Korea’s Road to a US Reactor: A 50-Year Timeline Built From Filings (1975–2026)

On Monday, September 7, 2026, KEPCO Engineering & Construction closed up 13.66% and Doosan Enerbility up 10.98%, on reports that Seoul and Washington were discussing up to eight large reactors on American soil. The reports are real. Nothing has been signed. But the runway underneath them is fifty years long, every plank of it is in a filing, and the last time a headline like this appeared it took seventeen months and six official non-answers to become a contract.

Updated September 29, 2026: the one scheduled event this article pointed at has happened. On September 17, 2026 Doosan Enerbility filed its seventh and final answer on the NuScale SMR reports, and it is a denial — the closing section has been rewritten from a forecast into a result. The reported September 18 investment signing also passed with no announcement from either government. Everything else here is unchanged and still carries its original verification date: the market data is the settled close of September 7, 2026 and has not been refreshed, so read the prices and flows as the point-in-time figures they are labelled as, not as current ones.

Method note. Money figures are shown in US dollars, converted once at ₩1,345.06/$1, the USD/KRW close on September 6, 2026. Share prices show the dollar first with the won in brackets, because a reader checking a Korean quote screen needs the won. Where a filing states its own amount, we use the filing’s figure and do not re-convert. Market data is cut off at the settled close of Monday, September 7, 2026; September 8 trading is not included anywhere in this article.

---

🇺🇸 Who Owned Korea’s Reactor Designer When It Was Founded?

An American engineering firm did. KEPCO E&C was incorporated in October 1975 as Korea Atomic Burns and Roe, Inc., and Burns & Roe of New Jersey was its largest shareholder from day one.

This is not a colourful footnote. It is the first entry in the company’s own corporate-history section, and it reframes the entire “Korea enters America” story. The traffic ran the other way first. Korea did not acquire reactor design capability by buying it or stealing it; it acquired it inside a joint venture with an American firm, then bought the American partner out inside a year, then spent five decades turning the licence into a product.

Source: DART rcpNo 20260814001293, KEPCO E&C H1 2026 semi-annual report filed August 14, 2026, section I.2 — “changes in largest shareholder,” “changes in company name,” “changes in head-office location,” and “other material events.” Every row above is quoted from that section.

---

📜 What Had to Exist Before an American Reactor Order Was Even Possible?

A design certification. Under US law, an applicant can build a reactor by referencing a design the NRC has already certified as a rule — and only five such designs are in effect, of which the Korean APR1400 is one.

This is the single most load-bearing fact in the whole story, and it is almost never stated in coverage of the current talks. The reported plan — two of the first four units as Korea’s APR1400, the rest as Westinghouse’s AP1000 — is not an arbitrary political compromise. It is the shortlist. Those two designs are Appendix F and Appendix D to 10 CFR Part 52. A design that is not already an appendix cannot be referenced, and certifying a new one takes years.

The NRC’s own rulemaking says so directly. Its 2025 rule revising certification duration lists the five design certifications currently in effect: Appendix A (US Advanced Boiling Water Reactor), Appendix D (AP1000), Appendix E (ESBWR), Appendix F (APR1400), and Appendix G (NuScale).

Sources: the 2019 rule is “Advanced Power Reactor 1400 (APR1400) Design Certification,” direct final rule, 84 FR, May 22, 2019; the 2025 rule is “Revising the Duration of Design Certifications,” direct final rule, 90 FR, July 2, 2025. Both were retrieved as published PDFs from the US Government Publishing Office. ⚠️ A companion proposed rule was published the same day; it is a procedural backstop, not the operative rule.

Two consequences follow, and both are invisible if you only read the news.

First, the certificate is 40 years old before it expires, not 15. Under the rule that applied until September 2025, the APR1400 certification issued in September 2019 could only be referenced until 2034. A large reactor takes roughly a decade from licence application to first power. A programme discussed in 2026 would have run straight into the expiry of the design it was built on. The NRC removed that problem eleven months ago, and almost nobody noticed. The certification now runs four decades from its issuance date.

Second, the certificate does not belong to the company you can buy. The applicant named in the rule is KEPCO and KHNP. KEPCO E&C did the design work; the certificate sits with the utility and the plant operator. Korea Hydro & Nuclear Power is not listed at all. That gap between “who did the engineering” and “who holds the paper” runs through every layer of this cluster, and it is the same gap we mapped in Korea’s $150bn shipbuilding package.

A translation trap worth naming

The KEPCO E&C half-year report gives two different dates for what English coverage calls “NRC approval.” The corporate-history section records October 2018 and uses the Korean term 표준설계승인 — standard design approval. The business section, thirty pages later, says 2019 and uses 최종 설계인증 — final design certification. These are different regulatory objects. An approval is a safety finding on a design; a certification is a rule an applicant can reference in a licence application. Only the second one is what makes a US project buildable, and only the second one is in the Federal Register.

---

🏭 Has a Korean Company Ever Built for an American Reactor Before?

Yes. Doosan Enerbility supplied AP1000 equipment for Vogtle Units 3 and 4 under a contract with Westinghouse signed on May 8, 2008, scheduled to December 31, 2022, with a base value of $205.0M and cost-input progress of 99.94% at the reporting date.

That one row in a Korean contract schedule does more work than any amount of speculation about whether Korea “can” build in America. It has already happened. It was small, it was long, and it was as a supplier to Westinghouse rather than as a principal — which is exactly the position the current reporting describes for the AP1000 units.

The scale is the part worth sitting with.

Source: DART rcpNo 20260814003782, Doosan Enerbility H1 2026 semi-annual report, contract-progress schedule, base date June 30, 2026. Won amounts in that schedule are ₩4,928.9bn and ₩711.1bn respectively; dollar figures here are our conversion at ₩1,345.06/$1.

The completed American nuclear contract is about 5.6% of the size of the single Czech reactor-system contract signed seventeen years later. If you want a sense of what “Korea builds reactors in America” would need to become to matter financially, that ratio is the honest starting point. We take Doosan’s counterparty structure apart in detail in our Doosan Enerbility report.

---

⏱️ How Long Does “Reported” Take to Become “Signed”?

Seventeen months, on the only recent example we have. A July 18, 2024 report that Korea had won a ₩24tn Czech nuclear order produced twelve official filings across two companies saying nothing was confirmed, before contracts were finally signed in December 2025.

Korea has a disclosure instrument with no clean English equivalent: the 풍문 또는 보도에 대한 해명(미확정), a “clarification of a rumour or media report (unconfirmed).” When a listed company’s shares move on a story, the exchange requires the company to answer on the record — and if it cannot confirm, it must set a date by which it will answer again. That re-disclosure date is a public, machine-readable clock on a rumour. Nothing in US or European disclosure practice looks quite like it, and it is one of the most useful things a foreign investor can learn to read.

Here is the whole Czech clock for KEPCO E&C. Doosan Enerbility filed an almost identical chain on the same dates.

Sources: DART clarification filings rcpNo 20240718800133, 20240814801930, 20241113800566, 20250508800077, 20250605800155 and 20251204800421, all filed by KEPCO E&C and all chained to the same July 18, 2024 report; supply-contract filings 20251212801320 and 20251224800396. Doosan Enerbility’s parallel chain runs 20240718800148 · 20240814803363 · 20241113800917 · 20250507800616 · 20250605800156 · 20251204800389. Translations from the Korean are ours.

Three things in that table are worth more than the headline that started it.

The interval is the signal. A company that expects an answer soon promises to re-disclose in one month. A company that does not promises six. KEPCO E&C went 1 → 3 → 6 within four months of the original story. The clock itself was telling you, in August and November 2024, that nothing was close.

The reason for the delay was a French lawsuit, not Korean execution. Between May and June 2025 the contract was frozen by a Czech court injunction obtained by EDF and then unfrozen by the Czech Supreme Administrative Court. That is a full year of the seventeen months, and it appears in Korean company filings before it appears in most English coverage.

The wording change was the tell. On December 4, 2025 the standard formula — “nothing has been specifically decided” — was replaced by “currently under discussion.” That is the only substantive edit in eighteen months of otherwise identical text. The company simultaneously scheduled its next non-answer for June 2, 2026. The contracts were filed eight days later.

---

🔎 What Is Actually Signed With the United States Right Now?

Two contracts, both at Doosan Enerbility, both for gas turbines rather than reactors, and both with the counterparty and the value legally withheld — which still tells you each is worth at least $301.7M.

This is where a lazy version of this article would have gone wrong, and we nearly did. The expected finding was zero. Instead, a search of the report type an American order would have to appear as turned up two.

What we searched. We pulled every filing of type 단일판매·공급계약체결 — “single sale or supply contract concluded,” the mandatory disclosure a Korean listed company must make when it signs a material contract — made by KEPCO E&C (052690) and Doosan Enerbility (034020) between January 1, 2015 and September 8, 2026. That is 198 filings: 130 from KEPCO E&C and 68 from Doosan. We read the 판매·공급지역 (“sale/supply region”) field on every one. Two name the United States. Both are Doosan’s.

Source: DART rcpNo 20251217800425 (signed Dec 16, 2025) and rcpNo 20260306801165 (signed Mar 6, 2026). Both state the reason for withholding as “maintenance of business confidentiality,” both state that the contract value equals or exceeds 2.5% of latest revenue, and both give that revenue basis as ₩16,233,054,771,559 — FY2024 consolidated revenue. Both promise re-disclosure once the confidentiality ground lapses.

Korean disclosure rules let a company withhold a contract’s value and counterparty for commercial reasons — but not the fact of the contract, and not the threshold that triggered it. Because the obligation bites at 2.5% of latest revenue, the floor is arithmetic: each of these two contracts is worth at least $301.7M. Not an estimate of what they are worth. A minimum they cannot be below.

And on the design side, the absence is real

KEPCO E&C has no American order of any kind, and we checked it two ways. None of its 130 supply-contract filings since 2015 names the United States as the supply region. And in the 260,942 characters of its H1 2026 semi-annual report, the word 미국 (“United States”) appears exactly nine times — none of them a customer, an order or a contract. They are: the 2012 ENR ranking; the two NRC milestones; two sentences describing the US as a competitor in the world reactor market; two footnotes explaining that “PE” and “PMP” are American professional certifications held by staff; and two macro paragraphs on global nuclear demand. “Westinghouse” appears zero times.

What this method cannot see — and you should know it

An intergovernmental memorandum of understanding is not a disclosable event for a Korean listed company. Disclosure obligations attach to contracts the company itself signs, above a materiality threshold. So the absence of an American nuclear filing today is exactly what you would expect whether or not an MOU is signed on any given date. It tells you that no company-level contract exists yet. It tells you nothing about the negotiation.

This is the same blind spot that hid a lost C$60bn Canadian submarine competition from a Korean shipbuilder’s filings entirely: when the decision-maker sits outside the company, the disclosure system does not record it. Read filings for what has been signed, and read them knowing they are silent on what is being discussed.

---

📰 Where on That Clock Is the American Story?

At the very beginning — earlier than the Czech story was on July 18, 2024, because in 2024 a Korean-led consortium had already been named preferred bidder, and here nothing has been awarded to anyone.

What is actually reported, as of September 8, 2026: Seoul and Washington are discussing up to eight large reactors in the United States, with roughly $120bn of the $200bn of unassigned funds inside a $350bn investment package potentially directed at them; the first project in the package is a gas-fired combined-cycle plant at Encinal, Texas; and a joint investment statement is expected around September 18 — a date that has since passed with no announcement from either government. Reporting also suggests two of the first four reactors would be APR1400s and the rest AP1000s. Every one of those items is attributed to unnamed government and political sources, and the reporting itself states that the scale and the Korean companies’ roles remain undecided (Source: Seoul Economic Daily, September 8, 2026). We take each of those reported figures apart against the government’s own documents — and count the fourteen ministry clarifications that answer them — in our companion piece on what is actually signed in the US investment package.

Hold that against the Czech chain. In July 2024 there was a named preferred bidder, a named project, a named client and a court-tested tender — and it still took seventeen months and a French lawsuit to reach signature. The American package today has no award, no named counterparty for either Korean company, and no contract. On the only clock we have, it sits earlier than the point at which the Czech story began generating filings.

---

🏢 The Timeline Also Explains 2025’s Record Profit

KEPCO E&C’s best profit year in a decade was an office building it stopped using in 2015 — and the ten-year gap between vacating it and being paid for it is its own lesson in Korean filing time.

The company moved out of Yongin in August 2015 when it relocated to Gimcheon. It resolved to sell the building on April 1, 2022 and signed the contract on April 13, 2022. The buyer is Solum Co., Ltd., an electronics-components maker; the price is ₩95,755,500,000 excluding VAT, structured as a 10% deposit and six equal semi-annual instalments; the transfer reference date — the buyer’s final payment — was February 6, 2025 (Source: DART rcpNo 20250206000060, amended material-event report on tangible-asset disposal).

The gain lands in FY2025 under a line called 매각예정자산처분이익 — gain on disposal of assets held for sale — at ₩78,975,791,993, against FY2025 profit before income tax of ₩109,416,438,076 and full-year operating profit of ₩35,452,646,024 (Source: DART rcpNo 20260326001003, FY2025 annual report, consolidated statements — all three figures are full-year consolidated, not half-year).

Our arithmetic on those three filed figures: the building gain is 72.2% of pre-tax profit and 2.2 times full-year operating profit. In dollars at ₩1,345.06/$1 that is a $58.7M gain inside $81.3M of pre-tax profit, on $26.4M of operating profit.

One label trap, because it is the kind of thing that ruins a number: the same FY2025 statements also carry a line called 유형자산처분이익 — gain on disposal of property, plant and equipment — of ₩65,940,998. That is sixty-six million won, not seventy-nine billion. Two adjacent lines, both “disposal gain,” differing by a factor of about 1,200. The Yongin sale is the first one.

---

🌏 What Did Foreign Money Do While the Runway Was Being Talked About?

It sold. Across three windows ending at the September 7, 2026 settled close, foreign investors were net sellers of five of the six Korean nuclear-linked names we track, and the sixth is the worst performer of the group.

Flow figures below are derived from share counts, not currency. We multiply each session’s net foreign shares by that session’s close and sum — an approximation, not execution prices — then convert at ₩1,345.06/$1.

Source: our own database of settled Korea Exchange closes and daily investor flows, all windows ending at the settled close of September 7, 2026. Price changes are measured from the close of the session preceding each window. The 56-session window is the full depth of flow history we hold for these names on a common start date.

Read the columns against each other, because a single window invents a story. Over 20 sessions this looks like a nuclear rally: BHI +35.67%, KEPCO E&C +20.02%, Hyundai E&C +15.47%. Over 56 sessions the same names are flat to sharply lower, and four of the six are down more than 8%. The rally is a recent layer on top of a drawdown, not a trend.

The flow signs disagree, and that matters more than the levels. LS Electric (010120) shows +$70.8M of foreign buying over the 20 sessions from August 10 to September 7, 2026. Over the 56 sessions from June 18 to September 7, 2026 the same stock shows −$46.0M. Anyone writing “foreigners are rotating into Korean grid equipment” from the 20-session column would be describing a reversal of five weeks of selling, not accumulation. The one name with a consistent positive sign across both longer windows is Hyosung Heavy Industries (298040), at +$48.3M over the 20 sessions from August 10 to September 7, 2026 and +$57.2M over the 56 sessions from June 18 to September 7, 2026 — and it is the worst 56-session performer in the table at −23.09%.

And the move was one session. KEPCO E&C fell from ₩116,400 on August 31 to ₩107,600 on September 4, then closed at ₩122,300 on September 7 — a single-session gain of 13.66% that accounts for more than the whole five-session move. Doosan Enerbility did the same thing, closing +10.98% on September 7. Buying a five-day chart here means buying one headline. For the weekly version of this analysis, see our Foreign Flow Watch series, and for how to read these categories at all, our guide to foreign and institutional flows in Korean stocks.

---

📅 The One Date to Watch Has Passed, and the Answer Was No

On September 17, 2026, Doosan Enerbility filed its seventh and final answer about its NuScale SMR equipment business, and it was a denial: the company stated that the business discussion between NuScale Power and Standard Power described in the original report “is not currently proceeding.” The clock that had run for almost twenty-eight months stopped on a no.

On May 28, 2024, Korean media reported that Doosan would supply equipment for NuScale’s small modular reactors, under a headline that noted the shares had jumped 17%. Doosan answered on the record, and kept answering: six filings across May 2024, June 2024, September 2024, March 2025, September 2025 and March 2026, each of the type 풍문 또는 보도에 대한 해명(미확정) — clarification, undetermined. The text barely changed. Doosan is a strategic investor in NuScale Power and holds a right to supply a certain volume of equipment on projects NuScale wins — “however, nothing specific has been confirmed to date.” The March 2026 filing set the next answer for September 17, 2026 (Source: DART rcpNo 20260318800949, filed March 18, 2026, which lists the five prior filings in the chain).

It answered on the date it had promised. The filing is the confirming form — 풍문 또는 보도에 대한 해명 without the “undetermined” qualifier, the version a company must file when the question finally resolves — and it resolves as a denial. Doosan restates the equipment right and then says of the 2024 report: it “has confirmed that the business discussion between NuScale Power and Standard Power referred to in the above media report is not currently proceeding,” and therefore that the reported equipment supply “is not true as of now.” The filing lists all six prior undetermined answers it closes out (Source: DART rcpNo 20260917800618, filed September 17, 2026; translation ours).

Two things follow, and the second is the more useful one.

The interval was the signal here too. The gaps between the seven filing dates — May 28 and June 27, 2024, then September 26, 2024, March 25 and September 19, 2025, March 18 and September 17, 2026 — run one month, then three, then six every single time after that. A company that widens its clock to six months and then holds it there for two full years is not managing a near-term announcement. The Czech chain’s tell was the opposite motion: a narrowing of language eight days before signature. This chain never narrowed, and it closed on a denial instead.

And a denial on one American SMR story is not a verdict on the reported reactor package. NuScale is Appendix G and a small modular design; the APR1400 is Appendix F and a 1,400MW plant. Different design, different customer, different scale. What the September 17 filing does establish is the mechanism: when a two-year-old American nuclear headline finally resolved for a Korean listed company, it resolved in a filing, on a date the company had published in advance, and the resolution was no. That is a base rate worth carrying into the next headline, and it is not a forecast about any particular one.

---

📚 Lingo Check

🎯 Why It Matters for K-Export Stars

The wires had this story before we did and will have the next one first too. What they will not do is open a 260,942-character Korean filing and notice that it dates the same NRC milestone twice with two different regulatory words, or pull 198 supply-contract filings to find out that the only signed American business in this cluster is gas turbines with the price redacted, or point out that the last identical headline generated twelve official non-answers before it generated a contract.

That is the arbitrage. The facts in this article were public the entire time — most of them for years — in a language and a filing format that almost no foreign investor reads. Our job is to read them and hand you the timeline, not the excitement. If you want the company-level versions, we have deep dives on the designer and the manufacturer, and a broader map of how to actually buy Korean stocks and which Korean names trade as US-listed ADRs.

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 as of the dates stated and were taken from primary filings and settled market data; prices and flows change. Translations from Korean regulatory filings are our own. Do your own research and consult a licensed adviser before investing.