Hyosung Heavy Industries (298040): The Order Book Has Two Sizes — and the Filing Says Why

Korea has three listed power-equipment makers and the world suddenly wants all of them. Hyosung Heavy Industries is the largest by market value, and after its second-quarter results the English coverage settled on one number for its power-equipment order book: ₩17.5 trillion, described as a record and as years of guaranteed work. The number is real and it is precisely reported. It is also not the number in the company’s own filing — the August 2026 half-year report puts the same order book, at the same date, 36% higher, and prints a one-line footnote explaining why. That footnote did not exist in the annual report. This is what it says, and what changes once you read it.

🧭 Why This Article Exists

Because we have been running the same test on Korean order books all year, and this one fails it in a way we have not seen before.

We asked Hanwha Aerospace’s $84.6bn order book not how big it was but how much profit it had produced, and found 27.7% of it sitting in a division earning 0.44% of group operating profit, on a VAT basis that changed between two consecutive filings.

Hyosung Heavy is the harder case, because here the problem is not a footnote that changed but a footnote that appeared. The FY2025 annual report says only that the heavy-industry order figures may not add up, because of currency translation and cancellations. The half-year report filed on 13 August 2026 adds a second sentence the annual report does not contain, and that sentence reframes every backlog number published about this company.

🏢 What Is Hyosung Heavy Industries, and What Does It Actually Sell?

Hyosung Heavy Industries (KRX 298040) builds the heavy iron of an electricity grid — ultra-high-voltage power transformers, circuit breakers, gas-insulated switchgear and industrial motors — and, in the same legal entity, it builds apartment blocks in Korea. Those two businesses are roughly two-thirds and one-third of revenue, and they share one balance sheet.

A power transformer is not the small device that charges a laptop. These are room-sized units that step voltage up for long-distance transmission and back down near the customer; the company’s Memphis plant builds them from 69kV to 765kV, with unit capacity up to 2,250 MVA. They take years to build, they are quoted years in advance, and there are only a handful of firms in the world that can make the largest of them. That is why a data-centre boom translates into a Korean industrial order book at all.

The legal company is young; the business is not. English Wikipedia dates the operation to 1962, when it was founded as HanYoung Industrial Co., Ltd. The DART filing dates the company to 4 June 2018, the registration date of the demerger that created it, with the shares listed on KOSPI on 13 July 2018. Both are correct and they answer different questions — a distinction worth getting used to, because this entire article turns on it.

The filing splits operations into two production units and a construction division. Power PU makes the ultra-high and medium-voltage equipment, and is pushing into HVDC, STATCOM, energy storage and AI-based grid asset management. Electro-Mechanical PU makes industrial motors, generators and gears, plus hydrogen refuelling stations, liquid-hydrogen plants and wind. Construction does housing, redevelopment, commercial buildings, civil works and public infrastructure. Group headcount was 3,567 at 31 December 2025, of whom 1,884 sat in the power division and 442 in construction, with average tenure of 13.2 years (Source: DART FY2025 annual report, rcpNo 20260323001004, section VIII-1).

Five entities do the power-equipment work, and their names matter for everything that follows:

Source: DART FY2025 annual report, rcpNo 20260323001004, section I-1 and section II-4, and DART half-year report rcpNo 20260813001554. The filings also consolidate ChinHung International (KRX 002780), a separately listed builder, inside the construction segment; 25 companies are consolidated in total. The Pittsburgh entity appears in filings as “HICO America Sales & Technology Inc.” and on its own website as “HICO America”.

The customer list is the first thing that does not look like the press coverage. In the first half of 2026 the parent company’s largest customers were HICO at 18.83%, Korea Electric Power (KEPCO) at 9.09%, then two Korean redevelopment associations at 4.95% and 3.95%, NSW Electricity at 3.54%, and Statnett — Norway’s transmission system operator — at 1.89%. A year earlier, across all of FY2025, HICO was 6.78% and KEPCO 3.40% (Source: rcpNo 20260813001554 and rcpNo 20260323001004, section II-1. The filing does not label this table’s basis, but HICO’s appearance in it implies the parent-company accounts, since a consolidated view would eliminate it).

HICO is not a customer in the ordinary sense. It is the group’s own US sales company, and its share of parent revenue nearly tripled. One basis caveat on that comparison, because the two figures are not the same length of period: 18.83% is a six-month concentration and 6.78% is a twelve-month one, and a single large shipment weighs more heavily in a half-year. The direction is not in doubt — no plausible seasonality turns 6.78% into 18.83% — but “nearly tripled” is 6 months against 12, not like against like. Hold that thought.

📊 Where Does the Revenue Actually Come From?

About two-thirds power equipment and one-third construction — 66.26% and 33.23% of consolidated revenue in the first half of 2026 — but the split in operating profit is nothing like that.

Basis: full year to 31 Dec 2025, consolidated, external (post-elimination) revenue. Won as printed: heavy industries ₩4,149,258m, construction ₩1,786,501m, other ₩32,753m, total ₩5,968,512m; operating profit ₩738,068m / ₩46,899m / ₩4,095m / adjustment −₩42,094m / total ₩746,968m. “% of segment OP” is our calculation, not a filed figure — we add the three segment operating-profit lines the filing prints and divide into that sum, because the consolidation adjustment cannot be attributed to a segment. The sum itself is not printed anywhere in the filing. Margins are operating profit over external revenue. (Source: DART rcpNo 20260323001004, consolidated note 6, “Operating segment information”.)

Construction took 30% of revenue and produced 5.9% of segment profit in FY2025. That is the cleanest single statement of what the two halves of this company are worth to a shareholder.

But we have to immediately argue against ourselves, because the first half of 2026 says something different. Construction’s operating margin rose from 2.62% in FY2025 to 7.02% in the six months to June 2026, and its share of segment operating profit rose to 12.6%. Over FY2024 it was 2.69% and 11.4%. Three periods, no trend: 11.4% → 5.9% → 12.6%. Anyone telling you Korean construction margins are structurally 2% is reading one year. What is stable is only the direction — construction earns a low-single-digit to mid-single-digit margin while power equipment earns a mid-teens to mid-twenties one.

Basis: six months, consolidated, revenue from contracts with customers, heavy-industries segment only. Won as printed, H1 2026: Korea ₩484,104m, Asia ₩529,263m, North & Central America ₩716,193m, Europe ₩155,702m, other ₩131,650m, total ₩2,016,912m. H1 2025: ₩596,787m / ₩495,502m / ₩383,501m / ₩180,885m / ₩134,864m / ₩1,791,539m. Construction revenue is 99.8% Korean in both periods and is excluded here. Both columns converted at the same ₩1,350.36/$1, so no part of the change is our FX. (Source: DART rcpNo 20260813001554, consolidated note, “Disaggregation of revenue from contracts with customers”.)

North America nearly doubled while Korea, the company’s home market, shrank 18.9%. The grid story is real and it is American. The home market is not carrying it.

🗓️ How Did Hyosung Heavy Get Here? A Short History

The short version: a 1962 industrial firm was carved out of a Korean conglomerate in 2018, bought an American transformer factory in 2019, and then the American grid started replacing itself.

Every row except the 1962 founding and the Memphis acquisition date comes from a Korean regulatory filing; those two are marked because they do not. The DART filing’s own “company history” section covers only the most recent five financial years, which is why the earlier rows need outside sources.

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

In large power transformers sold into the United States, Korea as a country is genuinely a top-two supplier — 18.6% of US imports in the first seven months of 2026, up from 13.1% in 2021 — but Hyosung’s own filings claim global leadership almost nowhere, and the one market share they do quantify is a domestic one.

This matters because “global leader” is the easiest phrase to repeat and the hardest to verify. Here is what is actually checkable.

What the filing claims. The FY2025 annual report says the company “leads the domestic market” in ultra-high-voltage transformers and breakers “based on a higher level of technology,” and that its quality is “recognised in global markets.” Those are company assertions with no number attached, and we are labelling them as such. The single quantified market share anywhere in the filing is for low-voltage industrial motors, where it states roughly 35% of a domestic market it sizes at about $370M (₩500bn) — a real number, and a small business (Source: rcpNo 20260323001004, section I-5, new business purposes).

What outside data says. US imports under HS 8504.23 — liquid-dielectric transformers rated above 10,000 kVA, the closest published line to “large power transformer” — show Korea moving from third supplier to second in four years, inside a market that itself quadrupled.

Basis: calendar years, US imports for consumption, CIF values. Source: UN Comtrade public API, reporter USA, commodity 850423, flow M — Comtrade’s mirror of US Census data, not a US-government site, and CIF is not the Census customs value. 2025 leaders were Mexico $803.0M, Korea $748.0M, Brazil $531.2M. ⚠️ Two cautions. First, this is Korea the country, not Hyosung the company — LS ELECTRIC and HD Hyundai Electric are in the same line. Second, and more important for this stock: output from Hyosung’s Memphis plant is US domestic production and never appears here at all. The trade data understates Hyosung’s US position, and as Memphis expands this line could flatten while Hyosung’s American revenue rises. Do not read a plateau as weakness.

And here is a fact published in English by the US government, absent from English coverage of this stock, and absent from Hyosung’s own filings: Hyosung Heavy pays an antidumping duty to sell large transformers into the United States from Korea. The order, case A-580-867, has been in force since 2012 and was continued after a second sunset review published on 5 September 2024. In the Final Results of Antidumping Duty Administrative Review; 2023–2024, published 14 August 2026 at 91 FR 52670 covering the period 1 August 2023 to 31 July 2024, Commerce assigned these weighted-average dumping margins:

Source: US Department of Commerce, Large Power Transformers From the Republic of Korea: Final Results of Antidumping Duty Administrative Review; 2023–2024, 91 FR 52670, 14 Aug 2026, read directly from the US Government Publishing Office. Hyosung and LS were non-selected respondents; Commerce applied the 4.32% rate to both by “other reasonable means.” ⚠️ This is a cash-deposit rate, not a permanent tax. The notice states the rates take effect on publication and “shall remain in effect until further notice” — these reviews are annual, so the figure can change at the next one, and it did change at this one: the preliminary results of 10 Feb 2026 had continued the prior rate for non-selected companies, and the 4.32% assignment was made at the final stage. The continuation of the order itself is at 89 FR 72373. There is no countervailing-duty order on Korean large power transformers — antidumping only, so do not write “AD/CVD”.

Now the part that is ours rather than Commerce’s. We searched both of Hyosung Heavy’s most recent periodic filings for any discussion of this. The word for antidumping — 반덤핑 — appears zero times in the FY2025 annual report and zero times in the half-year report filed on 13 August 2026 — the day before Commerce published the 4.32% rate. “Tariff” appears three times in each, all of it generic. A Korean investor reading only the Korean filings would not learn that the company carries a US dumping margin. That cuts against our usual finding, and it is worth saying plainly: this is a case where the English-language primary source has something the Korean one does not.

On the product itself, the company’s US subsidiary states it builds 69kV to 765kV units up to 2,250 MVA and is the only US producer of 765kV transformers, and Hyosung says it holds close to half the US 765kV market. We could not independently verify that share and are attributing it rather than asserting it.

📈 Is the Business Getting Better or Worse?

Better, and not marginally: operating margin went 7.41% → 12.52% → 13.68% across FY2024, FY2025 and the first half of 2026, and operating profit more than doubled in FY2025.

Basis: consolidated (K-IFRS). The H1 2026 column is six months, not annualised — do not compare it to a full year. Won as printed: revenue ₩4,300,572m / ₩4,894,980m / ₩5,968,512m / ₩3,045,187m; operating profit ₩257,837m / ₩362,479m / ₩746,968m / ₩416,609m; net profit attributable to owners ₩115,982m / ₩222,633m / ₩519,894m / ₩259,843m; and total equity as filed in single won: ₩1,224,159,409,651 / ₩2,055,618,494,514 / ₩2,489,720,957,283 / ₩2,728,789,773,671. FY2023 figures are the prior-year comparatives printed in the FY2025 filing. Unit note: the filings print every line in single won (FY2023 operating profit is ₩257,836,947,601); where we state millions, the last digits are our rounding, not a different figure. All four columns converted at the same ₩1,350.36/$1 so no part of the change is our FX. (Sources: DART rcpNo 20260323001004 and rcpNo 20260813001554, section III-1 summary financial information.)

Two things sit underneath that margin. The first is mix: heavy industries went from 59.91% of revenue in FY2023 to 69.52% in FY2025. The second is that the trend is not clean — the heavy-industries share fell back to 64.86% in the first quarter of 2026 and 66.26% in the first half. Construction did not shrink; power equipment simply stopped outrunning it for two quarters.

Customer cash is also arriving ahead of the work. Contract liabilities — money customers have already paid for equipment not yet delivered — rose to $1.48bn in advances alone by 30 June 2026, up 57.5% in six months (₩1,271,847m to ₩2,003,638m as filed). A backlog that arrives with prepayments is a materially better backlog than one that does not.

🔍 Why Does Hyosung’s Order Book Have Two Different Sizes?

Because the order book the company files is a simple sum of five companies that sell to each other, and the revenue the company reports is not. The filing says so in one sentence that appeared for the first time in August 2026.

First, the two published numbers, so you can check both. English coverage of the second-quarter results put the power-equipment backlog at ₩17.507tn at 30 June 2026 — The Elec, 4 August 2026, which prints the figure to three decimals alongside HD Hyundai Electric’s ₩12.14tn and LS ELECTRIC’s ₩6.9998tn; The Herald Business the same day wrote “approximately 17.5 trillion won”; Korea IT Times on 20 August called it “a record KRW 17.5 trillion.” The half-year report’s own order table, for the same date, prints ₩23,834,404m. That is 36.1% higher. Neither figure is wrong; they are struck on different bases, and only one of them comes with the basis attached.

For completeness: the higher figure has appeared in English once that we could find. A UPI story of 15 September 2026 reported “23.83 trillion won” and correctly attributed it to Hyosung Heavy “and its major heavy-industry subsidiaries” — which is the simple-sum basis, though it did not say that this makes the number non-comparable with the ₩17.5tn its own peers had been printing six weeks earlier. So the problem is not that English coverage picked the wrong number. It is that two numbers for one date are in circulation and nothing tells a reader why they differ. The filing does. Here is the sentence, from section II-4 of the half-year report, in full:

“The Heavy Industries segment’s order and revenue figures are a simple sum of the results of the controlling company and its major subsidiaries, and because intercompany transactions are not eliminated, they may differ from the consolidated figures.”

Our translation. Original: “중공업 사업부문의 수주 및 매출액은 지배회사 및 주요 종속회사의 실적을 단순 합산한 금액으로, 내부거래가 제거되지 않아 연결기준 금액과 차이가 있을 수 있습니다.” (DART rcpNo 20260813001554, filed 13 Aug 2026.) This sentence does not appear in the FY2025 annual report, which carries only a note about currency translation and cancellations.

Now the order-status table itself, exactly as filed:

Basis: six months to 30 Jun 2026, consolidated entities, intercompany NOT eliminated, translated by the company at the 30 Jun 2026 reporting-date rate. Won as printed (millions): heavy industries 15,936,349 + 11,010,478 − 2,924,567 = 23,834,404; construction (parent) 5,518,255 + 447,126 − 642,766 = 5,322,614; ChinHung 3,531,186 + 673,807 − 403,762 = 3,801,231. We show the “unexplained” column because the arithmetic does not close and we will not hide that — the two construction rows reconcile to within ₩1m (the parent row is out by exactly that), while the heavy-industries row misses by 0.78%. That gap is our subtraction, not a filed figure — the filing prints the four balances and leaves the difference implicit. The filing anticipates this: “the closing backlog may differ from the simple addition and subtraction of the opening backlog, new orders and revenue, because of order cancellations and changes in contract amounts.” It does not disclose how much of that roughly ₩188bn gap is cancellation and how much is currency. (Source: DART rcpNo 20260813001554, section II-4.)

Now the proof that the “simple sum” footnote matters, from a different part of the same document. The order table says heavy-industries revenue for the half was ₩2,924,567m. The consolidated segment note in the same filing says heavy industries had gross revenue of ₩2,938,968m — within 0.5% of the order table — of which ₩921,206m was intersegment and was eliminated, leaving external revenue of ₩2,017,762m.

Basis: consolidated segment note. FY columns are twelve months; the H1 2026 column is six months and is not annualised. Won as printed (millions): gross 3,800,297 / 5,239,937 / 2,938,968; intersegment 701,466 / 1,090,679 / 921,206; external 3,098,831 / 4,149,258 / 2,017,762. Each line subtracts exactly in won; the dollar columns are rounded to two decimals and will occasionally miss by $0.01. One basis note, because we quote two nearly-identical numbers for the same thing: the segment note’s H1 2026 external heavy-industries revenue is ₩2,017,762m, while the revenue-disaggregation note used for the regional table above totals ₩2,016,912m — a difference of ₩850m, or 0.04%. The two notes are drawn on different definitions: the regional table is revenue from contracts with customers under IFRS 15, the segment line is all external segment revenue. The filing does not reconcile them and neither will we; we are flagging the ₩850m so that a reader who adds up the regional table and compares it to the segment table is not left wondering which one we mistyped. (Sources: DART rcpNo 20260323001004 note 6 and rcpNo 20260813001554 note 6.)

Nearly a third of the heavy-industries division’s gross revenue in the first half of 2026 was one Hyosung company selling to another, and that proportion has risen by 12.8 percentage points in eighteen months. The order table is struck on the gross basis. The headline backlog therefore counts the same transformer twice: once when Changwon books an order from HICO America, and again when HICO America books it from the utility.

And the two largest power-equipment contracts Hyosung disclosed in 2026 are both sales to its own subsidiary. The filings say so outright.

We quote each filing’s own currency and do not re-convert it — every one of these disclosures prints its own foreign-currency amount and the exact exchange rate it used (USD at ₩1,470.3 and ₩1,346.40; NOK at ₩130.79 and ₩116.72; AUD at ₩1,055.85), so re-translating at any other rate would manufacture a number the company never filed. The 765kV filing states plainly: “this contract is a supply contract in which the company’s US sales subsidiary, HICO America Sales & Tech, Inc., won a project from a large US utility and then re-ordered it from the company.” The “percentage of revenue” each filing prints is not comparable across rows — they use whichever fiscal year was most recent at signing (FY2023, FY2024 or FY2025 revenue). (Sources, in row order: rcpNo 20260210800044, rcpNo 20260226801263, rcpNo 20260915800149, rcpNo 20260311802668, rcpNo 20260210801094.)

None of this makes the demand fake. A US utility really is buying 765kV transformers, and the Norwegian grid operator really has signed Hyosung through January 2032 — the company’s longest disclosed delivery date, and better evidence of a long runway than any ratio. What it means is narrower and more useful: the backlog number and the revenue number are not measured the same way, so dividing one by the other produces a figure that belongs to neither.

📅 So How Many Years of Work Is the Backlog Actually?

About four years, not five or six — 4.07 years if you divide the backlog by the revenue figure printed in the same table, and 5.91 years only if you divide it by a consolidated revenue figure measured on a different basis.

Our calculation, not a filed figure. Numerator is ₩23,834,404m throughout. Denominators (millions of won): 2,924,567×2 = 5,849,134; 2,938,968×2 = 5,877,936; 5,183,402; 2,017,762×2 = 4,035,524; 4,149,258. Annualising a half-year is itself an assumption — it presumes the second half looks like the first. We show all five rows rather than pick one, because the spread between the top and bottom row is 45% and it is entirely a definitional choice.

For what it is worth, the “five to six years” phrase in English coverage traces to a Business Korea article of 18 November 2025 where it describes four companies combined, not Hyosung, against an aggregate backlog of $24.47bn (₩33.04tn as the article prints it). A separate AJU Press piece of 27 April 2026 says the same four companies have secured “at least three years of work.” Three, five, six: when the same four companies produce three different answers in five months, the metric is soft, and the only honest response is to show the arithmetic and let the reader choose the denominator.

🏗️ What About the 27.7% of the Order Book That Is Apartments?

It is $6.76bn of the group’s $24.41bn order book — 27.7% — and its real risk is not the thin margin. It is a Korean contract structure called a completion undertaking, which turns a developer’s bank loan into Hyosung’s own debt if a building is late.

Start with the size. At 30 June 2026 the group carried completion undertakings on 31 project-finance construction sites with contract value of $4.39bn. The filing then discloses something most Western readers never see, because it is a Korean disclosure requirement: the maximum credit exposure attached to those undertakings.

Basis: consolidated, “maximum exposure to credit risk” as the note labels it — a ceiling, not an expected loss. Won as printed (millions): 30 Jun 2026 total 6,713,496 (redevelopment 1,655,054 + other 5,058,442); 31 Dec 2025 total 7,050,061 (2,633,025 + 4,417,036). Total equity as filed: ₩2,728,789,773,671 at 30 Jun 2026 and ₩2,489,720,957,283 at 31 Dec 2025. The multiple of equity is our calculation, as are both change percentages below. ⚠️ We have used the larger of two numbers in this note, and you should see the smaller one. The same table prints a third line — the credit exposure of the loan commitments and financial guarantees actually outstanding — at $3.37bn at 30 June 2026 (₩4,554,018m) against $3.60bn at 31 December (₩4,866,187m). That is 32% below the $4.97bn ceiling, and on it the multiple of equity is 1.67× rather than 2.46×. We lead with the ceiling because the note labels it “maximum exposure to credit risk” and because the ceiling is what a completion undertaking can actually cost, but choosing between the two lines is choosing a denominator, and the honest version shows both. Read the two sub-lines against each other, because the total hides them: redevelopment exposure fell 37.1% while other-project exposure rose 14.5%, so a headline “risk is falling” is only half true. The filing prints no change column, so every percentage in this caption is ours. (Source: DART rcpNo 20260813001554, consolidated note, “Completion undertakings related to PF projects”.)

Then, five weeks after that balance sheet date, one of them was called. On 31 July 2026 the board resolved to assume the project-finance debt of Jamwon-dong Complex Development PFV, because the completion deadline of that same day had arrived and the obligation had not been met. The amount is ₩346,300,000,000 — $256M, which the filing itself sizes at 13.91% of equity — and it is described as the unpaid principal and interest the borrower owed its lenders. The assumption took effect on 1 August 2026 (Source: DART rcpNo 20260731800934).

That date is the point. It falls after 30 June, so it is in none of the half-year figures quoted anywhere in this article or in the press. It will appear in the third-quarter report, due around mid-November 2026.

Construction orders also evaporate in a way transformer orders do not. On 2 January 2026 the company disclosed a ₩524,868,000,000 ($389M) contract to build apartments at Myeongjang Park in Busan — at 10.72% of FY2024 revenue, the largest single order disclosure of that period. The filing carried a condition: the contract could be cancelled if project financing was not arranged by March. On 17 April 2026 it was cancelled, the filing citing “the developer’s inability to satisfy the contract conditions.” It was one of four contract terminations Hyosung Heavy disclosed between January 2024 and September 2026 (Sources: rcpNo 20260102800499 and rcpNo 20260417800636).

⚙️ Can the Factories Even Build All of This?

Not faster than they already are. Hyosung’s power-equipment plants ran at 104.49% of stated capacity in the first half of 2026 — the third consecutive period above 100% — and the new plant that relieves it does not finish until July 2027.

Basis: Changwon plus Nantong, Pune and Memphis, combined. Capacity is measured in money, not units. Won as printed (millions): capacity 2,540,371 / 3,128,398 / 1,894,093; output 2,600,598 / 3,258,122 / 1,979,213. Output divided by capacity reproduces all three printed percentages exactly. The filing states its own method — capacity is “units producible × average shipment price per product” — and adds its own warning: “utilisation is simply production divided by capacity and will differ from the actual utilisation of individual entities and products.” Note also that capacity itself is growing fast: up 23.1% in FY2025 and running about 21% higher again in H1 2026, and output is still outrunning it. ⚠️ And read “104%” carefully, because the method allows a reading that has nothing to do with steel. Both the numerator and the denominator here are money, and the denominator is units producible × average shipment price. If units shipped match the plan but realised prices come in above the average price assumed in the capacity calculation, the ratio exceeds 100% with no extra unit built. In a market where transformer prices have been rising, that is not a remote possibility. So a figure above 100% is consistent with physically full plants, with better pricing, or with both, and the filing does not let you separate them — which is worth saying, since the scarcity story rests on the first reading. (Source: DART rcpNo 20260813001554, section II-3.)

This is the opposite of the pattern we found at LS ELECTRIC, whose extra-high-voltage lines run in the sixties and seventies. Here the filing backs the scarcity story — but in a way that cuts both ways: a plant at 104% cannot absorb a surge of new orders by working harder. The $8.15bn of new heavy-industries orders booked in the first half of 2026 — more, in six months, than the $7.31bn booked in all of FY2025 — has to wait for steel and buildings.

That is what the July 2025 board resolution is for: a ₩253,800,000,000 ($188M) dedicated HVDC transformer plant on the Changwon site, 12.35% of equity at the time, running from 30 July 2025 to a scheduled completion of 31 July 2027 (Source: DART rcpNo 20250730800037). Separately, on the US side, the group says it has put a combined $300M into Memphis across three expansion rounds and plans to lift capacity there more than 50% by 2028, and has a joint venture with Quanta Services in Pennsylvania to build ultra-high-voltage circuit breakers from October 2026 — company statements, dated 2 September 2026, which we are attributing rather than verifying.

Research spending, for context, is modest: $40M in FY2025 (0.90% of consolidated revenue) and $23M in the first half of 2026 (1.03%). The company also notes that its R&D organisation is split between Hyosung Corporation’s Heavy Industries Research Institute and Hyosung Heavy’s own team — meaning part of the research function for this listed company sits inside its holding-company parent.

💰 What Do You Actually Pay for Hyosung Heavy Today?

₩2,901,000 a share — about $2,148 — for a company worth just over $20bn, trading at 52.0× trailing earnings and 10.5× book, up 62.9% this year and still 37.0% below the all-time high it set in May.

All market figures are the settled Korea Exchange close of 22 September 2026, converted at ₩1,350.36/$1 like everything else in this article. ⚠️ Why 22 September, when 23 September was also a trading day: KRX has not published per-stock settled data for 23 September, and our two internal price feeds disagree on that date for 2,026 of the 2,555 stocks we track. Rather than quote a figure two sources dispute, we anchor to the last fully settled session. Korean markets were then closed for Chuseok. P/E uses the EPS the company prints (₩55,832, basic, owners’ interest); P/B uses owners’ equity of ₩2,587,324m at 30 June 2026 and is our calculation. Dividend yield is ₩7,500 over the 22 September close — note that DART’s own FY2025 figure of 0.40% is computed against a 2025 price base and is a different measurement, not a contradiction. ⚠️ Prices go stale. Check a live quote before acting on any of this.

Basis: market data at the settled KRX close of 22 Sep 2026; FY2025 revenue and margin from each company’s own consolidated annual filing. Everything is converted at the one article-wide rate of ₩1,350.36/$1. P/E is the 22 September close divided by each company’s own filed FY2025 basic earnings per share — ₩55,832 for Hyosung Heavy, ₩20,354 for HD Hyundai Electric, and ₩1,929 for LS ELECTRIC, which is our split adjustment: LS files ₩9,647, struck before the 1-for-5 split, so it has to be divided by five to sit against a post-split price. P/B is market cap over owners’ equity at 30 June 2026. Both ratios are our calculations. ⚠️ A different and equally defensible method gives different numbers, so here are those too: market capitalisation over consolidated net profit (the whole group’s, including the part that belongs to minority shareholders) gives 53.8× for Hyosung, 35.2× for HD Hyundai Electric and 111.9× for LS ELECTRIC. The difference is non-controlling interests, not a contradiction — and it is largest at Hyosung because Hyosung’s minority interests lost ₩17.1bn in FY2025 against ₩0.8bn at HD Hyundai Electric and ₩2.4bn at LS ELECTRIC, so its consolidated profit is below its owners’ profit. We lead with price-over-filed-EPS because it is the figure a reader can reproduce from one line of one filing. Won as printed: revenue ₩5,968,512m / ₩4,079,498m / ₩4,965,799m; operating profit ₩746,968m / ₩995,313m / ₩426,407m. ⚠️ LS ELECTRIC split its shares 1-for-5 on 13 April 2026, so any LS per-share figure sourced from a pre-April article or an unadjusted database is wrong by a factor of five; the market-cap and margin figures here are split-neutral. The bottom row is not a listed company — it is Hyosung’s heavy-industries segment on its own, shown because it is the only like-for-like comparison with the two pure-play peers.

The bottom row is the whole valuation argument in one line. Hyosung has the largest revenue of the three and the second-worst group margin — because a third of it is a construction company. Strip the construction out and its power-equipment margin of 17.79% sits neatly between LS ELECTRIC’s 8.59% and HD Hyundai Electric’s 24.40%. An investor buying 298040 for the grid is paying for a mid-pack transformer business and receiving, attached to it, a builder with $4.97bn of completion exposure.

Which is also why the cheapest-looking multiple here is not obviously a bargain. Hyosung trades at 52.0× earnings against HD Hyundai Electric’s 35.1× despite earning half the margin; LS ELECTRIC’s 109.9× is the outlier in the other direction — and we have taken each of those two apart in its own filings, because a peer multiple you have not checked is just a rumour with a decimal point. On book value all three are between 10× and 15×, which tells you these are being priced on the order book, not the balance sheet.

🏷️ Which Hyosung Ticker Are You Actually Buying?

Only 298040 is the transformer maker. There are seven other listed companies with “Hyosung” in the name, two of them are holding companies, one is not part of the group at all, and one changed its name in 2024 — so any English article written before then will point you at the wrong entity.

The 2018 demerger split one company into five, and a second demerger in 2024 split the parent again. The FY2025 annual report prints the original arithmetic: of Hyosung Corporation’s 35,117,455 pre-split shares, Hyosung Corporation kept 13,795,151 and Hyosung Heavy Industries received 9,324,548 — the share count it still has.

Ticker codes and Korean names verified against the KRX listed-company universe as at 27 Sep 2026; the 2018 split arithmetic from DART rcpNo 20260323001004, section I-3, and the 2024 HS Hyosung demerger from the same filing, section VII-1. There is no ADR and no preferred share for 298040 — the authorised capital allows 50,000,000 preference shares, and none have ever been issued. For how to reach KRX-listed shares from outside Korea and how to read Korean prices in dollars, see our guide to Korean share prices in USD. 298040 is a KOSPI 200 constituent, so index funds tracking that benchmark hold it.

Korea does not require the low nominal share prices US investors are used to, which is how a single share comes to cost ₩2,901,000. Hyosung Heavy has never split its stock, never issued a bonus share, never done a rights issue and never bought back a share: the share count has been 9,324,548 every single day since listing in 2018.

🌏 Is Foreign Money Buying or Selling Hyosung Heavy?

Foreign investors bought in both windows we measured. Everything else — institutions, retail, and the share price itself — changes sign depending on which window you pick, which is why we will not give you only one.

Method, stated because it changes the answer: investor flows are recorded as net shares, so every won and dollar figure here is derived. We multiply each day's net shares by that day's settled close and sum — an approximation, not execution prices. Windows are the 20 sessions from 26 Aug to 22 Sep 2026 and the 60 sessions from 29 Jun to 22 Sep 2026, both ending at the last fully settled KRX session. No flow rows are missing from either window. Won equivalents: foreign +₩24.71bn / +₩75.00bn; institution +₩7.17bn / −₩79.43bn; individual −₩34.58bn / +₩5.99bn, converted at ₩1,350.36/$1. Price change is computed from market capitalisation, not the raw close, and is measured from the session immediately before each window opens (25 Aug and 26 Jun), which is our house convention. Measured instead from the first close inside each window, the same two figures are +3.72% and −11.39% — the 20-session gain shrinks and the 60-session fall deepens, so nothing in the argument below depends on which convention you prefer. Source: settled Korea Exchange closes and our own daily investor-flow database.

Three of the four rows reverse. Over twenty sessions this reads as a recovering stock that institutions and foreigners were both buying while retail sold into it; over sixty it reads as a stock down 6.5% that institutions sold nearly $59M of while retail quietly accumulated. Both are arithmetically correct and they describe different companies. An article publishing only the 20-session column would have called institutions accumulators; the 60-session column makes them the largest sellers on the board.

What survives both windows is narrow and real: foreign investors were net buyers throughout, and foreign ownership rose from 26.87% on 22 June to 27.29% on 22 September. We are not going to tell you that buying caused anything. Over the longer window foreigners bought and the stock fell; over the shorter one they bought and it rose. That is the ordinary case, and it is why the price row sits in the same table as the flow rows. For how these investor categories are defined and what they can and cannot tell you, see our guide to foreign and institutional flows in Korean stocks, and for the weekly version across the whole market, the Foreign Flow Watch series.

One more reason not to over-read any of it: the flows are small relative to the stock's own swings. Hyosung Heavy is up 62.9% year to date and down 37.0% from the all-time high it set on 7 May 2026, when its market value was $31.77bn against $20.03bn now. About $11.74bn of market value has come off in four and a half months — roughly 200 times the 60-session foreign net purchase. Whatever is moving this price, it is not these flows.

🏛️ Who Controls Hyosung Heavy, and What Do Minority Holders Get?

The Cho family controls it through Hyosung Corporation with 43.96% between them, the National Pension Service holds 10.26%, and minority holders get a 0.26% dividend, no buyback, and one genuine victory at the 2026 annual meeting.

Basis: register closed 31 Dec 2025. Controlling-group percentages are against total issued shares; the NPS and small-shareholder lines are as the filing states them. The “nine other related parties” row aggregates nine individually disclosed holdings of 0.05% or less and is our sum. (Source: DART rcpNo 20260323001004, section VII-1 and VII-3.) Note that Hyosung Corporation is itself 41.02% owned by Cho Hyun-joon, a stake that rose from 33.03% in May 2024 on the inheritance of the late honorary chairman Cho Seok-rae’s bequeathed shares.

The chairman sold 456,903 shares during 2025 and landed at 9.99%. The filing records the method as off-hours block trades and gives no reason. We do not know the motive and the filing does not supply one.

And we are not going to tell you the stake was parked just below a legal threshold, because on the basis the law uses it is not. 9.99% is 931,320 shares against the total issued 9,324,548. The same annual report’s “shareholders of 5% or more” table prints the identical holding as 10.00%, footnoting its basis: “percentages are on the basis of the number of issued shares carrying voting rights.” That base excludes the 12,807 treasury shares, giving 9,311,741 — and 931,320 of 9,311,741 is 10.0016%. The “major shareholder” test in Article 9(1)2 of the Financial Investment Services and Capital Markets Act runs on voting shares, so on the statutory base the stake is fractionally above 10%. One report prints both numbers for the same holding, two sections apart, and they answer the obvious question in opposite directions. What is not in doubt is the effect on control: none. Hyosung Corporation’s 32.47% did not move, and the family retains 43.96% between them.

Now the governance event, which is genuinely unusual and exists nowhere in English. At the annual meeting on 19 March 2026, the board put forward two articles-of-association amendments. The first — deleting the clause that excludes cumulative voting — passed with 99.9% of votes cast. The second, agenda item 2-2, bundled together a change to board size, new director qualification requirements, a director-term change, the renaming of “outside directors” to “independent directors,” and an increase from one to two in the number of audit-committee members who must be elected separately from the rest of the board.

Item 2-2 was defeated. It needed a two-thirds special majority and received 58.8% of votes cast, with 41.2% against or abstaining. The consequence is spelled out in the filing: because 2-2 failed, the term in item 4-1 had to be cut from three years to two, and item 4-2 — the second separately-elected audit committee member — was never put to a vote at all (Source: DART rcpNo 20260319801907).

Ten weeks later the company came back. At an extraordinary general meeting on 2 June 2026 the same articles amendment passed with 99.9% of votes cast, and Park Jong-bae, a Konkuk University electrical-engineering professor, was elected as the separately-elected audit-committee director with 90.6% of the shares voted — and 65.7% of all voting shares outstanding, the filing marking that item “3% rule applied” (Source: DART rcpNo 20260602800611).

Those two percentages are not evidence of the 3% rule, and the reading that they are is an easy one to fall into. Every Korean meeting-result filing prints both columns because they have different denominators: approval as a share of all voting stock on issue, and approval as a share of the stock actually voted. The gap is mostly turnout — shares nobody voted — and the proof is in the same filings. At the 19 March annual meeting, approval of the financial statements, an ordinary resolution with no 3% cap anywhere near it, came in at 81.1% of voting shares against 98.1% of shares voted: a 17-point gap of exactly the same shape. At the June meeting the articles amendment showed 84.5% against 99.9%.

What the 3% rule does is cap each holder at 3% when an audit-committee director is elected, so the controlling block’s 32.47% counts as 3% and the National Pension Service’s 10.26% counts as 3% too. That shrinks the shares eligible to be counted and widens the gap further on those items — and it is the mechanism that stops a 43.96% family block from choosing its own auditors. But it is not what creates the two columns, and a reader handed only “90.6% versus 65.7%” would misjudge how much support the director actually had.

We want to be careful about what that sequence proves. It does not prove minority shareholders were blocking a governance improvement, because item 2-2 bundled five unrelated changes into one vote and the filing never says which one drew the opposition — bundling is itself the governance problem. What it does show is a board that lost a vote and reconvened to win it, and an audit committee that now has two independently elected members instead of one, on a timetable driven by Korea’s amended Commercial Act, effective 23 July 2026.

On shareholder returns the record is thin and easy to state. The FY2025 dividend was ₩7,500 a share ($5.56), $51.7M in total — a payout ratio the filing prints as 13.40% and a yield of 0.4% on the 2025 price base, 0.26% on today’s. Beyond that: no share buyback, no cancellation, and no Value-Up filing. We read every one of the 263 filings Hyosung Heavy made to DART of every type between 1 January 2024 and 27 September 2026, and found zero treasury-share purchases, zero cancellations, zero bonus issues, zero rights issues by the company itself, and zero corporate value-enhancement plans. The same search correctly returned the three corporate-governance reports and all 34 supply-contract filings we already knew existed, so the test is working rather than silently finding nothing. The only treasury stock the company holds is 12,807 shares, or 0.14%, and those are the fractional shares created when it was listed in 2018.

Board composition, as at the 19 March 2026 meeting: seven directors, four of them outside directors (57.1%), with three audit-committee members. The chief executive, Woo Tae-hee, also serves as chairman of the board — there is no independent chair. Chairman Cho Hyun-joon, who joined the board in March 2025, attended 75% of FY2025 board meetings; every other director attended 100%. Five sub-committees operate, two of them — ESG and compensation — created only in February 2025. This is the kind of detail that sits behind the phrase “Korea discount” and almost never makes it into an English research note.

⚠️ The Bear Case

Two further items belong on the record without being investment arguments. On 8 May 2026 a 200-tonne mobile crane overturned on a road in Pyeongchang near the Gangwon wind-repowering site and a crane operator was killed; the filing notes the worker was not a Hyosung employee but that the incident was reportable, and that the cause was under investigation (rcpNo 20260508801298). And Gangwon Wind Power was Hyosung Heavy’s fourth-largest customer in FY2025 at 3.89% of parent revenue.

📚 Lingo Check

Definitions follow our house editorial glossary; the terms new to it here are order backlog, intercompany elimination, gross vs external revenue, completion undertaking, project financing, horizontal demerger, the 3% rule, separately elected audit-committee director, rumour-response disclosure and HVDC.

🎯 Why It Matters for K-Export Stars

Because the single most quoted number about this company is a number nobody has checked against the filing that produces it.

Everything in this article was public on the day it was filed: the non-elimination footnote, the three values for one date’s backlog, the 41.2% vote against the governance amendment, the antidumping rate, the $256M debt assumption, the thirteen identical answers to a five-year-old question. All of it sits in Korean, in documents running to several megabytes, and none of it needs a source, a call or a subscription. It needs reading the filing instead of the press release about the filing.

That is the whole of our edge and we would rather state it plainly than dress it up. Reconciling a segment order table against a consolidated segment note against thirty-four individual contract disclosures is not a job anyone can do on a news deadline. We are not faster than Yonhap and never will be. We are reading a different document.

If you follow Korean order books, the specific transferable lesson here is a question to ask of any of them: is the backlog measured the same way as the revenue you are about to divide it by? We found the answer was no at Hanwha Aerospace, where the VAT basis changed between filings, and no again here, for an entirely different reason. Two for two is not a coincidence; it is what an unstandardised disclosure looks like.

Disclaimer: This article is for informational and educational purposes only and is not investment advice, a recommendation, or an offer to buy or sell any security. Figures are drawn from public regulatory filings and settled exchange data on the dates stated and will go out of date; share prices, valuation multiples and flow figures were current at the settled Korea Exchange close of 22 September 2026. Currency conversions are our own at the rates stated. Where a figure is a company assertion rather than an independently verifiable fact, we have labelled it. Do your own research and consider your own circumstances before making any investment decision. K-Export Stars publishes a corrections policy, and we will correct any error in this article on the record.