipment order announced May 19 originally ran to August 13, 2026. Techwing amended it on August 4 (rcpNo 20260804900350) and again on August 13 (rcpNo 20260813900654), both times with the same stated reason: “고객사 요청에 의한 제품 납기일 변경으로 계약종료일 변경” — “change of contract end date due to a change in product delivery date at the customer’s request.” The end date moved to October 30, 2026 and the contract value did not change. Measure the slip on delivery rather than on the end date and it is starker: the original May 19 filing expected final delivery on July 12, 2026; both amendments now say September 29, 2026 — a 79-day slip.

There is a reason it was filed twice rather than once, and it is visible only if you open both. The August 4 amendment announced the new end date in its correction table and updated its own explanatory note — but left the contract-terms block below still printing 종료일 2026-08-13, the very date it was moving. The August 13 filing, made on that now-stale date, is the one that finally carries 종료일 2026-10-30 in the contract table itself. So this is a two-and-a-half-month slip on the company’s flagship HBM order, disclosed twice because the first disclosure did not fully take, in Korean, with no English coverage we could find.

Third, the backlog is dated later than the report it appears in. The order-backlog table is headed “기준일: 2026.06.30” but carries a footnote saying the figures are as of the half-year report’s drafting date, August 7, 2026. It also excludes parts backlog entirely — which, given that parts are 44.9% of revenue, means the backlog understates forward business by construction. Techwing says so itself: order-to-delivery is short, amounts and dates change frequently, and parts orders are too frequent to track, so “it is difficult to project revenue on backlog alone.”

Excludes parts backlog, per the filing’s own footnote. Source: DART half-year report 2026.06, section 4-4 “수주 현황”.

One more balance-sheet line supports the demand picture, and it is audited rather than announced: customer advances (선수금, a contract liability for a performance obligation satisfied at a point in time) rose from $1.2M at end-2025 to $23.9M at June 30 — a 19.8× increase. Customers have prepaid for equipment not yet delivered. That is a real cash-in-hand demand signal, and it sits in Note 16 where nobody reads it.

💰 What Do You Actually Pay for Techwing Today?

About $1.30bn for a company that earned $7.0M last year — 185.6× trailing earnings, the second-highest multiple in its peer group and the lowest operating margin. The multiple is a statement about what the market thinks 2027 looks like, not about what 2025 was.

Price, share count and market capitalisation from the Korea Exchange settled close of September 4, 2026. P/E uses FY2025 net income attributable to controlling interests of ₩9,402,064,061 (DART annual report 2025.12). Dividend of ₩130 per share from the cash dividend decision of Feb 9, 2026, rcpNo 20260209900763. The 52-week window runs September 5, 2025 to September 4, 2026. We show closing and intraday ranges separately because they are not interchangeable and one argument later in this article turns on which you use. Prices go stale — check a live quote before acting on any of this.

How does that compare to its peers?

All market caps at the KRX settled close of September 4, 2026. FY2025 revenue, operating margin and equity are from each company’s DART annual report, and the basis is not uniform: Hanmi, ISC and Techwing file consolidated statements, while Leeno Industrial and Doosan Tesna file no consolidated statements at all — DART returns none for FY2025 — so those two rows are separate-entity (별도) figures. We say so rather than smooth it: a consolidated row and a separate-entity row are not the same measurement. P/B is computed on FY2025 total equity for every row so that at least that basis is identical across the table (Techwing’s 8.2× here therefore differs slightly from the 8.31× above, which uses June 30, 2026 controlling equity). These are back-end test and inspection names, not a matched peer set: Hanmi makes HBM bonders, Leeno and ISC make test consumables, and Doosan Tesna is a test service house rather than an equipment maker. We flag that rather than smooth it, and the same caution applies to Hanmi’s own half-year numbers and ISC’s.

Two things stand out. Techwing has the lowest FY2025 operating margin of any profitable name in the group — 10.0% against 27.3% to 47.5% for the consumables makers — and it trades on the highest multiple of any of them. Both are explicable and neither is comfortable. The margin gap is structural: Techwing sells machines with bought-in parts, while Leeno and ISC sell high-margin consumables. The multiple is a bet that the first-half operating margin of 20.3% is the new run rate rather than a cycle peak.

R&D spending is a useful cross-check on that bet, and it is going the wrong way as a share of sales: 15.80% of consolidated revenue in FY2024, 13.22% in FY2025 and 9.93% in the first half of 2026 ($21.8M, $15.6M and $8.1M respectively). A note on the denominator, because it matters: the filing prints its own ratios of 17.88% / 14.20% / 10.57% under the heading “(ratio to sales)” and states no basis — but those divide by the parent company’s separate-entity (별도) revenue ($121.8M / $110.1M / $76.2M, i.e. 88.4% / 93.1% / 94.0% of consolidated revenue), not by the consolidated revenue this article uses everywhere else. It is not the segment note either: that note puts first-half semiconductor revenue at ₩103,071,585,323, a different number again. On the consolidated base the FY2024 figure is 2.1 points lower. The direction survives either way; the label does not. In absolute won, first-half spend annualises close to the FY2025 level; the intensity fell because revenue rose. For a company whose stated growth path runs through a product category that Tokyo Electron currently owns, that is a number to watch rather than dismiss.

🏷️ Which Techwing Ticker Are You Actually Buying?

There is exactly one: KOSDAQ 089030. No ADR, no preferred share class, no listed holding company, and no listed subsidiary. That is unusually simple for a Korean industrial, and it is worth stating because most of the companies we cover are not.

Source: DART half-year report 2026.06, subsidiary history and share-total sections; large-holding report rcpNo 20260824000064.

The absence of an ADR matters practically. Techwing is not on our list of Korean stocks available as US-listed ADRs, which means a foreign investor buys it on KOSDAQ or not at all — see how to buy Korean stocks as a foreign investor for the mechanics, and our note on converting Korean share prices into USD for why the dollar figure you see on a screener may not match the one you compute.

🌏 Is Foreign Money Buying or Selling Techwing?

Both, depending entirely on which sessions you count — and the trailing windows every screener shows you are the ones most likely to mislead. Foreign investors are net buyers over 5, 20 and 55 sessions, but all three windows end on the same day, so they are one observation dressed as three. Split the same series into stretches that do not overlap and the sign flips.

Source: our own KRX-derived investor-flow series, which for this stock begins on June 18, 2026 — the 55-session column is therefore the longest window available, not a chosen one. Method: each day’s net share count is valued at that day’s close and summed, then converted at ₩1,345.99/$1. This is an approximation of value traded, not execution prices. Foreign ownership of 17.4% as of September 4, 2026 is from Naver Finance, not from the KRX series — it is the one figure in this section we did not compute ourselves.

The reversal is in the institutional row and it is worth saying out loud rather than picking the convenient half. Over 55 sessions, institutions were net buyers of 147,419 shares. Over the most recent 20 of those sessions they were net sellers of 324,446. That means essentially all of the institutional accumulation happened in the earlier stretch and has since been given back and then some. A note that said only “institutions have been buying Techwing” would be true and useless; so would one that said only “institutions are selling.” For how these flows work and why the window matters, see our primer on foreign and institutional flows and the running Foreign Flow Watch series.

Set the flows against the price and the trailing windows fall apart. Over the 55 sessions the stock went from ₩60,500 on June 18 to ₩47,100 on September 4, a fall of 22.1%. The 20-session window starts from a ₩43,600 close on August 7 and is +8.0%. The most recent five sessions are −5.0%. So the three columns above describe a long decline, a rebound and a pullback — and foreigners are net buyers in all three, which reads like conviction independent of direction.

It is not. Every one of those windows ends on September 4 and therefore contains the same recent buying; the only thing that changes is how much earlier history gets included. Cut the series into calendar months instead, which do not overlap and which we did not choose:

Same series, same method (each day’s net share count valued at that day’s close, summed, converted at ₩1,345.99/$1) — an approximation of value traded, not execution prices. The June and September buckets are partial: our flow series for this stock begins June 18, 2026 and the table ends at the September 4 settled close.

“Foreigners have been buying Techwing” survives that test in three buckets out of four and fails badly in the first, where they sold 800,209 shares into an 11.4% drop. And the sharpest version is not a monthly split at all. Cut the same 55 sessions into two halves that do not overlap:

Same series and same method as above — an approximation of value traded, not execution prices. The two columns share no sessions, so the figures add to the 55-session column in the first table of this section. July 9 is not a date we picked for effect — and it is not the most favourable one either. Cumulative foreign net buying in Techwing (089030) actually bottoms two sessions earlier, on July 7, at −1,463,740 shares. Splitting there instead gives foreigners −$52.7M in Techwing (089030) from 18 June to 7 July 2026, into a 32.1% fall, then +$84.1M from 8 July to 4 September 2026, into a 24.1% rise. We show the July 9 cut because it is the more conservative of the two. Because it is the most favourable cut for this argument, here is a second one fixed by the price rather than by the flow — split at the July 30 closing low instead and foreigners are flat into a 47.3% fall (+35,714 shares, −$0.1M) and +$31.5M into the rebound. Same story, smaller first leg.

So foreigners sold into the collapse and bought the rebound. That is a $126M swing inside a series the trailing table above shows as one tidy positive number.

The window-free way to say it is a correlation. Across all 54 daily observations we hold, foreign net buying and the same day’s price move correlate +0.60 — foreigners are net buyers on up days and net sellers on down days, consistently. The correlation with the previous day’s move is −0.005, essentially nothing, so this is not a next-day chase. And here is the honest limit: a same-day correlation cannot separate “foreigners follow the price” from “foreign flow moves the price,” and on a $1.30bn KOSDAQ name a single large foreign day can plausibly do the latter. What it does establish is the thing that matters for reading any flow table, including ours: because foreign flow tracks the day’s direction, the net over any multi-week window depends on the mix of up and down days inside it, not on where the window started and ended. That is why June nets −$31.2M in a falling month and July nets +$41.0M in a worse one.

We are applying the same discipline here that we applied to the institutional row above. It would have been easy not to: the institutional reversal is a $16.5M swing across the trailing windows and we called it out, while the foreign reversal would have read better left alone. Measured like for like on the non-overlapping split, the institutional swing is $24.6M against the foreign $126.0M — about five times, not eight.

Retail is the clearest line on the other side: net sellers of 1,046,263 shares over 55 sessions, roughly 2.8% of the entire share count, into a stock that is up just 2.8% year to date and sits 33.8% below its 52-week closing high. Foreigners have absorbed most of it.

Flow data tells you how many shares moved. It does not tell you whose. For that there is a second and much slower record — Korea’s 5% disclosure register — and Techwing has four institutional names on it. It is worth reading before deciding who the buyers and sellers above were, because in all four cases the headline percentage turns out not to mean what it appears to.

🧾 Who Actually Filed on Techwing’s Share Register?

Four institutions appear on Korea’s 5% disclosure register for Techwing (089030) over the past fourteen months — Samsung Asset Management, J.P. Morgan Securities PLC, Manulife and the National Pension Service — and in every one of the four, the headline percentage means something other than “this institution bought this much of the company.” One crossed the threshold on stock that was two-thirds borrowed. One saw its declared stake grow while it was selling on the exchange. One is five separate legal entities filing under a single name. One is a fund reported by an agency that holds no shares itself.

Korea’s 5% rule works roughly like a US 13D/13G: cross 5% of the voting shares and you file, then file again whenever your stake moves by a percentage point or more. The filings are public, in Korean, and almost never read in English. Here is every such filing on Techwing that we could find on DART, oldest first.

Every figure is the filing’s own, taken from the abbreviated major-holding reports (주식등의 대량보유상황보고서(약식)) linked in the “Filed” column. Each report prints three separate dates on its cover — the date the reporting obligation arose, the date the holding is measured, and the date it reaches DART — and we show all three because the gaps are the point. The “trigger → filed” column is our subtraction in calendar days. One denominator caveat: the July 2025 National Pension Service filing computes against 37,353,645 voting shares; every filing from November 2025 onward uses 37,053,645, after the 300,000-share cancellation described later in this article. The largest shareholder’s own August 24, 2026 filing is a different form — the “intent to influence control” version — and is covered in the governance section below.

J.P. Morgan’s 5.53% was two-thirds a lending book

J.P. Morgan Securities PLC appeared on the register at 5.53% on June 16, 2026 having bought 114,714 shares — 0.31% of the company — on the two days that produced the filing. Everything else was already there or was borrowed.

The detail schedule is itemised line by line, so it can be added up. Of the 2,050,004 shares reported, 1,687,641 were flagged 기존 보유주식 — “existing holding” — carried in on the schedule’s first day. The rest is nine further dated lines across June 15 and 16: two on-exchange purchases of 46,108 and 68,606 shares, at filed average prices of ₩69,831 and ₩62,676; four borrowings from named agents — Citibank NA as agent 8,812 shares, Kiwoom Securities 5,000, BlackRock Advisors (UK) as agent 15,881 and Vanguard Group Inc as agent 45,000; two lines marked “On-Lend” of 52,762 and 116,473; and a 3,721-share line marked simply “Adjustment” (Source: DART, rcpNo 20260618000400, Part 3 section 2, detail of changes).

Add those up. 4.55% of the company was already on the book before any of this. Of the 362,363 shares that actually moved, 114,714 — 0.31% of Techwing — were on-exchange purchases. The other 247,649, or 68.3%, were borrowings, On-Lend entries and an adjustment: stock that arrived on nobody’s decision to own the company. Those four figures are our sums of the lines above, not the filing’s.

The filing’s own layout gives the game away. Every on-exchange line carries a price in the 취득/처분단가 column. Every lending line leaves it blank.

The exit filing two weeks later runs the other way, and this is where Techwing differs from the last stock where we opened these schedules. It reports a net −531,942 shares across June 17–30 in fourteen dated lines: eight on-exchange sales, two on-exchange purchases, an On-Lend Return of 177,729 shares, a 3,900-share return of borrowed stock to Vanguard Group Inc as agent, and two small fresh borrowings from BlackRock entities (Source: DART, rcpNo 20260702000581, Part 3 section 2).

Sorted by method, −356,839 shares of that, 67.1%, was genuine on-exchange trading — 410,948 sold and 54,109 bought back — against −175,103 of lending mechanics. So J.P. Morgan walked onto the register mostly on borrowed stock and walked off it mostly by selling. Those totals are ours; the fourteen line items behind them are the filing’s.

Two checks worth stating because they are the ones most often skipped. First, the JPM filings list no special related parties — the 5.53% is J.P. Morgan Securities PLC standing alone, unlike Leeno, where the same bank’s 5.11% turned out to be J.P. Morgan Securities PLC at 2,884,338 shares (3.78%) plus J.P. Morgan Securities LLC at 1,011,868 (1.33%) (rcpNo 20260710000204). Second, both JPM filings leave the proprietary-account versus client-account table blank, so we cannot say how much of the position was the firm’s own risk.

On scale: JPM’s own filing records net on-exchange sales of 304,344 Techwing (089030) shares between June 18 and June 30, 2026 — 38.0% of the 800,209 shares foreigners sold net in that same nine-session bucket in the table above. We are putting those two numbers side by side, not equating them: we cannot verify that the KRX classified JPM’s trades in the foreign category, and the filing’s dates are contract dates, not settlement dates.

“Manulife IM (Hong Kong), 6.79%” is five entities in four countries

No single Manulife entity holds 5% of Techwing. The largest holds 3.99%, and the Hong Kong company whose name DART prints on the filing holds 0.96%. The reporting obligation arises on the combined book, which is why the aggregate appears next to one entity’s name — the same trap we flagged at Leeno, in a different shape.

Source: DART abbreviated major-holding report, rcpNo 20260907000214, Part 3, base date August 3, 2026. The filing states the group total as 2,514,814 shares, 6.79%, and its previous report as 2,021,475 shares, 5.46%. Per-entity share counts and percentages are both the filing’s own, printed in Part 3; computed against the 37,053,645 voting shares the same filing states, they add to the filing’s own 6.79% without rounding drift.

The second thing the Manulife schedules show is that every line that moves is an on-exchange trade with a price against it. There is no borrowing, no On-Lend, no adjustment anywhere in either filing — the opposite of the JPM book. (The July filing also carries five 신규보고 opening-balance lines with no price, which are the pre-existing position rather than trades; the September filing is all traded lines.) Whatever else this is, it is a manager buying stock on the KOSDAQ.

It is also mostly older than the filing. The July 31 report crossing 5.46% shows opening balances struck on July 22 totalling 1,665,817 shares — 82.4% of the reported position — with a footnote saying plainly that these were not acquired on that date but accumulated on several earlier occasions and are shown as an account balance for reporting purposes. Only 355,658 shares, bought on July 23, actually triggered the disclosure. A reader who saw “Manulife takes new 5.46% stake” on July 31 would have concluded that a foreign manager had just built a position. Four fifths of it was already there, and had been for an unstated length of time.

The second filing is the interesting one. It records 521,862 shares bought and 28,523 sold between July 24 and August 3 — a net 493,339. And 479,837 of those purchases, 91.9% of the gross, are dated July 30, July 31 and August 3 — the ₩31,900 close of July 30, the low of the 55-session window in the tables above, and the two sessions after it, at filed average prices of ₩33,926, ₩39,934, ₩45,827 and ₩45,914. We are reporting the dates the filing prints, not crediting anyone with timing: a daily average price says nothing about intent, and one manager’s book is not the market.

Set against the tape on the same window — the JPM comparison above matches its windows, so this one must too — Manulife’s disclosed net on-exchange buying between July 23 and August 3, 2026 is 848,997 shares against foreign net buying of 443,462 shares in Techwing (089030) over those same eight sessions: 191.4% of it. One manager’s disclosed buying is larger than the entire foreign net for the window, which means other foreign accounts were net sellers into it — or that some of these fills were not booked as foreign. (Over the longer July 10 – September 4 window, foreigners bought 2,207,722 shares net, and Manulife’s eight sessions are 38.5% of that; the two ratios measure different things.) The same caveat as with JPM applies, in the same words: side by side, not equated.

Samsung Asset Management’s stake grew while its trading sold

Across its three 2026 filings, Samsung Asset Management’s declared Techwing holding rose by 478,232 shares — and its own on-exchange trading over exactly the same span was net −508,278 shares. The register says the position grew. The trading says it sold. The difference is ETF creation and redemption.

Korean major-holding filings itemise the mechanism in a remarks column, and Samsung’s schedules run to 197, 302 and 85 lines respectively. Sorting them by that column gives this:

Our aggregation of the line-item schedules in rcpNo 20260206001611, rcpNo 20260602000340 and rcpNo 20260708000416, grouped by each line’s own 취득/처분방법 and remarks fields. The three filings state their own net changes as +1,148,207, −369,378 and −300,597 shares, and their closing balances as 3,006,275, 2,636,897 and 2,336,300 shares against 37,053,645 voting shares. “On-exchange” is 장내매수 less 장내매도; “ETF / fund mechanics” is every line the filings label ETF 설정 (creation), ETF 해지 (redemption), 일반펀드장외ETF설정 and 일반펀드장외ETF해지 (off-exchange transfers between funds and ETFs). The two columns add exactly to each filing’s own stated net change and to its stated closing balance, which is how we know nothing has been dropped.

The February filing is the clearest case. Samsung Asset Management’s declared stake jumped from 5.01% to 8.11% — by far the largest single move on this register — while it was a net seller of 220,823 shares on the exchange. The entire increase, and more, came from ETF units being created against Techwing stock.

Two further details make the point structural rather than incidental. The filings’ proprietary-versus-client account table shows zero shares in Samsung Asset Management’s own account and 100% in client accounts — this is money it manages, not money it owns. And the June 2 filing adds a parenthetical to its stated reason for filing: the figure includes assets delegated to it by Samsung Active Asset Management, a second manager whose name appears nowhere in the headline.

One thing this section is not: evidence about foreign flows. Samsung Asset Management is a Korean institution. Its trading belongs in the institutional row of the tables above, not the foreign one, and any reader tempted to fold an 8.11%-to-6.31% decline into a story about foreigners leaving Techwing is combining two different columns of the same data.

The register is stale on arrival — sometimes by a quarter

The National Pension Service has been in and out of the 5% line on Techwing three times, and its filings are the slowest on the register — 67, 90 and 65 days from trigger to DART. The October 2024 crossing was disclosed at 5.10%, still above the line. The next two filings were declines, and by the time each reached DART the position was already below 5%.

The earliest, filed January 6, 2025, reports the crossing itself — new, 1,904,539 shares, 5.10%, measured on October 31, 2024 (rcpNo 20250106000109) — 67 days after the obligation arose. Note the denominator: that filing predates the October 2025 treasury cancellation, so its percentage is struck on the larger share count.

The next, filed October 1, 2025, reports the fall from 5.10% to 4.03% measured on July 3, 2025 — a 90-day gap between the date the obligation arose and the date anyone could see it. The third, filed January 2, 2026, has a trigger date of October 29, 2025, when the fund held 1,891,160 shares — 5.10% of the 37,053,645 voting shares the filing itself states, on our calculation — and a base date of November 26, by which point it was down to 4.06%. Sixty-five days.

This is not unique to the pension fund, only most extreme there. J.P. Morgan filed in three days both times. Samsung Asset Management took 9, 21 and 15. Manulife took 8 days on its first filing and 35 on its second: the 6.79% stake measured on August 3, 2026 did not reach DART until September 7. Anything written about Techwing’s shareholder base between those two dates — including the first draft of this article — was working from a register that was already out of date.

We are deliberately not telling you what the legal deadline is. Every one of these filings is the abbreviated form, and its own cover states the category: “약식서식 : 자본시장과 금융투자업에 관한 법률 제147조에 의한 보고 중 ‘경영권에 영향을 주기 위한 목적’이 아닌 경우 및 보고자가 동조 제1항 후단에 따른 전문투자자인 경우” — the short form used for Article 147 reports where the holding is not for the purpose of influencing control, or where the filer is a professional investor. What deadlines follow from that sits in the Financial Investment Services and Capital Markets Act and its Enforcement Decree, which we have not read for this piece and will not summarise from memory. What we can show is what each filing prints on its own cover: three dates, and the distance between them.

One more identity check, since this is the section for them. The reporting person on both pension filings is 국민연금공단, the National Pension Service — the administering agency — and it holds zero shares. All 1,505,135 shares sit with 국민연금기금, the National Pension Fund, filed jointly as a related party. “NPS owns 4.06% of Techwing” is a sentence about the fund, reported by the agency.

A loose end we could not close

Samsung Asset Management’s first Techwing filing, measured November 25, 2025, put it at 1,858,068 shares — 5.01%, one hundredth of a point above the threshold. Against the pre-cancellation share count of 37,353,645 the identical holding is 4.97%, below it. The 300,000-share cancellation described later in this article cut the denominator by 0.80% at some point between the July 2025 and November 2025 filings, and the arithmetic is what it is. Whether Samsung Asset Management crossed 5% by buying or by the company shrinking underneath it, we cannot say: that filing records the whole holding as a single “new report” line with no breakdown, and its stated reason is simply that a reporting obligation arose from trading. We are flagging the coincidence and leaving it open rather than resolving it in the direction that would read better.

What the register does and does not tell you

It names holders on a date. It does not identify the flow. A 5% report is a point-in-time balance struck on a base date; our flow tables are daily net share counts across every account in a category. The two can be laid alongside each other, and we have done that twice above with explicit percentages, but they cannot be joined. Four filers are four books among thousands, and none of the four filings names a counterparty on the other side of a single share.

The register’s own construction is the strongest argument against reading it as a flow proxy. Of the four institutions on it, one crossed the threshold largely on borrowed stock, one had its stake pushed up by ETF creation while it sold on the exchange, one is a five-entity aggregate reported under the name of a member holding 0.96%, and one is a pension fund whose crossings were public only after they had reversed. This is a record of who held what and when it was measured — a genuinely valuable record, and one almost nobody reads in English. It is not a ledger of who bought.

🏛️ Who Controls Techwing, and What Do Minority Holders Get?

A family group controls 24.13% through the CEO plus three unlisted companies — and 38.4% of that block is pledged against $44.6M of broker loans. This is where the Korea Discount lives, and none of it is in English.

The control chain

Source: DART large-holding report (일반), rcpNo 20260824000064, filed Aug 24, 2026, and annual report 2025.12, rcpNo 20260623000173, minority-shareholder section. The individual percentages are the filing’s own and are computed against 37,053,645 voting shares — with one exception: the related-parties row is our aggregation of three separately filed holders (Na Ki-hoon 40,032 shares / 0.11%, Na Hyun-jong 40,172 / 0.11%, Jang Nam 18,000 / 0.05%), so both its total and its percentage are ours, not the filing’s. The annual report separately reports 24,114,642 shares — 65.08% — held by 73,553 minority shareholders as of December 31, 2025.

The structure is a small pyramid: Na Ki-hoon controls Trutech, Trutech controls Dasomtech and Nanoace, and those three unlisted companies hold 10.36% of Techwing between them. Note also that Dasomtech sits at 4.99% and Trutech at 4.75% — both just under the 5% level at which a Korean holder incurs standalone large-holding reporting obligations. We are describing arithmetic, not motive; the filing offers no explanation and we are not going to invent one.

The pledges

The August 24 filing was made for one reason: “보유주식에 대한 담보계약 변경 · 특관자 담보계약 변경” — changes to share-pledge contracts by the reporting person and related parties. Pledged shares rose from 3,234,861 to 3,436,084 across fourteen separate contracts.

The filing states the pledged stake as 9.37%, and prints that figure three times. It does not reconcile: 3,436,084 ÷ 37,053,645 voting shares is 9.27%. The 9.37% is the sum of the filing’s own fourteen rounded per-contract percentages, and one of those fourteen rows is wrong — Dasomtech’s 300,000-share pledge to Daishin Securities is printed as 0.89% when 300,000 ÷ 37,053,645 is 0.81%. The 0.89% appears to have been copied from the Trutech row two lines down, which pledges 330,000 shares to the same brokerage and for which 0.89% is correct. Every other percentage in this article divides cleanly by 37,053,645; this one does not, and the gap is 0.10 of a point. We use 9.27% where the arithmetic matters and flag the filing’s 9.37% where we quote it.

Fourteen pledge contracts across eleven counterparties — ten brokerages (Daishin, Yuanta, IBK, Woori, Shinhan, KB, LS, Hyundai Motor Securities, DB, Korea Investment) plus Korea Securities Finance, which is a state-linked securities-finance institution rather than a brokerage. The fourteen individual loan amounts printed in the filing add to $44.6M (₩60.0bn), and the fourteen share counts add exactly to the filing’s own stated total of 3,436,084 — both additions are ours. Source: DART large-holding report, rcpNo 20260824000064, section 2-나 and 2-다.

At the September 4 close of ₩47,100, the pledged shares are worth $120.2M against $44.6M of loans — aggregate collateral coverage of about 270%. That aggregate is where most write-ups would stop, and it is misleading, because the filing prints the loan amount, share count and required maintenance ratio for each of the fourteen contracts separately. Do the division contract by contract and four of the fourteen sit below their own stated maintenance ratio on the pledged Techwing shares alone: LS Securities (53,938 shares against a $1.5M loan — 127% against a required 140%), the smallest KB contract (29,198 shares against $0.7M — 137% against 140%), DB Securities (66,778 against $1.5M — 157% against 200%) and Korea Investment (72,675 against $1.5M — 171% against 180%). There is also no relationship between contract size and required ratio: the smallest KB loan carries a 140% requirement while the 200% requirement sits on a loan twice its size — and the book’s lowest tier, 110%, sits on a larger loan again.

Three honest caveats, and they matter. The loan amounts are as filed at each contract date and may since have been partly repaid; a pledger can post cash or other assets that a share-pledge filing never shows; and two of these four contracts have since reached the end of their disclosed term — the KB pledge term ended on 7 September 2026 and the DB pledge term on 14 September 2026 (two contracts outside this four, IBK’s 181,052 shares and KB’s 393,384, also ended on 14 September). No updated large-holding report has been filed, so whether those contracts were repaid, rolled or replaced is not in the public record; the expiry of a pledge term is not itself a disclosable event. What the expiries do not do is retire the finding: re-run at the regular-session close of ₩43,400 on 14 September 2026 — the official 09:00–15:30 close, not the after-market print, a distinction that only came into existence that day — and it is still four of the fourteen below their ratio: the same four, with less cushion (117%, 127%, 145% and 158%) and aggregate coverage down from 270% to 248%. Valuing the same collateral at the after-market price instead moves each ratio by under a point and leaves the count at four. So the correct reading is not “a margin call is coming” — it is that on the collateral this filing actually discloses, four of these fourteen contracts are already in top-up territory at ₩47,100, and the aggregate 270% conceals that completely. We are deliberately not computing forced-sale trigger prices, because a trigger price built on an assumption about undisclosed collateral is worse than no trigger price.

The treasury stock that was already sold

On October 2, 2025 Techwing’s board took two decisions on the same day. It resolved to cancel 300,000 common shares with a book value of ₩2,652,809,551 ($2.0M), “for the purpose of enhancing shareholder value and expanding the shareholder return policy.” And it resolved to dispose of 1,313,171 common shares at ₩71,060 each — ₩93,313,931,260 — with the stated purpose “issuance of an exchangeable bond with treasury shares as the exchange asset.”

Before that day, Techwing held 1,613,171 treasury shares (4.3%). It cancelled 18.6% of them and committed the other 81.4% to bondholders.

Sources: material-event report, treasury share disposal decision, rcpNo 20251002000526; share cancellation decision, rcpNo 20251002900612; treasury share disposal result report, rcpNo 20251020000171; half-year report Note 17-2.

This is a genuinely elegant financing. Techwing raised $69.3M at zero interest for five years by selling a call option on 3.54% of itself, struck 20% above where its stock was trading. It is also, from a minority shareholder’s point of view, an overhang that a screener will not show you — and it is the same structure we decoded at ISC (095340) — a KOSDAQ component maker turning its treasury block into an exchangeable bond rather than cancelling it. We are not going to call two instances a trend, and the pattern is not confined to small caps either — we found SK hynix releasing 6,179,194 treasury shares to holders of a 2023 exchangeable bond on the same logic. What it is enough for is a rule of thumb — the treasury-share line is now the first thing we check on any Korean chip-supply-chain name, because a screener will show it as a buffer when it may already be spoken for.

Three things follow that are worth stating precisely.

One: two official filings disagree about how many treasury shares exist, and both are correct. The October 20 disposal result report says the shares are “deemed to have been exchange-requested on the same date” and reports treasury holdings after disposal of zero. The annual report, as of December 31, 2025, says: “당사는 자기주식으로 총 1,313,171주를 보유하고 있으며, 해당 주식수는 총 발행주식의 3.54%에 해당합니다” — the company holds 1,313,171 treasury shares, 3.54% of shares issued. The reconciliation is in the footnote: the shares are physically deposited at the Korea Securities Depository and will only actually transfer when a bondholder exercises. One filing reports the accounting deeming; the other reports the physical position.

Two: the cancellation did not increase anyone’s dividend by one won. The February 2026 dividend filing states the dividend base explicitly — 37,053,645 shares issued at December 31, 2025, less 1,313,171 treasury shares, equals 35,740,474 dividend-bearing shares. The prior year’s base was the same 35,740,474, because 37,353,645 issued less 1,613,171 treasury also equals 35,740,474. The 300,000 cancelled shares were treasury shares that had never received a dividend. Dividend per share was ₩130 in both years; total dividends were ₩4,646,261,620 in both years. A $2.0M cancellation that was announced as “expanding the shareholder return policy” left the per-share dividend, the total dividend, and the dividend-bearing share count all exactly unchanged.

Three: the annual report did not carry any of this until a second amendment, filed eight months after the event. The disposal itself was not hidden — Techwing filed three material-event reports about it in October 2025, all cited above. What was missing was its appearance in the annual report, the document a foreign investor actually reads. Techwing filed the FY2025 annual report on March 16, 2026, amended it the same day to add an omitted accounting-policy note, and then amended it a second time on June 23, 2026 — eight months and three days after the October 20, 2025 disposal. What that second amendment added was a new section, “3. Treasury share direct acquisition and disposal implementation status,” disclosing the October 2025 disposal — ₩93,314 million, or $69.3M — in the annual report for the first time, alongside disposals from 2015, 2018 and 2019. The amendment form gives the reason for change as 미기재 — “not stated” — and marks “related to a correction demand or order: no,” so it was not filed at a regulator’s instruction.

Two smaller corrections rode along in the same filing, and one of them is not what it looks like. A period label in the three-year loss-allowance table changed from “2025년 3분기 (24th term)” to “2025년 (24th term)” — the annual report had been carrying a Q3 heading over full-year figures. And an address was corrected from “경기도 화성시 동탄산단6길 37” to “경기도 화성시 동탄구 동탄산단6길 37,” inserting a missing district. That row is in the subsidiary detail table and belongs to ENC Technology, not to Techwing — Techwing’s own cover address already read 동탄구 in all three versions. We flag it because the correction is easy to misread as the parent fixing its own headquarters address, and it isn’t.

Value-Up, and the $11.1M That Moved Quietly

Techwing filed a Corporate Value-Up plan on March 25, 2026. It contains no numeric target of any kind. The entire content is eight bullet phrases — “enhance growth,” “pursue a shareholder return policy,” “pursue sustainable management,” “strengthen investor communication,” and four matching implementation lines. The filing explains why there is no attached plan document: the company qualifies as a “high-dividend enterprise” under Article 104-27 of the Restriction of Special Taxation Act, and such companies may state the key points in the disclosure instead of attaching a plan.

The disclosure’s own numbers are the interesting part. It reports FY2025 dividends of ₩4,646,261,620, FY2024 dividends of ₩4,646,261,620, and — in its own field — a year-on-year dividend growth rate of 0%. A company can be a designated high-dividend enterprise on a 49.4% payout ratio while its dividend has not moved and its yield is 0.3%. Both facts are true at once, because the payout ratio is measured against earnings that are small relative to the market capitalisation.

One thing did move, at the same annual general meeting. On March 24, 2026 shareholders approved the transfer of $11.1M of capital surplus into retained earnings (₩15,000,000,000). The consolidated statement of changes in equity shows it as a single line: capital surplus down ₩15.0bn, retained earnings up ₩15.0bn, net effect on total equity zero, cross-referenced to Note 22. Mechanically, this moves money from a reserve that generally cannot be distributed into the pool that dividends and buybacks are paid from — $11.1M is 3.2× the FY2025 dividend.

And the company said exactly why it was doing it — in the AGM convocation notice, in Korean, three weeks before the vote. The half-year report’s note records only the resolution, which is why this looks unexplained if you read the financial statements alone. The notice does not leave it open:

“상법 제461조의2(준비금의 감소)에 따라 자본잉여금(주식발행초과금) 일부를 이익잉여금으로 전환하는 건입니다. 본 건 자본잉여금의 이익잉여금 전입의 목적은 배당가능이익을 추가로 확보하는 것으로, 전입된 금액은 관련법령에 의거하여 향후 비과세 배당 재원으로 사용 가능합니다.”

“Under Article 461-2 of the Commercial Act (reduction of reserves), this converts part of capital surplus (share premium) into retained earnings. The purpose of this transfer is to secure additional distributable profit, and under the relevant legislation the transferred amount can be used as a source of tax-exempt dividends in future.” — DART AGM convocation notice, March 4, 2026, rcpNo 20260304000354

Two things follow that a foreign reader would not otherwise get. First, this is a deliberate, disclosed enlargement of the dividend capacity of a company whose dividend has not moved in two years — the capacity was created before the intent to use it, and the intent is still not stated. Second, the notice’s own phrase is 비과세 배당 재원 — a source of non-taxable dividends. We are quoting the company, not modelling the tax: whether and how a particular distribution out of transferred share premium is taxed is a matter of Korean tax law and the holder’s own residence, and Techwing does not spell that out. What a foreign holder should note is only that the tax character of a future Techwing dividend is not automatically the same as the ordinary case set out in our guide to Korean dividend withholding tax. What is still genuinely unknown is whether the capacity gets used — that is the thing to watch, not the purpose, which the company has already given.

⚠️ What Would Make This Article Wrong?

The most likely way this piece is wrong is that the forward book is treated as a one-off when it is structural. Six specific things would break the reading above, and they are worth stating before the conclusion rather than after it.

📚 Lingo Check

🎯 Why It Matters for K-Export Stars

Techwing matters here because it is a Korean exporter whose entire investment case is written in a language its potential foreign buyers cannot read. It is close to a pure expression of what this site exists to cover. It is a Korean exporter — 79.3% of revenue leaves the country — supplying the back end of a global semiconductor supply chain, with essentially no English-language footprint: no Wikipedia entry, no ADR, no analyst coverage a foreign investor can read.

What makes it a clean demonstration rather than just another deep dive is that every item in this article is a case of a screener and a filing disagreeing, and the filing winning. A screener reports a half-year net loss; the three-month column in the same document shows the company was profitable again by Q2. A screener reports 1,313,171 treasury shares; the filing says they were committed to bondholders eleven months ago. A screener reports a ₩130 dividend and a cancellation announcement; the filing says the dividend-bearing share count did not move by one share. A screener reports a 6.79% holder called Manulife IM (Hong Kong); the filing says five Manulife entities across four countries, the Hong Kong one holding 0.96%. A screener reports Samsung Asset Management building a stake from 5.01% to 8.11%; the filing says it was a net seller of 220,823 shares on the exchange while that happened. A screener cannot report a purpose the company only stated in a Korean AGM notice.

None of that is hidden. It is written in Korean, in a 227,000-character document, across four separate notes and three filings that nobody cross-reads. That is the whole business model of this site, and it is why we treat the Korea Discount as partly an information problem rather than only a governance one.

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 and market data as of the dates stated; prices, valuations and flows change and go stale quickly. Do your own research and consult a licensed financial adviser before making any investment decision.