KEPCO KPS (051600): Korea’s Power-Plant Maintenance Contractor, Decoded

Every headline about Korea’s nuclear export push is about building something new. Reactors for the Czech Republic, reactors maybe for America. But there is a company in Naju, four hours south of Seoul, that has never built a reactor and never will — and it has been paid, every single year since 1984, to open Korea’s reactors up, inspect them, and close them again. Its stock spent exactly one day, June 12, 2026, closing 29.6% higher — one tick short of the daily price limit — and has since given back more than a quarter of its value. That gap between the story and the schedule is what this article is about.

🏢 What Does a Power-Plant Maintenance Company Actually Sell?

KEPCO KPS sells labour, diagnostics and specialist tooling to keep other people’s power plants running — it owns no plants, buys almost no raw materials, and its customers supply the parts. The half-year report puts it plainly: “the power-plant maintenance service business does not carry out work using raw materials; it is performed using materials procured by the ordering party.” What it brings instead is 6,253 employees with average tenure of 13.6 years, a fleet of non-destructive testing rigs, and the accumulated right to be inside a reactor building.

The formal name is KEPCO Plant Service & Engineering Co., Ltd. (한전KPS, KRX: 051600). It was established on April 1, 1984 as Korea Electric Power Maintenance Co., wholly funded by Korea Electric Power Corporation — the state utility known abroad as KEPCO, still its 51% owner — and listed on the KOSPI on December 14, 2007. Headquarters moved to Naju, South Jeolla Province in December 2014 under Korea’s public-institution relocation law. It runs 66 domestic and 9 overseas offices, and has operated without borrowings since 2000: the capital-management note records borrowings of exactly zero at both June 30, 2026 and December 31, 2025.

The work divides into four kinds. Routine maintenance (경상정비) is the standing crew at every plant, on annual contracts. Planned preventive maintenance (계획예방정비) is the scheduled outage — the plant stops, the reactor is opened, fuel is replaced. Refurbishment is life-extension work. Commissioning maintenance is what happens on a new plant before commercial operation. Only the last connects to a new-build order, and it arrives at the very end of a construction programme.

📊 Which Reactors and Boilers Pay the Bills?

Thermal power, not nuclear, was the larger revenue segment in the first half of both 2025 and 2026 — but on a full-year basis nuclear was larger in 2024 and 2025, and smaller in 2023. There is no trend here. There is a schedule.

This is the most misread number in the company. An investor who opens the half-year report, sees nuclear at 37.5% and thermal at 39.2%, and concludes “the nuclear play is only a third nuclear” has been fooled by seasonality: reactor outages are scheduled away from Korea’s summer and winter demand peaks and cluster into the second half. Read the full years instead.

Converted at the rate used throughout this article, half-year 2026 revenue of ₩790,333mn is $588M, and FY2025 revenue of ₩1,576,517mn is $1.17bn. Nuclear and pumped-storage revenue grew 17.1% year on year in the first half of 2026 while thermal grew 3.3% — the nuclear side is accelerating, off a low base set by an unusually quiet 2025.

One housekeeping note. The filing prints the parent’s overseas half-year revenue as 47,369 in one table and 41,369 in another four pages later. Only 47,369 reconciles to the parent subtotal of 770,089 and to the 5.99% share the company itself prints, so we used it — and flagged the discrepancy rather than quietly picking one.

🗓️ How a Repair Shop Became a Listed Public Enterprise

KEPCO KPS has been bought, absorbed, dissolved and re-founded — the company that exists today was created in 1984 specifically to be KEPCO’s in-house maintenance arm, and it has never stopped being that. The history is short enough to print in full, and it explains why the state still owns half of it.

That 2017 reclassification is not cosmetic. Under Korea’s Act on the Management of Public Institutions, a quasi-market-type public enterprise has its budget, executive appointments, executive pay and annual performance evaluation supervised by the government. The subsidiaries are recent and small: a Philippine entity (2013), a facility-management arm (2020), and a South African entity (January 2023).

⚙️ Why the Outage Calendar Is This Company’s Real Income Statement

KEPCO KPS’s profits track the number of reactor and boiler overhauls completed in a year, and 2025 was a trough year — 13 nuclear units against 19 in 2024, which is most of why operating profit fell a third.

The company discloses the unit counts, but buries them in a narrative section on maintenance methodology rather than the financial statements. Here they are next to the profit line — the comparison the filing never makes for you.

Read the thermal row across all four columns before drawing a trend from it: 93 units in 2023, 71 in 2024, 87 in 2025 and 96 planned for 2026. 2024 was the trough, not the baseline — thermal overhauls in 2025 recovered towards, but did not reach, the 2023 level. The nuclear row is the one with a genuine 2025 slump.

Now the profit. Between FY2024 and FY2025, revenue rose 1.2% but cost of sales rose 6.1% — five times faster. Gross profit fell $42.8M and selling, general and administrative expense rose $8.8M, producing an operating profit decline of $51.6M (₩57,583mn, ₩11,814mn and ₩69,397mn as printed). 83.0% of the fall came from gross margin, not from overheads. Operating margin went from 13.45% to 8.89%.

The first half of 2026 has already turned: operating profit is up 9.8% year on year and margin has recovered to 10.20%. Seven nuclear overhauls were completed in the first half against three in the same period of 2025. If the 20-unit plan holds, 2026 is a recovery year — but note that “plan” is the company’s own forecast, and 2025’s plan was 22 units against 13 delivered.

🌍 How Much of Korea’s Fleet Does It Actually Maintain?

KEPCO KPS performs 68.0% of the maintenance on Korea Hydro & Nuclear Power’s reactor fleet and 49.0% on the five state gencos’ thermal fleet, measured by capacity — and the nuclear share has drifted down 2.6 points since 2023 while the thermal share climbs.

These are the company’s own capacity-weighted figures, disclosed with an explicit warning that “market share differs depending on the calculation basis.” That caveat is the company’s, and removing it would be dishonest.

A more concrete way to say it: of the 26 reactors in commercial operation at June 30, 2026, KEPCO KPS maintains 16 on its own and 10 through joint consortia. It touches every reactor in the country — but its capacity-weighted share of the fleet is 68%, not 100%, because the consortium units are shared and some scopes go to competitive tender.

That is the liberalisation story, and it is slow. Korea ran a first phase of maintenance-market competition from 2013 to 2017, under which KEPCO KPS kept core equipment at the newest large plants while equivalent volumes went to tender. The filing states the second phase has been deferred — the phase-two policy decision “is in a state of reservation.” Private competitors, cultivated since 1997, now work at the Taean, Hadong, Dangjin and Yeongheung thermal stations. The important word is deferred: an unresolved policy overhang, not a closed question.

🔧 What Is the Only American Name in These Filings?

Westinghouse — and it appears as a supplier, not a customer. The 30-year technology agreement under which KEPCO KPS licensed reactor primary-side maintenance technology from Westinghouse ran from May 1, 1994 to April 30, 2024 and lapsed; a replacement, running ten years from December 31, 2025 to December 30, 2035, appears for the first time in the FY2025 annual report.

First, a check: is there any American order in the file at all?

We found no American order, and here is exactly how we looked: we pulled all 671 KEPCO KPS filings from January 2015 to September 2026, isolated the 242 single-sale-and-supply-contract disclosures, and read the “sales and supply region” and “counterparty” fields on every one that the DART document API would return. None names the United States.

The method matters more than the answer, so here are its limits. Of the 242, 177 were machine-readable and 65 were not — the API returns no document for those, and they are almost all old (2016–2019). So the honest claim is narrower than “none”: no supply-contract filing we could open names an American region or counterparty, and our blind spot sits in 2016–2019. Separately, the word “미국” (United States) appears in only two contexts in the entire half-year report — once describing 1960s equipment imports, and once as Westinghouse’s nationality, in the two copies of the same technology-agreement table.

What the sweep does show is where the overseas work is: UAE Barakah on 32 filings, India on 10, Pakistan on 4, Uruguay on 2, Madagascar on 2, Romania on 1. Two caveats matter more than the count.

First, government-to-government talks and memoranda of understanding are not disclosable events in Korea. If Korean and American officials are discussing reactors — and press reporting says they are — no Korean company is obliged to file anything about it, and none has. An absence in DART is evidence about DART, not about the world.

Second, and more interesting: the one American entity that does appear is not a customer. It is a supplier of technology, and the dates around it are the most striking thing in this filing.

The technology agreement, in three filings

The three annual filings tell the story by themselves. Each contains a table of “technology introduction” agreements — inbound licensing, not outbound. The Westinghouse row reads:

Set that against the public timeline. Westinghouse sued in a US federal court in October 2022 to block KHNP and KEPCO from exporting the APR1400 design without its consent; the case was dismissed in September 2023; and on January 17, 2025 the three companies announced a settlement of all APR1400-related intellectual-property disputes, saying future cooperation would cover engineering, construction services, maintenance, fuel supply and training. The commercial terms were confidential, and no royalty or licensing figure has ever been published. (Sources: World Nuclear News, Jan 2025; Korea Herald, Jan 17, 2025.)

So: the old agreement expired in April 2024, mid-dispute. A settlement was announced in January 2025. A new ten-year agreement begins at the end of December 2025. Be precise about what that does and does not establish — the filing gives dates, not reasons, and never mentions the settlement. We are not asserting the settlement produced this contract. We are pointing out that a Korean regulatory filing contains a dated, ten-year, named commitment in exactly the area the settlement said the parties would cooperate in, and that no English-language source we could find reports it.

It also complicates a comfortable story. Korea’s nuclear champions are usually described as having achieved technological independence. Korea’s dominant nuclear maintenance contractor has just committed to another decade of licensing primary-side maintenance technology from an American vendor, having already done so for thirty years.

🌐 Why the Backlog Table Tells You Almost Nothing

Overseas work is 6.2% of the parent’s half-year revenue and 56.1% of the disclosed contract backlog, because a Jordanian diesel O&M contract running to 2039 books a quarter-century of remaining value today while a year of Korean routine maintenance books almost nothing.

Duration, not size, drives this table. Read it as a map of contract length rather than of business importance.

And the overseas line is not one contract. The filing’s detailed order table lists eight overseas projects, of which the Jordan IPP3 600MW diesel O&M contract with Amman Asia Electric Power carries a remaining balance of $659M (₩886,880mn) — 60.0% of the overseas backlog on its own, and 33.7% of the whole table. It was signed on August 5, 2013 and runs to February 4, 2039. (Source: rcpNo 20260814001938, detailed table 4, “Order status”.)

Thermal is the extreme case: 40.24% of revenue, 2.43% of backlog. That is not a collapsing business — it is one run on one-year contracts re-signed every December, so at any moment there is almost nothing left to report. Anyone screening Korean industrials on backlog-to-revenue will misprice this company in both directions.

🇷🇴 Who Actually Hired KEPCO KPS for Romania?

Korea Hydro & Nuclear Power did — not Romania’s utility. Both of KEPCO KPS’s overseas nuclear jobs are subcontracts from KHNP, an unlisted state company that foreign investors cannot buy.

This governs the entire Korean nuclear complex: the entity that signs export contracts is Korea Hydro & Nuclear Power (DART corporate code 00382001), which has no stock code and is not listed. Every listed “Korean nuclear stock” sits at least one step below it. Here is where KEPCO KPS sits.

In the dollar terms this article uses throughout, the Romanian contract is $361M and the Barakah contract $348M. For scale, Romania’s Nuclearelectrica signed the overall Cernavodă-1 refurbishment EPC contract in December 2024 at approximately €1.9bn (about $1.97bn), with a consortium of Candu Energy (AtkinsRéalis), Ansaldo Nucleare, the Canadian Commercial Corporation and KHNP; KHNP leads the construction phase. (Source: World Nuclear News; Nuclearelectrica, Dec 19, 2024.) KEPCO KPS’s own slice, at $361M, is roughly 18% of that headline number — and it reaches the company through KHNP.

Cernavodă is a CANDU-6 pressurised heavy-water reactor, which is why the work is “retubing”: in a CANDU the fuel sits inside hundreds of horizontal pressure tubes, and replacing them is the central act of a mid-life refurbishment. The filing states site work begins in September 2027. So this contract, signed in August 2025, contributes very little revenue for another year.

The Barakah row deserves its own paragraph, because what matters is not its current end date but how many times that date has moved. The contract was first filed on August 9, 2013, and the April 2026 filing lists sixteen earlier disclosures against this one contract, running back to that original notice. We read the five most recent of those, plus this April 2026 filing itself; our DART list query started in 2022, which is a limit of our search window and not of the record. Those six run as follows. June 2022 pushed the end date from June 30, 2022 out to May 31, 2024 and lifted the value from ₩335.97bn to ₩439.93bn. May 2024 moved it to December 31, 2024 and ₩462.16bn. December 2024 moved it to December 31, 2025 and ₩463.14bn. October 2025 raised the value again, to ₩467.04bn, for price escalation and design-change volume — but left the end date untouched, so it is a revision and not an extension. Then, on December 31, 2025 — the day it was due to expire — the company deleted the end date altogether, stating the contract was under renegotiation with the customer. On April 17, 2026 it filed the date back in as December 31, 2026, with the value at ₩467,531,980,000.

So the honest reading is not that a thirteen-year revenue line has suddenly acquired an expiry. It is that this contract has been extended roughly a year at a time, at or just before each expiry, four times since 2022, and the filing’s own list of sixteen earlier disclosures shows the same amend-and-extend running back to 2013 — a longer base rate than the stretch we read line by line. The current expiry is real and it is under four months away; the base rate says it gets extended again; and the extension is never on file until it happens. We could find no English-language source that tracks the sequence. (Sources: DART supply-contract amendments rcpNo 20220616800296, 20240530800264, 20241211800702, 20251017800201, 20251231800780 and 20260417800376.)

🧾 Why the 2025 Maintenance Contract Was Signed on December 30, 2025

KEPCO KPS routinely performs a full year of maintenance before the contract for that year is signed, is paid in the meantime at the previous year’s contract price, and settles the difference retroactively afterwards — the company says so in the disclosure itself.

This is the most consequential sentence we found, and it is one line of boilerplate. The $54.0M (₩72.6bn) 2025 routine-maintenance contract with Korea Midland Power covers January 1 to December 31, 2025 — and its contract date is December 30, 2025, the second-to-last day of the work it covers. The filing explains why in the company’s own words: the work was “started first at the ordering party’s request, with the contract concluded afterwards,” and progress payments are made “applying the 2024 contract amount, with retroactive settlement carried out after the contract is concluded.”

Three otherwise puzzling things then line up at once.

First, margin. If your costs are this year’s and your interim revenue is last year’s price, wage and subcontractor inflation lands on you until the true-up. In FY2025 cost of sales grew 6.10% against revenue growth of 1.25%, and payroll is not a small variable: the parent’s first-half 2026 employee compensation was $199.8M (₩268.75bn, printed as 268,750,447 in a table denominated in thousands of won) against the parent’s own half-year revenue of $572.5M (₩770,089mn), or 34.9% of it. (Source: rcpNo 20260814001938, employee status table, parent company.)

Second, unbilled work, which behaves exactly as that arrangement predicts.

The balance builds through the first half and drains at year-end when contracts are signed and settled — every year. That is structural, not a warning sign, which is why we checked the prior year before writing anything alarming. One more year-end is worth naming, because it changes the level rather than the pattern: unbilled work stood at $409M (₩549,806mn) at December 31, 2023 — more than double either of the two year-ends in the table. The seasonal build-and-drain is a permanent feature; the balance it drains to is not fixed. What is worth watching: the mid-year build was larger this year than last (+54.7% versus +41.6%), and the June-to-June balance grew 14.8% against 6.5% revenue growth.

Third, the auditor. For FY2025, the audit report’s Key Audit Matter was: “Revenue recognition on the percentage-of-completion basis using the input method — increased uncertainty in estimating total contract costs, and the risk that revenue is overstated as total contract cost amounts change.” That is the auditor naming the same mechanism. It is not new, though, and the article would mislead you if it implied otherwise: the FY2025 annual report prints the identical Key Audit Matter for FY2024 and FY2023 as well, so this is a standing feature of the audit, not a fresh flag. Separately, the auditor changed for FY2026, from EY Han Young to KPMG Samjong.

🏛️ Who Actually Sets This Company’s Dividend?

The government does, in substance: KEPCO KPS anchors its payout target to a 40% payout-ratio target for state-invested institutions set by the finance ministry, and says that changing its dividend calendar requires “consultation with the supervising authority” as well as an amendment to its articles of association.

The ownership chain runs: Korean government → Korea Electric Power Corporation → KEPCO KPS. The filing spells out the top link: the state owns 51.1% of KEPCO, split between the Korea Development Bank at 32.9% and the government directly at 18.20%, and by law it must hold a majority; the voting rights attached to both are exercised by the Ministry of Trade, Industry and Energy in consultation with the Ministry of Economy and Finance. KEPCO in turn owns 51.00% of KEPCO KPS, unchanged through the period. Below it sit the National Pension Service at 9.09% and the employee stock ownership association at 1.20%, leaving 38.71% outside those three, on our own subtraction. The filing separately reports 65,344 minority holders owning 19,083,267 shares, or 42.41%; that line overlaps the ones above and is why the four percentages do not add to 100. (Source: FY2025 annual report, rcpNo 20260323000091, shareholder section, as at Dec 31, 2025; parent stake reconfirmed at Jun 30, 2026 in rcpNo 20260814001938.)

Two details from the dividend-policy section exist only in Korean. First, the benchmark: the company sets its payout target “using the 재정경제부 payout-ratio target for government-invested institutions (40%) as the standard,” measured on separate — not consolidated — net profit. 재정경제부 is the Ministry of Finance and Economy, a body abolished in 2008; its successor is today’s Ministry of Economy and Finance. The company is still quoting a target set by a ministry that no longer exists. Second, the calendar. For each of the last three financial years the record date preceded the date the amount was fixed: for FY2025, a record date of December 31, 2025 and an amount confirmed on March 31, 2026. The filing marks “dividend predictability provided” as X for all three years, and says fixing it requires “consultation with the supervising authority and improvement through amendment of the articles of association.”

Korea’s dividend-procedure reform exists precisely to stop investors buying a stock for a dividend whose size they cannot yet know. Here is a listed company saying it needs a ministry’s agreement before it can comply — the mechanism behind the Korea Discount.

💸 A “High-Dividend Company” Whose Dividend Fell 33%

KEPCO KPS is a designated high-dividend company under Article 104-27 of Korea’s Restriction of Special Taxation Act, and its dividend fell 33.1% year on year — from ₩111.105bn to ₩74.295bn. Both statements come from the same filing, on the same page.

Two disclosures document the sequence. On March 19, 2026 the company filed its 2026 Corporate Value-Up plan. Under “whether a high-dividend company under Article 104-27” it answered Not applicable, every supporting field blank — while its shareholder-return chapter promised to “establish an identity as a high-dividend stock with a payout ratio exceeding 50%” and to hold a minimum 50% payout through 2028. (Source: Corporate Value-Up plan, rcpNo 20260319800476, board decision Mar 18, 2026.)

On April 20, 2026 it re-filed. The reason given was “correction consequent on qualifying as a high-dividend company.” The changed fields were:

How does a company with a −33.1% dividend growth rate qualify under a rule whose first condition is that the dividend has not been reduced? The filing answers, in a note the company wrote itself:

“Because the ‘immediately preceding business year’ in Article 104-27(1)(2) of the Restriction of Special Taxation Act means the 2024 business year, the Company’s dividend income for the immediately preceding business year (2024) compared with the base year (2024) is interpreted as not having decreased, and the high-dividend company requirement is therefore satisfied.”

The qualification rests on 2024 being compared with 2024. The company also invokes a transitional provision — it says it falls within the case of a high-dividend company that had already disclosed a Value-Up plan before the enforcement date of the relevant Presidential Decree — and re-filed the March plan on that basis. And the standard header on every such disclosure states that “whether a company is a high-dividend company is determined by the company’s own judgement.”

We are not saying this is improper — transitional rules exist and companies are entitled to read them. We are saying that anyone looking at the “designated high-dividend company” label should know it is self-assessed, that it turns on which year counts as the preceding one, and that the dividend itself fell by a third.

One more thing, because it is the part that actually affects our readers. The Article 104-27 benefit is a separate taxation election that applies to resident individual shareholders — it takes their dividend income out of Korea’s global progressive tax base. A non-resident is not taxed that way; a non-resident’s Korean dividend is withheld at source under the statutory rate or a treaty rate. The high-dividend designation is very unlikely to reduce a foreign investor’s tax bill on this stock. What you actually pay is set out in our guide to Korean dividend withholding tax, and the programme this designation belongs to is explained in our guide to Korea’s Value-Up programme.

📉 What You Pay, and What the Competition Costs

KEPCO KPS trades at 17.2 times FY2025 earnings and 1.58 times book while its three listed private-sector counterparts trade at 6.2 to 9.0 times earnings — and all three of them earn a higher operating margin.

The valuation snapshot first. All figures are as at the KRX settled close of September 7, 2026.

Now the peers. Korea has three listed independent power-plant maintenance contractors, and they are the direct beneficiaries of the liberalisation policy described above.

One larger company is deliberately absent from that table, so we should name it rather than let its absence flatter the comparison. Korea Electric Power Industrial Development (한전산업개발, KEPID, KRX 130660) serves the same state gencos and is worth $309M on KRX’s listed-share count — more than any of the three above. Two things keep it out. It is not independent: KEPCO owns 29.00% of it, second to the Korea Freedom Federation’s 31.00%. And it does different work — its own segment table puts 95.7% of FY2025 revenue in fuel handling, flue-gas desulphurisation and ash handling, the plant services around the boiler rather than the boiler itself. (Source: DART FY2025 annual report, rcpNo 20260319000705, shareholder and segment sections, as at Dec 31, 2025.) For the record, its numbers would not rescue the comparison in either direction: an operating margin of 1.67% on revenue of $293M, and roughly 102 times earnings and 4.0 times book.

KEPCO KPS is by far the largest of them all, has the only nuclear franchise, and holds net cash worth 13.4% of its market capitalisation. Those are real reasons for a premium. But it is 2.6 times Kumhwa PSC’s P/E against a company earning a higher operating margin — 10.34% versus 8.89% — and 3.0 times its P/B against a company whose ROE is only 1.3 points lower. Whatever you are paying for, it is not superior unit economics.

One note undercuts the word competitor entirely: every company in the peer table above is one of KEPCO KPS’s own consortium partners on nuclear maintenance. The half-year report’s notes name three such contracts — Shin-Hanul 1&2 (KEPCO KPS 60%, Doosan Enerbility 20%, Kumhwa PSC 20%), Saeul 1&2 (KEPCO KPS 60%, Soosan Industries 20%, Kumhwa PSC 20%) and Hanbit 3&4 (KEPCO KPS 80%, Iljin Power 20%) — with each party disclosing only its own slice as a contract value. So the peer table is not a list of outsiders trying to prise the franchise open. It is a list of the firms the franchise is already shared with, at the newest units. (Source: rcpNo 20260814001938, notes 4–6 to the major-contracts table.)

🏷️ Four KEPCO Names, and the One You Cannot Buy

There is no ADR and no US listing for KEPCO KPS; the only KEPCO-family security on a US exchange is the parent, Korea Electric Power Corporation, on the NYSE — a completely different company. Confusing the two is easy and expensive.

Buying 051600 requires a Korean-market route rather than a US ticker; the practical options are covered in our guide to buying Korean stocks as a foreign investor, and the full list of what is genuinely available as an ADR is in our Korean ADR guide.

🔗 How Concentrated Is the Customer List, Really?

86.8% of KEPCO KPS’s first-half 2026 consolidated revenue came from related parties, and KHNP alone accounted for 37.7% — essentially level with the five state gencos combined. This is not a company with customer concentration risk. It is a company with one customer group.

Even the overseas names on that list — Jordan’s Amman Asia Electric Power and the UAE’s Nawah Energy — are related parties, because KEPCO holds stakes in them. Geographic diversification and counterparty diversification are not the same thing here.

One warning about a number quoted elsewhere. The same filing’s large-transaction disclosure puts KHNP at ₩298.0bn and “19.28%.” That percentage uses FY2025 full-year separate-basis revenue as its denominator, not first-half revenue — the filing says so in its own footnote. Against the parent’s actual first-half revenue of ₩770,089mn — rather than the consolidated ₩790,333mn used in the table above — the same ₩298.0bn is 38.7%. Both are correct; they answer different questions. Quoting 19.28% as a first-half concentration figure understates it by half.

🌏 Three Windows, Three Different Stories About Who Is Selling

Foreign investors have been net sellers of KEPCO KPS (051600) over the last 20 sessions and the last 56, but essentially flat over the last five — and domestic institutions and retail investors each flip sign depending on which window you choose.

This is why we never publish a single window. Below, all three investor groups, three windows, one table.

Read the foreign row carefully. The 56-session net sale is smaller than the 20-session one — only possible if foreigners were net buyers over the 36 sessions before August 10, which they were, by roughly 51,000 shares. The selling is recent and concentrated, not a steady exit, and it has stopped: over the five sessions to September 7 foreigners sold a net 2,152 shares, which on this stock is noise.

The price windows disagree too, and one of them disagrees so violently that we are not going to print it as a number. Against the KRX settled close of ₩47,350 on September 7, 2026, the stock is up 4.41% over five sessions (from ₩45,350 on September 1) and up 2.71% over twenty (from ₩46,100 on August 10).

The sixty-session window is a trap, and it is worth showing you why rather than picking a number. On June 12, 2026 KEPCO KPS closed at its high of ₩63,900, a one-day gain of 29.61% on about two and a half times normal volume — and gave back 15.49% the very next session. It did not actually lock limit-up. The previous close was ₩49,300, so the +30% ceiling was ₩64,000 after Korea’s ₩100 tick rounding, and the stock never touched it: high and close were both ₩63,900, one tick short. Other Korean stocks did lock limit-up that session — HL Mando closed at ₩65,000 from ₩50,000, HPSP at ₩71,500 from ₩55,000, each exactly +30.00%. (Source: KRX Open API daily trading data; open ₩50,100, high ₩63,900, low ₩49,400, volume 439,122 shares against a prior-20-session average of 172,567.) A sixty-session lookback from September 7 lands its baseline on either June 11 or June 12 depending on whether you count observations or intervals. Pick June 11 and the stock is down 3.96%. Pick June 12 and it is down 25.90%. A one-day offset moves the answer by 22 percentage points, so any “three-month performance” figure for this stock is an artefact of that single session unless the writer tells you which side of it they stood on.

The measures that survive the spike: the stock is down 26.59% from ₩64,500 on April 27, 2026 — the highest close in our daily price coverage, which starts March 31, 2026, not a 52-week high — and down 14.68% from ₩55,500 on March 31, 2026, the start of our daily price coverage.

That June 12 session is also the answer to the question this article opened with. The market did not ignore KEPCO KPS while the nuclear theme ran. It bid the stock to within one tick of its daily price limit for exactly one day, then spent the next three months handing the gain back — which is roughly what you would expect of a stock whose FY2025 profits fell a third for reasons that had nothing to do with anybody’s export ambitions. For how to read Korean flow data generally, see our guide to foreign and institutional flows.

📚 Lingo Check

🎯 Why It Matters for K-Export Stars

Almost everything an English-speaking investor can read about Korean nuclear is about export announcements. That coverage is fast and well-resourced, and we cannot beat it. What it leaves out is the part that exists only in Korean regulatory filings: who is actually paying, on what contract, signed when, expiring when.

Here those documents contain a thirty-year Westinghouse licence that quietly lapsed and a ten-year one that quietly replaced it; a Romanian contract whose counterparty is a company nobody can buy; a Barakah revenue line whose expiry has been pushed out four times since 2022 and was deleted outright for three and a half months; a maintenance contract signed on the second-to-last day of the year it covers; and a high-dividend designation resting on comparing a year with itself. None of it was hidden. All of it was in Korean, and every figure above carries a receipt you can open.

Disclaimer: This article is for informational and educational purposes only and does not constitute investment advice, a recommendation, or an offer to buy or sell any security. All figures are drawn from public regulatory filings and market data as of the dates stated and may since have changed; share prices, valuation multiples and flow data go stale quickly. K-Export Stars holds no position in the securities discussed. Do your own research and consult a licensed financial adviser before making any investment decision. If you believe any figure here is wrong, please tell us — our corrections policy is public.