BHI Co. (083650): Korea’s HRSG and Boiler Maker, Decoded

Every Korean nuclear name has run this year, and BHI has run with them. But BHI does not file like a nuclear company. Its half-year report has no nuclear segment, no nuclear product line and no nuclear revenue disclosure — because on the company’s own numbers, the nuclear work is roughly one dollar in twenty-five. What actually moved BHI’s 2026 accounts was a currency hedge, and almost nobody outside Korea has read it.

🏢 What Is BHI Co., and What Does It Actually Make?

BHI is a Korean maker of the large steel heat equipment that sits behind a power plant’s turbine — heat-recovery steam generators, boilers, condensers, heat exchangers and deaerators — built to order, one project at a time, mostly for export.

The company describes itself in its own words as “a power-plant equipment specialist that designs, fabricates, installs and constructs the equipment required for power-generation and steelmaking processes,” with customers that include KEPCO and the Korean generating companies, domestic and international EPC contractors, and steelmakers including POSCO. (Source: DART half-year report, rcpNo 20260814000764, filed Aug 14, 2026)

The distinction that matters for a foreign investor is between prime equipment and balance of plant. BHI’s filing sets it out plainly: the prime mover is determined by the fuel — a reactor for nuclear, a boiler for coal, a gas turbine plus an HRSG for combined cycle — while the balance of plant is broadly common across plant types. BHI builds the boilers and the HRSGs, which are prime equipment in coal and gas plants, and it builds condensers and heat exchangers, which are balance of plant everywhere.

It does not build reactors. That distinction is not a quibble — it is the entire investment case, and we return to it below.

The basics, all from the same filing: incorporated June 12, 1998 to make heat exchangers and power equipment; renamed BHI Co., Ltd. on March 14, 2009; listed on KOSDAQ on December 2, 2005. Head office and two of its three plants are in Haman County, South Gyeongsang Province, with a third at Sacheon. Headcount at June 30, 2026 was 763, of whom 249 were on fixed-term contracts, with average tenure of 5.6 years. It is classified as a mid-sized enterprise, not an SME.

📊 Where Does BHI’s Revenue Actually Come From?

Two-thirds of it comes from HRSGs, three-quarters of it is export, and the product line that touches nuclear plants at all — balance of plant — was 4.7% of first-half revenue.

The half-year report breaks revenue into exactly four product lines. There is no nuclear line, no renewables line and no services line. This is the table, as filed, for the six months to June 30, 2026 and the two prior full years.

1H 2026 is the cumulative six months to June 30, 2026 (제29기 2분기 누적), not the second quarter alone. Filed figures in won: HRSG ₩377,362m / ₩605,913m / ₩259,014m; Boiler ₩124,742m / ₩97,227m / ₩68,413m; B.O.P ₩27,076m / ₩26,113m / ₩14,305m; Other ₩47,886m / ₩44,864m / ₩63,010m; total ₩577,066m / ₩774,117m / ₩404,742m. (Source: DART rcpNo 20260814000764, section II-4-가; FY2024 column cross-checked against DART rcpNo 20260320000821.) Converted at ₩1,345.06/$1 throughout this article — see the note in the valuation section.

The export split is just as stark. Of first-half revenue, 75.4% was export and 24.6% domestic, and within HRSG specifically 84.6% was export. Everything is sold direct — the filing states that 100% of sales go through direct purchase orders, with no distributors, supported by overseas subsidiaries in Indonesia, Thailand, Israel, Japan and the United States plus agents in Europe, the Middle East, Asia and Africa.

That export share is the setup for the currency problem we come to later. A company that books three-quarters of its revenue in dollars, on contracts running four to seven years, has to decide what to do about the won. BHI decided, and in the first half of 2026 the decision cost it more than the operating business earned in improvement.

🗓️ How Did a Heat-Exchanger Shop Become a Power-Plant Contractor?

By buying the technology it had been licensing, and then moving one step up the chain from supplying equipment to building the plant.

BHI’s registered corporate history in DART covers only board changes over five years, so the operating history has to be assembled from the business sections of the filings. It is a clear arc.

(Sources: DART FY2025 annual report, rcpNo 20260320000821, sections I-2, I-3, I-5 and II-1; DART half-year report, rcpNo 20260814000764, section II-6-가, which prints the purchase price in the filing’s own units as 410억원 (₩41.0bn), and section III-8-가-2), which prints the balance-sheet effect.)

Two of those rows do most of the work. The 2017 and 2020 technology acquisitions turned BHI from a licensee into an owner of its core designs, which is why it can now bid HRSG packages directly against Japanese and European majors. The 2023 articles amendment and the 2025 EPC division turned it from a component vendor into a contractor that will take an entire power block.

The company is candid about the trade. In the mandatory disclosure explaining why it added construction to its articles, it wrote that the new business “will increase the degree of exposure to risk, because the scale, scope of work and level of responsibility are larger than the existing business.” That is BHI describing its own balance sheet three years before it reached a 444.6% debt-to-equity ratio.

That convertible bond deserves a footnote of its own. It was issued in February 2022 for working capital — the annual report records the proceeds as ₩20,000m, about $14.9M, used in full for working capital — and converted on July 23, 2023 into 4,944,375 shares, which is 16.0% of the 30,944,375 shares outstanding today. The filings do not print the conversion price; dividing the proceeds by the shares issued implies roughly ₩4,045 a share. Against the September 7, 2026 close, that is a sixteen-fold move for whoever held it.

🌍 Is BHI a Global Leader in Anything?

BHI declines to state a market share at all, and says so in writing — the only competitive-position claim it makes is that Korean balance-of-plant supply has historically been a three-way contest between itself, Doosan and Hyundai Heavy Industries.

This is worth quoting, because most Korean industrials do put a share number in. BHI’s annual report says: “The world power-generation market is conducted locally, country by country, and objective statistics are difficult to obtain. Because the nature of the industry makes any reasonable estimate of market share difficult, this disclosure has been omitted.” (Source: DART rcpNo 20260320000821, section II-3)

What it does claim is structural. On balance of plant, the filing says the domestic market “has been contested by our company, Doosan Heavy Industries (now Doosan Enerbility) and Hyundai Heavy Industries,” that the expansion of the nuclear market drew in new entrants, that those entrants “led unreasonable price competition to make up for a lack of track record and technology,” and that several have since failed or exited. On HRSG, it says competition intensified as many firms entered via technology tie-ups, but that industry restructuring has “somewhat eased” the intensity.

Treat all of that as company-reported. It is an assertion about competitive structure, not an audited number, and it is the assertion of a company that benefits from the reader believing it.

What is independently checkable is the customer list, and it is a genuinely international one. The half-year report’s progress-basis order table names Toshiba Plant Systems, Mitsubishi Heavy Industries, IHI, Fuji Electric, Siemens, Harbin Electric International, SEPCO 3, Elsewedy Electric, Formosa Heavy Industries, Alghanim, and the Korean majors Samsung C&T, Hyundai E&C, Daewoo E&C, POSCO E&C and Doosan Enerbility. A Korean small cap that sells HRSGs to Toshiba, Mitsubishi and IHI — in their home market — is not a domestic-only business.

⚛️ How Much of BHI’s Order Book Is Actually Nuclear?

Between 3.4% and 4.7% depending on which of the company’s own three tables you use — and every won of it is one domestic project, Shin-Hanul Units 3 and 4, where BHI supplies containment steel plate, a stainless-steel liner and a condenser.

This is the part that no English-language source has put together, so it is worth being precise about method. BHI publishes three separate views of its order book, in two filings. We counted all three. They land in a narrow band — but they do not print the same numbers, and that is worth showing rather than smoothing.

Cut one — the progress-basis order table. The half-year report’s note on percentage-of-completion contracts lists every project whose contract value exceeds 5% of the prior year’s revenue: 43 projects, with customer, contract date, deadline, total order value and percent complete. Three of the 43 are with Korea Hydro & Nuclear Power, and the filing prints their order values as ₩59,747,467,700, ₩42,452,691,457 and ₩47,950,266,329. (Source: DART rcpNo 20260814000764, section III-8-라, as of June 30, 2026)

Summing the filed rows ourselves, those three come to ₩150.15bn against ₩4,366.38bn for all 43 — 3.44%. By product category the same 43 projects split HRSG 72.59%, balance of plant 12.42%, boiler 10.27% and EPC 4.71%. These are our additions of the filing’s rows, not figures the filing prints.

Cut two — the single-supply contract table. A separate section of the same filing lists the 30 disclosed single-supply contracts still running. Three are labelled “nuclear power equipment supply contract,” and the filing prints their confirmed values, in millions of won, as 43,931, 62,300 and 47,950. (Source: DART rcpNo 20260814000764, section XI-1, as of June 30, 2026)

Adding the 30 rows gives ₩3,292.96bn, so the nuclear three are 4.68% of that table — again our arithmetic on the filing’s rows.

Two of those three values do not match the progress-basis table, and the filing does not reconcile them. Pair the rows by completion deadline and the gap is visible: the contract due 2029-06-30 is ₩62,300m in the contract table against ₩59,747,467,700 in the progress table; the one due 2029-07-31 is ₩43,931m against ₩42,452,691,457; the third, also due 2029-07-31, is ₩47,950m in both. The two tables also date the same contracts differently — the contract table by disclosure date (2025-02-11, 2025-02-11, 2025-06-05), the progress table by contract date (2024-08-22, 2024-11-15, 2025-05-21). We can show you the discrepancy; we cannot explain it from the filing, so we will not invent a reason. In aggregate the two tables land 2.6% apart, which is why the percentages below are a band and not a point.

Cut three — new orders won in FY2025. The annual report reports order intake by product: HRSG ₩1,149,570m, boiler ₩522,282m, B.O.P ₩70,422m and other ₩28,008m, totalling ₩1,770,282m, up 19% on FY2024. (Source: DART rcpNo 20260320000821, section IV-3)

Balance of plant — the only line that can contain nuclear work at all — was 3.98% of the year’s new orders. HRSG was 64.9%.

Three independent measures, 3.44%, 4.68% and 3.98%. The nuclear exposure is real but it is small, and it has not been growing as a share of the book.

What is the nuclear work, exactly?

It is heavy steel fabrication and a condenser at one domestic project — the containment steel plate, the stainless-steel liner and the condenser package for Shin-Hanul Units 3 and 4. The filings get specific in a way that no summary conveys: the same three KHNP contracts appear in the project-level note under their actual project names.

Project names from section III-8-라 and the revenue-from-contracts note; customer, value, deadline and progress from the progress-basis order table, all as of June 30, 2026. The filing prints the order values in won as ₩59,747,467,700, ₩42,452,691,457 and ₩47,950,266,329; the dollar column converts those at ₩1,345.06/$1 and the total is our addition of the three filed rows. (Source: DART rcpNo 20260814000764, section III-8-라.)

Containment steel plate. A stainless liner. A condenser. This is heavy steel fabrication for the building and the cooling end of the plant — not the reactor, not the steam supply system, not fuel. Doosan Enerbility is the company that forges Korea’s nuclear steam supply systems; the design work sits with KEPCO E&C; the reactor owner and orderer is Korea Hydro & Nuclear Power, which has no stock code and cannot be bought. BHI sits further out than any of them.

One more detail is telling. In its own progress-basis order table, BHI files all three Shin-Hanul contracts under the product category “HRSG.” Not a nuclear category — there isn’t one. The company’s product accounting has four lines and nuclear is not one of them.

For completeness on the language question: the word 원자력 (“nuclear energy”) appears 24 times in the 222,607-character half-year report, of which 12 are the customer’s corporate name, 한국수력원자력. The ordinary Korean word for a nuclear power station, 원전, appears zero times. So does SMR. That is not a gotcha — it is simply what a company writes when nuclear is 3–5% of its book.

📈 Is the Business Getting Better or Worse?

The operating business is improving fast — revenue has more than doubled in two years and the operating margin has gone from 4.1% to 14.0% — while the balance sheet is getting worse, with debt-to-equity at 444.6% and unbilled receivables at 63.7% of half-year revenue.

Full-year figures as filed in won: revenue ₩367,399,208,640 / ₩404,741,038,609 / ₩774,116,878,915; operating profit ₩15,086,698,166 / ₩21,927,258,374 / ₩75,496,618,005; net profit ₩7,463,782,053 / ₩19,585,052,430 / ₩65,196,402,061 (DART rcpNo 20260320000821). Half-year figures as filed, cumulative six-month column: revenue ₩577,066,229,772 vs ₩304,573,841,433; operating profit ₩80,807,667,784 vs ₩32,832,792,893; net profit ₩25,417,360,229 vs ₩38,568,354,610 (DART rcpNo 20260814000764). Debt-to-equity is the ratio the company itself publishes; FY2023 computed from filed liabilities ₩351,747m and equity ₩73,703m.

The order book supports the growth. At June 30, 2026 BHI reported 69 projects with a total contracted value of $2,835M, of which $1,008M had been delivered, leaving a backlog of $1,827M — roughly 3.2 years of FY2025 revenue. Overseas is 69% of the backlog. (Filed in won: domestic 41 projects, ₩1,498,432m total / ₩740,714m delivered / ₩757,718m backlog; overseas 28 projects, ₩2,314,165m / ₩614,997m / ₩1,699,168m; total 69, ₩3,812,597m / ₩1,355,711m / ₩2,456,886m — DART rcpNo 20260814000764, section II-4-다.)

One caveat on that backlog that an English reader will not get anywhere else: the filing’s own footnote states that the figure “includes amounts under MOUs, LOIs and similar.” A Korean backlog line labelled 수주잔고 is not necessarily a book of signed contracts. This one is not.

Two other things are deteriorating quietly. Unbilled construction receivables — revenue recognised on percentage of completion but not yet invoiced — rose from $161M at the end of FY2025 to $273M at the half, equal to 63.7% of half-year revenue. Both figures are the consolidated balance-sheet line, so the comparison is like for like: ₩216,792,588,075 at December 31, 2025 and ₩367,332,946,502 at June 30, 2026. (Source: DART rcpNo 20260814000764, consolidated statement of financial position.) And the auditor’s key audit matters in each of FY2023, FY2024 and FY2025 — the 26th, 27th and 28th fiscal years, three consecutive unqualified opinions from Samjong — were exactly the same two estimates that produce those numbers: “uncertainty in estimated total contract costs” and “appropriateness of the calculated percentage of completion.” BHI’s auditor for FY2026 changed from Samjong (KPMG) to Seohyun.

💱 Why Did Operating Profit Rise 146% While Net Profit Fell 34%?

Because of a currency-forward book with more notional value than a year of revenue: BHI had sold $526.5M forward at an average of ₩1,461.07 when the won closed the half at ₩1,541.73, and the resulting mark-to-market loss accounts for the entire non-operating deficit.

This is the most consequential thing in BHI’s 2026 filings and it has nothing to do with reactors. Start with the arithmetic, which is exact.

Derivative components as filed, in thousands of won. 1H26: valuation gain 0, valuation loss 39,566,300, transaction gain 330,030, transaction loss 16,552,546. 1H25: valuation gain 17,499,569, valuation loss 1,217,699, transaction gain 7,324,844, transaction loss 349,128. Total non-operating result as filed: −₩47,194,160,965 (1H26) and +₩15,182,218,421 (1H25). (Source: DART rcpNo 20260814000764, section II-5-바 and the consolidated statement of comprehensive income.)

Netting those four components ourselves gives −₩55.79bn for 1H26 and +₩23.26bn for 1H25. The half-year report prints the components, not the net — but BHI’s separate derivative-loss disclosure of the same day prints the 1H26 net to the won as −₩55,788,815,998, which is the same figure. The 1H25 netting is ours.

On the prior year end, read the whole book. At December 31, 2025 the dollar leg of the forward book — $438,197,500 struck at an average ₩1,431.18 — was marked at +₩2,900,207 thousand, but the euro leg, €20,000,000 struck at ₩1,428.30, was marked at −₩3,107,202 thousand. The filing’s own total line is −₩206,996 thousand. The 2025 book was flat to slightly negative, not ahead. We flag this because the dollar leg alone is the number that would flatter the comparison, and it is not the book. (Source: DART rcpNo 20260320000821, derivatives note.)

Subtract the derivative line from the total non-operating result and everything else — interest, other income, other expense, the lot — nets to a positive $6.4M. The derivative book is not part of the explanation for the profit fall. It is the whole of it. Year on year the swing was $58.8M, against a pre-tax profit decline of $10.7M.

Why did it happen? BHI sells in dollars on multi-year contracts and hedges the receipts by selling dollars forward. At June 30, 2026 it had 38 open contracts: $526,462,500 sold forward at an average agreed rate of ₩1,461.07, plus €10,000,000, for a total contracted amount of ₩783,487,018 thousand and a fair value of −₩39,566,300 thousand. (Source: DART rcpNo 20260814000764, section II-5-바.)

That forward notional is larger than BHI’s entire FY2025 revenue. And on the balance-sheet date the won was at ₩1,541.73 to the dollar — 5.5% weaker than the average rate at which BHI had agreed to sell. A company that has promised to sell dollars at 1,461 takes a mark against it when the dollar is worth 1,542.

Keep the mechanism honest, though: the mark is struck against the forward curve, not against spot. The company’s own buckets show it. The under-six-month tranche, struck at ₩1,449.97, marked −6.5% of its notional; the two-year-plus tranche, struck higher at ₩1,511.63, marked just −0.2%. And at the end of 2025 the same dollar book showed a small gain while spot sat at ₩1,437.91, above its ₩1,431.18 strike. Korean forward points are negative — the far-dated forward rate sits well below spot — so spot direction sets the sign of the mark, not its size. (Bucket detail: DART rcpNo 20260814000764, section II-5-바 — fair values of −26,203,128, −4,728,158, −7,817,479 and −46,963 thousand won across the four maturity buckets.)

Two things follow, and it is important to keep them separate.

First, this is a timing effect, not a cash loss on the business. The filing says so in its own risk section: the purpose of the hedging programme is “to minimise the uncertainty caused by exchange-rate movements and secure stable operating profit,” by “fixing the won value of future foreign-currency export receipts through forward contracts.” A mark-to-market loss on a hedge is the mirror image of a gain on the underlying receivable that has not landed yet. Reading “net profit −34%” as operating deterioration would be a mistake.

BHI itself put a headline number on the loss, in a filing separate from the half-year report. Korean disclosure rules put a derivative loss on its own dedicated form once it is large enough relative to equity — the form itself carries a “% of shareholders’ equity” box and a field recording whether the filer counts as a large company, and BHI filed one on August 14, 2026 — the same day as the half-year report. It reprints the same components as the accounts, to the won: valuation gain ₩0, valuation loss ₩39,566,299,840, transaction gain ₩330,029,659, transaction loss ₩16,552,545,817. What it puts in the headline box is the part not previously disclosed — ₩21,535,506,996, which it states as 12.1% of shareholders’ equity of ₩177,646,394,222. A first filing on May 15 had already reported the first quarter at ₩34,253,309,002, or 19.3% of the same equity base, valued at that quarter’s own rate of ₩1,513.40. (Sources: DART rcpNo 20260814900461, filed Aug 14, 2026, and DART rcpNo 20260515900610, filed May 15, 2026.)

Two pieces of arithmetic follow, and both are ours rather than the filings’. Net those four components and you get −₩55,788,815,998 — the same −$41.5M as the table above, now confirmed by a second document that reprints the components independently. Add the two quarters and the first-half derivative loss is 31.4% of equity. Neither filing prints that; each reports only its own quarter’s increment, which is how a loss of this size gets disclosed twice and read as small twice.

The same filing also complicates the tidy version of the timing story. BHI states in it that although it enters derivative contracts to hedge currency risk, “for accounting purposes it applies trading-purpose accounting” — in other words, not hedge accounting. The hedge is economically real; the accounting offset is not deferred. So every mark on the forward book lands in reported profit in the period it happens, while the matching currency gain on the export receivable only appears when that revenue is recognised, contract by contract, across the four to seven years these projects run. The filing goes on to say the two will offset and that the effect on cash flow and financial soundness is limited — that is management’s assertion, and the accounts cannot confirm it in advance. What the accounts do show is that the mismatch is structural rather than accidental: it reappears every time the won moves, in whichever direction it moves.

Second, the direction has since reversed hard, and that is checkable. BHI’s half closed on June 30, sixteen sessions after the won’s weakest print of 2026. The won has strengthened substantially since.

Daily USD/KRW closes from the Seoul session, cross-checked two ways. BHI’s own filings print an applied rate on the day of two 2026 contract disclosures — ₩1,411.00 on August 19 and ₩1,370.30 on September 2 — and both sit within 0.2% of this series on those dates. A note on why these dates matter: the FinanceDataReader USD/KRW series we normally reach for carries date labels one day early. The tell is in the calendar. From late March 2026 that series stops carrying Friday closes altogether: its last Friday row is March 27 and its first Sunday row is March 29, and from there to September 8 it holds 24 Sundays and zero Fridays. Used at face value it would have put the June 30 balance-sheet rate at ₩1,548.61, which is in fact the July 1 level. A third check agrees: BHI’s own derivative-loss disclosure values its book at ₩1,541.50 for June 30, 2026 — 0.015% from the close used here, which moves none of the ratios in this section. Bank of Korea publishes the official reference rates at the Economic Statistics System; those are fixed at a different hour and will differ slightly in level from a market close.

We are not forecasting a number. We are pointing at a mechanism: a short-dollar forward book marks the other way when the won strengthens, and between the balance-sheet date and early September the won strengthened 12.8%. Anyone underwriting BHI’s second half needs to model that line, and anyone reading only the headline “net profit fell 34%” will model it wrong.

💰 What Do You Actually Pay for BHI Today?

About $1.56bn, which is 40 times trailing earnings and 11.8 times the book value at the last audited year end — the most expensive balance sheet in its own peer group by a wide margin.

Shares outstanding are 30,944,375, all common stock of ₩500 par, of which 2,241 are held in treasury. (Source: DART rcpNo 20260814000764, section VII-3.)

And that share count is not going to grow. This is the one unambiguously good thing in BHI’s capital structure, and it is worth stating because a reader scanning a 445%-geared balance sheet will assume the opposite. BHI has no dilution overhang at all: the annual and half-year reports both answer the stock-option section with “not applicable” — no options have ever been granted, and every option column in the directors’ remuneration tables is a dash across three years — and every outstanding-balance table for securities is blank at June 30, 2026: no corporate bonds, no commercial paper, no short-term notes, no hybrid capital, no contingent convertibles. The 2022 convertible bond — ₩20.0bn, privately placed, drawn for working capital on February 28, 2022 — was converted in full on July 24, 2023 at ₩4,045, issuing 4,944,375 shares — 16.0% of the share count, on the filing’s own figure — and leaving an unconverted balance of zero. Nothing replaced it. BHI’s leverage is bank debt, which dilutes nobody. (Sources: DART rcpNo 20260814000764, section VIII-2-라 for the stock-option answer and section III-7-1, which carries the debt-securities issuance record and the five outstanding-balance tables, both as of June 30, 2026; DART rcpNo 20260320000821, section VIII-2-라; conversion detail from DART rcpNo 20230724900223, filed July 24, 2023.)

BHI closed at ₩67,700 (about $50) on September 7, 2026. Multiplying that settled close by the share count gives a market capitalisation of ₩2,094,934,187,500, or $1.56bn. (Price: Korea Exchange settled close, September 7, 2026.)

Currency note. Every dollar figure in this article converts at ₩1,345.06/$1, the USD/KRW close on September 7, 2026, held constant throughout so that figures from different periods stay comparable. Share prices are quoted in won first because that is what you will see on a Korean quote screen.

The peer comparison is where it gets uncomfortable. These are the listed Korean power-equipment names an investor would screen against.

Market caps at KRX settled closes on September 7, 2026. Revenue, operating profit, equity and liabilities from each company’s FY2025 consolidated statements via DART. P/B and debt/equity use total equity for all four so the column is comparable; for BHI, SNT Energy and KEPCO E&C non-controlling interests are immaterial, while Doosan Enerbility’s total equity of $8.93bn includes $3.14bn of non-controlling interests — on owners’ equity alone its P/B would be 7.23×. P/E is deliberately omitted: non-controlling interests took $89.5M of Doosan’s $152.6M FY2025 net profit, so a consolidated P/E for it is not comparable with the others — see our Doosan Enerbility deep dive.

BHI trades at 11.8 times book with a 368% debt-to-equity ratio. SNT Energy, which makes similar heat-transfer equipment, earns nearly twice BHI’s operating margin, carries a sixth of the leverage, and trades at 2.1 times book. That is a 5.5-fold valuation gap between two Korean power-equipment makers, and it is not explained by profitability.

🏷️ Which BHI Ticker Are You Actually Buying?

There is exactly one — KOSDAQ 083650, common stock, no preferred class and no ADR. Everything else carrying the BHI name is private: an affiliate, seven consolidated subsidiaries across six countries, and three associates, none of them listed anywhere.

(Sources: DART rcpNo 20260814000764, sections VII and XII-1 and the consolidated subsidiary note; DART rcpNo 20260320000821, section I-4 and the impairment note.)

There is no ADR, so a US or European investor has to buy the Korean line directly — see our guides to buying Korean stocks as a foreign investor and to the Korean names that do have US-listed ADRs. BHI is not on that list, and given its size it is unlikely to be.

Three of the private entities are worth a second look, and one of them is not where an English reader would expect to find it. BHI-FW in New Jersey is the vehicle for the Foster Wheeler technology relationship — the annual report lists it as a technology-tie-up counterparty running from August 2008 to August 2026 — and the parent’s separate accounts carry it 32.8% below cost after a $1.6M write-down of the investment, with a further $2.4M of goodwill impaired at the consolidated level on a 13.4% discount rate and zero terminal growth. It is a licence vehicle, not a US order book: on FY2025 numbers BHI-FW turned over $1.5M and earned $0.4M. BHI E&C Israel is carried at one won, has negative equity of $5.5M, and lost $0.8M on $6.3M of revenue; the parent wrote off a loan to it in full, citing the war. And Ehwa Private Real Estate Investment Trust No. 57, 98%-owned on an $8.9M subscription, was consolidated for the first time in FY2025 — the quiet answer to what the 2022 real-estate articles amendment was for. (Sources: DART rcpNo 20260320000821, subsidiary, impairment and related-party notes — acquisition cost 11,427,612 and carrying value 7,679,376 thousand won for BHI-FW, a 2,208,000 thousand won investment impairment in the separate accounts and 3,220,000 thousand won of goodwill impairment in the consolidated ones, and an 11,913,000 thousand won subscription to the trust.)

🌏 Is Foreign Money Buying or Selling BHI?

Selling, in every window we can measure — and the price direction depends entirely on which window you pick, flipping from +20.7% over twenty sessions to −26.8% over the full settled history we hold.

Method: investor flows arrive as share counts, so we multiply each session’s net share count by that session’s settled close and sum — an approximation, not execution prices — then convert at ₩1,345.06/$1. Source: Korea Investment & Securities Open API investor-flow data held in our own database, with closes cross-checked against Korea Exchange settled data. Our flow coverage for this ticker begins June 18, 2026, which is why 56 sessions is the longest window shown. In share terms the foreign net was −99,945, −971,989 and −1,167,655 respectively. One caveat on the retail column: our retail series is missing two sessions, June 18 and June 19, 2026, so the 56-session retail figure is the sum of 54 sessions. The foreign and institutional columns are complete in all three windows.

The sign never changes. Over twenty sessions foreigners sold $43.9M of a $1.56bn company — 2.8% of the market capitalisation — while domestic institutions bought $28.5M and retail bought $14.7M. Whatever repriced BHI this summer, it was not international money. If the mechanics of that are unfamiliar, our primer on how foreign and institutional flows move Korean stocks covers who these categories are.

Now the price, in four windows, because one window would be a story rather than a measurement.

Each window’s base is the settled close on the first session named. Korea Exchange settled closes; our settled-price history for this ticker begins March 31, 2026, so we cannot compute a 52-week range from it — see the primary-source figures below instead.

Because our own price history is short, we went to the filings for the longer view. Both the annual and half-year reports print monthly closing high, low and average, which the company sources from the Korea Exchange Data Marketplace. They show the 2026 peak was not in April, where our database sees it, but in March, at a closing high of ₩109,500. Against September 7, BHI is 38.2% below its 2026 closing peak.

Closing basis, sourced by the company from KRX Data Marketplace. (Source: DART rcpNo 20260814000764, section VII-4.) For a longer arc, the same filings print year-end closes of ₩8,100 (Dec 28, 2023), ₩15,440 (Dec 30, 2024), ₩52,300 (Dec 30, 2025) and ₩54,600 (Jun 30, 2026).

Look at January’s volume: 30,584,040 shares traded in one month against 30,944,375 shares outstanding. Essentially the entire register turned over. That was also the month the controlling family’s block shrank and a foreign 5% holder appeared and vanished — which is where we go next.

🧾 What Was JPMorgan’s 6.06% Stake in BHI, Really?

It was never one investor holding 6% — it was three JPMorgan entities aggregated, two-thirds of it sitting in the prime-brokerage arm, built partly from stock borrowed from BlackRock and Vanguard as lending agents, and it was already unwinding by the time the 6.06% figure reached the market.

JPMorgan filed a Korean 5% ownership report on BHI on January 15, 2026. Its cover page dates the event that created the obligation to file — the day the aggregate crossed 5% — to January 9, and the holdings themselves are measured as of January 13. The summary page shows 6.06%. In English-language coverage that is the number that would travel: JPMorgan takes 6% of Korean nuclear play.

The summary page is not where the answer is. Part 2 of the filing breaks the position down by entity:

No single JPMorgan entity was anywhere near 5%. The disclosure obligation arises on the aggregate, and the largest single piece — 4.20% — sat in the prime-brokerage entity, the part of an investment bank that holds and lends securities on behalf of clients rather than taking house investment positions. The purpose of holding was filed as 단순투자, simple investment: explicitly not for influencing management.

The transaction detail in the follow-up filing eight days later settles the character of the position. Its line-by-line movement log is full of entries typed 차입 (borrowed) and On-Lend Return, with named counterparties including “BLACKROCK ADVISORS (UK) LIMITED (AS AGT)” and “VANGUARD GROUP INC (AS AGT)”. (Source: DART rcpNo 20260123000059, filed January 23, 2026.)

And it did not last — the dates are the story, and they are easy to misread. The stake crossed 5% on January 9; the 6.06% picture is measured as of January 13; it reached the market on January 15. January 13 is also the trigger date of the next report — the day the holding moved by a full percentage point or more, which is what creates the obligation to file again. In other words the unwind had already started by the time the 6.06% snapshot was published. The day the stake actually crossed back under 5% is not a date any filing prints, and we do not claim to know it. By that second filing’s own measurement date of January 21, the aggregate was 1,107,786 shares, or 3.58%, with J.P. Morgan Prime Inc.’s entire 1,299,750 shares gone from the register: 386,504 shares left at Securities PLC and 721,282 at Securities LLC. (Trigger, measurement and filing dates from the cover pages of rcpNo 20260115000015 and rcpNo 20260123000059.)

The same pattern repeated with a domestic manager, and we date it the same way. Samsung Asset Management crossed 5% on February 10, 2026, reporting 1,558,391 shares or 5.04% as of that date (filed February 20); by its next measurement date, June 12, the holding was 1,220,665 shares, or 3.94% (filed June 18). (Sources: DART rcpNo 20260220001958 and DART rcpNo 20260618000254.)

Every institutional holder that crossed 5% of BHI in 2026 was back below 5% within months. That is worth knowing before reading a headline about who “took a stake.”

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

A founding family and one joint holder hold 32.09% between them — down from 40.30% on December 8, 2025, almost all of it lost in one off-market block trade on January 7, 2026 — and minority holders have received no dividend for five years, no buyback, and no Value-Up plan.

The largest shareholder is Park Eun-mi, who holds 5,200,000 shares, or 16.80% in her own name. The 32.09% figure printed beside her name on the front page of the ownership filing is the aggregate of her holding and those of six related parties — a distinction Korean 5% reports collapse on the summary page and only disclose properly in Part 2.

Share counts behind the totals: 12,471,494 (Dec 8, 2025), 9,940,537 (Jan 7, 2026), 9,929,000 (Jul 16, 2026), against 30,944,375 voting shares throughout. (Sources: DART rcpNo 20260320000909 and DART rcpNo 20260716000816.) The company rounds Cha Mi-rim’s January holding to 3.26% in its annual report; 1,007,000 of 30,944,375 shares is 3.2542%, so we print 3.25%. One more rounding note, because it is visible: the July 16 column adds to 32.08, while the filing prints a total of 32.09. The filing is right and the column is the artefact — 9,929,000 of 30,944,375 shares is 32.086%. We reproduce the filed total.

The moving part is Woo Jong-in. He was elected an inside director and co-chief executive at the March 2025 annual meeting, described in the meeting minutes as “a major shareholder of the company.” His 16.42% is not our back-calculation: the FY2025 annual report’s note on general information prints the register verbatim at December 31, 2025 — “Woo Jong-in 5,082,000 shares 16.42%, Park Eun-mi 5,200,000 shares 16.80%, Cha Mi-rim 1,007,000 shares 3.26%, treasury 2,241 shares 0.01%, others 19,653,134 shares 63.51%, total 30,944,375 shares 100.00%.” (Source: DART rcpNo 20260320000821, consolidated note 1.) He resigned on December 8, 2025 “for personal reasons,” and on the same day the block filed a report noting that “a change in the reporting person’s own shareholding is expected in accordance with the pre-trade plan for executives and major shareholders disclosed on 2025-12-08.”

On January 7, 2026 he sold 2,460,000 shares off-market at ₩52,528 — proceeds of roughly $96.1M — cutting his stake from 16.42% to 8.47%. Two much smaller related-party sales bracketed it: Cha Mi-rim sold 31,000 shares at ₩51,270 on December 9, 2025, and BHI Construction sold 39,957 at ₩59,699 on December 12. (Source: DART rcpNo 20260107000014, Part 3, as corrected.) The affiliate then cleared its remaining holding in three on-market lots between February 20 and May 7, 2026, at ₩89,900, ₩102,000 and ₩101,000. (Source: DART rcpNo 20260716000816, Part 3.)

The shape of that decline matters more than its size. This was not a controlling family drifting out of its own company over seven months. It was one pre-announced off-market block trade: 7.95 of the 8.21 points the block has lost went in a single transaction on a single day, thirty days after the seller resigned as co-chief executive and filed a pre-trade plan. The remaining 0.26 points are a relative and an affiliate selling small lots: 0.23 points before the block trade, and 0.03 points after it. Between January 7 and the block’s last filing on July 16 — six months — the total movement was those 0.03 points, and no report has been filed since, which under Korean rules means no holder has moved a full point. A planned exit followed by silence is a different fact from a slow bleed, and it is the one the filings actually describe.

Note what that affiliate did over the cycle: BHI Construction bought 19,700 shares at ₩10,360 in June 2024 and 8,994 at ₩55,500 in October 2025, then sold out. The 8.47% Woo retains is held under a joint-holding agreement with Park Eun-mi, which is why it still counts toward the block.

Meanwhile the board has become a family board with two outsiders. The Vice Chairman and co-CEO is the largest shareholder’s spouse. The President, appointed to the board in 2022, is her daughter, born in 1990, with a Penn State degree in international politics. There are four inside directors and two independent directors; one of the two, Kim Hak-do, is a former Vice Minister of SMEs and Startups who previously ran MOTIE’s trade negotiation and energy resources bureaus.

What does a minority holder actually receive?

Nothing, so far. BHI has paid no dividend in five years, holds 2,241 treasury shares (0.01% of the company), has never cancelled a share, and has never filed a Value-Up plan.

The dividend policy section is unusually direct about it: “Owing to deteriorating results caused by changes in the business environment, the company has not paid a dividend for the past five years; when results improve, it plans to establish dividend targets, the financial metrics and calculation method underlying them, and plans for share buybacks and cancellation.” Results did improve — FY2025 net profit was $48.5M — and BHI set a record date of December 31, 2025 for a possible FY2025 year-end dividend. The dividend table in the annual report then records “X” against all four periods of 2025: no quarterly, no interim, no year-end dividend was declared. (Source: DART rcpNo 20260320000821, section I-6.)

There is a mitigating fact worth stating plainly: BHI’s retained earnings were negative until FY2025. Accumulated losses ran ₩27,178m at end-2023 and ₩8,461m at end-2024 before turning positive at ₩56,054m at end-2025. A Korean company cannot distribute out of an accumulated deficit. The zero dividend through 2024 was arithmetic, not policy. The zero dividend for FY2025 was policy.

Two further governance markers. First, BHI’s annual report answers “X” to whether it can set the dividend record date after the dividend amount is decided, and “no” to whether it plans to adopt that practice — meaning it has declined Korea’s dividend-procedure reform, which exists precisely so investors know what they are buying before the record date. Second, on the Value-Up question we searched rather than counted: querying DART’s filing index for report names containing 기업가치제고계획 across January 1, 2024 to September 8, 2026 returns zero results among BHI’s 170 filings. The same query returns three for KEPCO E&C and one for SNT Energy. Our guide to Korea’s Value-Up programme explains why that matters, and our explainer on the Korea Discount explains what it costs.

One disclosure-quality note for the record. On July 2, 2024 the KOSDAQ market division placed BHI under a six-month suspended designation as an unfaithful disclosure corporation, arising from a supply-contract cancellation notice that constituted a full reversal of a previously disclosed contract; the suspension was granted because the cancellation was made at the counterparty’s request. (Source: DART rcpNo 20260814000764, section XI-3.) Separately, BHI amends its supply-contract disclosures constantly — six of the eight contract filings it made between July 10 and September 8, 2026 were amendments to earlier ones. Reading them, they are mundane: nearly all are exchange-rate restatements or schedule extensions, not order losses. One example makes the point. On August 19, 2026 a contract’s won value was revised down from ₩56,576,139,200 to ₩54,913,822,792, which looks like a 2.9% cut; the amendment note shows the dollar amount actually rose, from $38,414,000 to $38,598,832 plus €276,000, and only the applied exchange rate changed, from ₩1,472.80 to ₩1,411.00. (Source: DART rcpNo 20260819900396.)

🇰🇼 Who Is the Hong Kong Company on BHI’s Newest Contract?

Tie Jun International (HK) Limited is not a new counterparty — it has been buying from BHI since 2019, and the half-year report names its current project: Az Zour North 2&3 IWPP in Kuwait.

BHI’s most recent contract disclosure, on September 2, 2026, is for $106.0M of LNG combined-cycle equipment for delivery to Kuwait, running to April 15, 2029, with a counterparty named as Tie Jun International (HK) Limited, whose own most recent revenue is filed as “—”. (Source: DART rcpNo 20260902900118, which states the contract as USD 106,000,000 and converts it at ₩1,370.30 to ₩145,251,800,000.) A blank counterparty revenue line and a Hong Kong address is the sort of thing that reads badly in translation.

The half-year report fills in the gap without any inference required. Its progress-basis order table shows a contract with TIE JUN INTERNATIONAL(HK) LTD dated December 29, 2025, running to September 30, 2027, worth $115.9M, categorised as EPC and 1.0% complete; and an earlier, completed HRSG contract with the same counterparty dated April 3, 2019, worth $42.6M. The project-level note dates the December 2025 contract to Az Zour North 2&3 IWPP. (Source: DART rcpNo 20260814000764, section III-8-라, which prints ₩155,827,007,425 and ₩57,292,346,837.)

So the relationship is seven years old, the country is one where BHI also holds a contract with Alghanim International, and the September contract is the second Tie Jun award in eight months. We are not claiming the September contract is the same project — the disclosure does not name a project — only that the counterparty and the country are already in the book.

The other disclosed terms are worth reading as risk. The contract carries a 10% advance payment and invoice terms of 45 days, and the filing states that BHI had already begun work and collected part of the money under a limited notice to proceed before the contract was signed. We print the filing’s own dollar figure rather than reconverting its won total at the house rate used elsewhere in this article, which would read $108.0M; where a filing states a number, we quote it rather than re-derive it. Note that the “18.8% of revenue” ratio printed on that filing is measured against FY2025 revenue, which is the correct basis for the disclosure rule but will understate the ratio if you compare it to an annualised 2026 run rate.

📚 Lingo Check

🎯 Why It Matters for K-Export Stars

Every number in this article was public on August 14, 2026, in one document, for free. None of it required a source, a broker or a subscription. What it required was reading 222,607 characters of Korean and being willing to count.

The counting is the point. The English-language description of BHI available to a foreign investor is, in practice, “Korean nuclear equipment maker” — and that description survives because nobody checks it against the three tables in the company’s own filing that price it at 3–5% of the book. Nor does anyone outside Korea read the derivative note that explains why the headline profit fell while the business doubled, or the Part 2 of a 5% report that turns “JPMorgan takes 6%” into a prime-brokerage inventory that was already unwinding before the 6% headline was published.

This is what our Foreign Flow Watch series measures from the other side: foreigners sold $43.9M of BHI over twenty sessions while it rose 20.7%. That is not a coincidence and it is not an accusation. It is what happens when the marginal buyer is working from a translated label and the marginal seller has read the filing.

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. All figures are sourced from public filings and market data as of the dates stated; prices, valuations and flows change and may be out of date by the time you read this. Currency conversions use ₩1,345.06/$1 as of September 7, 2026. Do your own research and consult a licensed adviser before investing. Our corrections policy is at kexportstars.com/about/#corrections.