KEPCO Engineering Stock: The Multiple Moves the Wrong Way

Three nuclear engineering companies, three multiples, and only one of them moves in the direction you would expect. BWX Technologies in the United States trades at 43.96 times trailing earnings and 33.91 times forward. AtkinsRealis in Canada trades at 5.77 times trailing and 21.39 times forward. KEPCO Engineering and Construction, listed in Seoul under the code 052690, shows 45.89 times trailing, and every forward number I can build for it lands somewhere between 40 and 65. I spent a session on KEPCO Engineering stock trying to work out why two of those three multiples climb when you look ahead, and the answer turned out to be a building.

This is my journal entry on that company. I do not own it. What follows is the arithmetic I ran, where I got each number, and the reason I am not placing an order yet.

That second row is the whole piece. The screen says 45.89. My arithmetic says 116.6. Below I show both where the gap comes from and why I do not trust either number enough to act on it.

Contents16 min read

What sits under the KEPCO Engineering stock multiple

Start with the screen. On the August 5, 2026 close the shares were 102,500 won, market capitalisation was 3.9176 trillion won, and the trailing price to earnings ratio read 45.89. Multiplying 38,220,488 shares by 102,500 won reproduces the market capitalisation to the last won, so the share count and the price agree with each other. That much I could confirm without anyone’s help.

The rest of the screen did not hang together. Fiscal 2025 revenue was down 6.25 percent year on year. Operating profit was down 35.31 percent. And yet net income was up. Operating profit of 35.5 billion won, net income of 85.4 billion won. A company that earned 2.4 times more at the bottom of the income statement than it did from operations is telling you something, and it is usually not about operations.

The explanation was one line inside a February 13, 2026 (Fri) earnings review by Seong Jong-hwa at Korean broker LS Securities: a 79 billion won gain on the disposal of the old headquarters, booked in the first quarter of 2025. The same report put fiscal 2025 pretax profit at 105.6 billion won and net income at 82.4 billion won. So 74.8 percent of the pretax line was one property transaction. That gain is inside the fiscal 2025 net income figure, and the 45.89x on my screen is the share price divided by exactly that figure.

Two sources disagree, so I tested whether it matters

My data vendor (Kiwoom, priced off the August 5, 2026 close) records fiscal 2025 operating profit at 35.5 billion won and net income at 85.4 billion won. The LS report records 31.7 billion and 82.4 billion. A 3.8 billion won gap on operating profit and a 3.0 billion won gap on net income. I could not establish which is preliminary and which is final, so I stopped trying to settle it and asked a different question: does the gap change the answer?

It does not. On the vendor’s figures the trailing four-quarter multiple works out at 116.6x and the operating profit multiple at 110.4x. On the LS figures the same two calculations give 128.0x and 123.6x. Four numbers, one conclusion: the multiple has three digits under any of these inputs.

The two sources also agree somewhere useful. Reversing each one’s own decline rate gives fiscal 2024 operating profit of 54.88 billion won on the vendor path and 54.84 billion won on the LS path. They differ by 40 million won. So the disagreement is confined to last year and does not extend backwards, which reads to me like a preliminary versus audited distinction and not a definitional one.

Net income fell 78.6 percent in the quarter operating profit rose 194.8 percent

The size of the one-off shows up cleanly in this year’s first quarter. In its May 8, 2026 (Fri) consolidated disclosure the company reported revenue of 113.3 billion won, up 13.3 percent; operating profit of 14.1 billion won, up 194.8 percent; and net income of 14.1 billion won, down 78.6 percent. Converted to ratios, operating profit became 2.9 times what it was and net income became 0.21 times what it was, in the same quarter, at the same company.

Those two growth rates let me reconstruct last year’s first quarter. Reversing the 78.6 percent decline puts Q1 2025 net income near 65.89 billion won (my calculation). Reversing the 194.8 percent increase puts Q1 2025 operating profit near 4.78 billion won (my calculation). A quarter that earned under 5 billion won from operations and booked nearly 66 billion at the bottom line.

A second route to the same figure

I did not want to lean on a single reversal, so I rebuilt that quarter from unrelated inputs. Take the reconstructed operating profit of 4.78 billion won, add the 79 billion won disposal gain, and pretax comes to 83.78 billion won. The LS report’s annual pretax of 105.6 billion and net income of 82.4 billion imply an effective tax rate of 21.97 percent. Apply that rate and the quarter nets 65.38 billion won.

65.89 billion from the growth rates. 65.38 billion from the disposal gain and the tax rate. The two land 510 million won apart, a difference of 0.78 percent, using no shared inputs. When two unrelated calculations meet inside one percent, I treat the underlying facts as confirmed. That is the point at which I decided this was worth writing up.

Recomputing KEPCO Engineering stock across the last four quarters

Now the rebuild. Take fiscal 2025 net income of 85.4 billion won, remove the 65.89 billion won that belonged to Q1 2025, add the 14.1 billion won from Q1 2026, and the trailing four quarters produce 33.61 billion won. Divide by 38,220,488 shares for earnings per share of 879 won. Divide 102,500 won by that and the answer is 116.6 times.

Why the screen still says 45.89 is a question of accounting periods, not of anyone being careless. Vendor trailing multiples in this market are commonly built on the completed fiscal year, so the January to March 2025 quarter is still inside the earnings base. On a rolling twelve month view that quarter dropped out when the May 8, 2026 (Fri) disclosure landed. The price on my screen is current. The earnings base under it is not.

Because the 116.6x figure depends on a reversal, I built a third path that depends on nothing at all. Market capitalisation of 3.9176 trillion won divided by fiscal 2025 operating profit of 35.5 billion won is 110.4 times. No tax assumption, no reconstruction, no quarterly surgery. That 110.4 and the 116.6 sit close together, and their agreement is the single most reassuring thing I found all session.

The most punitive version

For completeness: strip the 79 billion won gain out of LS’s 105.6 billion won pretax and you are left with 26.6 billion won, which at the 21.97 percent effective rate nets 20.76 billion won, or 543 won per share. That is 188.7 times. I kept it out of my tables because it assumes the average effective rate applied to a property disposal, and disposal gains are frequently taxed on a different basis. I mention it so the range stays honest instead of flattering.

The backlog is real, and Dukovany has started to land

None of the above is a claim that the business is weak. Doubting a multiple and doubting a company are separate activities, and I only mean the first.

Kim Tae-hyoung at Mirae Asset Securities, initiating coverage on January 12, 2026 (Mon), put 2025 new orders at 1.7346 trillion won, a record. Within that, the Czech Dukovany units 5 and 6 brought 1.2508 trillion won of architect engineering work and 373.4 billion won of nuclear steam supply system design, or 1.6242 trillion won together. That report describes the pricing as roughly 56 percent above the comparable Barakah scope in the United Arab Emirates. LS Securities frames the same work as about 1.6 trillion won, near 6 percent of a 26 trillion won total project, running from 2026 through 2038.

Revenue recognition has begun. Coverage of a May 11, 2026 (Mon) review by Cho Jae-won at Korean broker Kiwoom Securities reports first quarter nuclear division revenue of 103.7 billion won, up 38 percent, which is 91.5 percent of the 113.3 billion won consolidated total (my calculation). Czech work and the domestic Shin-Hanul units 3 and 4 are both contributing. The first quarter operating margin came out at 12.44 percent (my calculation) against 6.84 percent for full-year 2025.

So the recovery is not hypothetical. It is in one quarter of reported numbers. My question is narrower: how far ahead of that recovery has the price already run?

Four brokers on KEPCO Engineering stock, and a 1.9x spread

Korean sell-side coverage of this name is unusually scattered. All four below are named analysts at named firms, and the figures are the valuations each of them published. Doosan Enerbility showed the same two-to-one spread elsewhere in the Korean nuclear chain, though there the split opened up after the share price had already halved.

The lowest and highest sit 1.9 times apart on the same company, the same disclosures, and in three cases the same week. That spread is itself a piece of information. And the market price of 102,500 won is 15.3 percent below even the lowest of the four.

The estimates split too

For 2026, Kiwoom models 595.3 billion won of revenue and 74.4 billion won of operating profit; LS models 621.8 billion and 66.3 billion; Mirae Asset models 655.6 billion and 73.5 billion. Revenue estimates span 60 billion won while operating profit estimates sit in a narrow band. One detail I checked: reversing Kiwoom’s numbers against fiscal 2025 gives revenue growth of 14.75 percent and operating profit growth of 109.6 percent, which matches the “+15%” and “+110%” quoted in the coverage. My vendor’s fiscal 2025 revenue and operating profit therefore agree with what the sell-side is working from, which is a useful independent check on data I otherwise have to take on trust.

One firm’s multiple, another firm’s earnings, and what KEPCO Engineering stock costs

The Mirae Asset initiation interested me for its method more than its conclusion. The report states that it applied a target multiple of 33.3 times to a 2030 estimated earnings per share discounted back four years, with 33.3 described as the historical average price to earnings ratio excluding the pandemic period. The applied earnings per share is 3,640.3 won, and 3,640.3 multiplied by 33.3 gives 121,222 won, which reconciles with the published valuation.

So even the most cautious of the four numbers is built on 2030 earnings pulled forward. The report’s reference price was 97,400 won on January 9, 2026 (Fri); the August 5, 2026 (Wed) close of 102,500 won is only 5.24 percent above that (my calculation), though the 250-day intraday high of 198,000 won records that the price went a long way in between and came back.

That set up the calculation I actually wanted to run. Take the 33.3 times that Mirae Asset itself calls a historical average, and apply it to the 71.6 billion won of 2027 net income that LS Securities estimates. The result is 71.6 billion multiplied by 33.3 and divided by 38,220,488 shares, or 62,382 won per share. The current price is 64.3 percent above that.

I am mixing two houses deliberately and I want that on the record: the multiple belongs to Mirae Asset and the earnings belong to LS. What the combination says is that a multiple the sell-side calls normal, applied to sell-side 2027 earnings, does not reach today’s price. Mirae Asset could publish 121,000 won because it went out to 2030 to find the earnings.

Forward multiples confirm the shape. LS’s 2026 net income estimate of 60.5 billion won is 1,583 won per share and 64.8 times (my calculation); its 2027 estimate of 71.6 billion is 1,873 won per share and 54.7 times. Mirae Asset’s own 56.3 times for 2026 and 38.0 times for 2027 were struck against 97,400 won, so at 102,500 won they become 59.25 and 39.99 (my calculation). Every route lands the forward multiple in the high forties to the mid sixties.

The direction of the multiple, in three companies

Back to where this piece opened. I picked two overseas comparisons. BWX Technologies makes nuclear components and provides nuclear services in the United States. AtkinsRealis, listed in Toronto and quoted in Canadian dollars, owns the CANDU reactor design intellectual property and sells engineering. The second is the closer business analogue.

The column to read is the last one. BWXT’s multiple falls as you look forward, which is what happens when earnings are growing into the price. KEPCO E&C’s rises. AtkinsRealis leaps. A rising ratio means the trailing earnings are the inflated end of the comparison.

AtkinsRealis makes the case vividly. Its trailing twelve month net income is up 762.5 percent and its net margin sits at 23.14 percent (my calculation), a number no engineering firm produces from engineering. That compressed the trailing multiple to 5.77, and the forward figure of 21.39 is where the market thinks the real earnings power is. The direction matches what I found in Seoul, and the magnitude is larger. Design and engineering houses carry one-off items, whether asset disposals or claim settlements, that are large next to their revenue. Trailing multiples in this corner of the market are contaminated more easily than in most.

Which means the closeness of 45.89 to BWXT’s 43.96 is a coincidence, and reading across from one to the other would be a mistake. For scale: BWXT’s market capitalisation of 15.55 billion dollars is 5.7 times KEPCO E&C’s (my calculation), its price to sales ratio is 4.43 and its net margin 10.12 percent, both reversed from the same page.

Where my KEPCO Engineering stock case breaks

I went looking for arguments against everything above. Six of them held up, and two are strong.

• Stripping a one-off always produces an ugly multiple. Divide by the earnings of the year before a turn and any company prints three digits. Mirae Asset called 2025 the last pause before the turnaround, and on that reading 116.6x is evidence that the ratio is uninformative, not evidence that the shares are expensive.
• The first quarter pace is respectable. Annualising 14.1 billion won of operating profit gives 56.4 billion, which is 85 percent of LS’s 66.3 billion estimate for 2026 and 76 percent of Kiwoom’s 74.4 billion. Revenue annualises to 453.2 billion against LS’s 621.8 billion, only 73 percent, but design work is recognised unevenly across the year. Q4 2025 operating profit of 22.6 billion won was 71.3 percent of that full year on the LS basis (my calculation).
• The price has already fallen a long way. The August 5, 2026 (Wed) close of 102,500 won is 48.23 percent below the 250-day intraday high of 198,000 won and 32.94 percent above the low of 77,100 won (both my calculation off that same close). The three month return is −44.71 percent. It is fair to argue the market has priced the problem I am describing.
• Price to book of 6.17 cuts the other way. My checklist scores 71 with price to earnings and price to book the two failing items out of seven. This is not a stock that is cheap on assets. It is not cheap on earnings either, which at least is consistent.
• There is a cash conversion caveat from a separate source. Korean business daily Hankyung notes the company’s debt-free posture but flags unbilled receivables and collection delays as cash flow risks. Design services invoice on milestones that do not line up neatly with revenue recognition.
• The dividend record is genuinely good. 283 won for 2022, 515 for 2023, 999 for 2024 and 1,347 for 2025, three consecutive increases and a three year compound rate of 68.21 percent (my calculation). The caution is that 51.48 billion won of 2025 dividends against 85.4 billion of net income is a 60.28 percent payout (my calculation) computed on earnings that include the building. Yield on the August 5, 2026 (Wed) close is 1.31 percent.

KEPCO Engineering stock: I am waiting for a ruler, not a price

I hold none of this company and I did not open a position this session. The reason differs from my usual one.

Normally I wait for a price to arrive. Here I am waiting for something else. The instrument that measures this company is broken at the moment. The 45.89x on my screen contains a property sale. The 116.6x I built divides by the earnings of the year before a recovery, which overstates in the opposite direction. Neither measures the business as it stands today, and arguing about length while the ruler is bent is the wrong order of operations.

The repair date is fixed

The useful part is that this fixes itself on a schedule. Once four quarters have accumulated from the first quarter of 2026, the quarter containing the disposal gain falls out of every basis, fiscal year or rolling. The multiple that emerges then is the first one built purely on operations. Until that point I am not copying any number off any screen.

What I check while I wait, and what would prove me wrong

Two things. First, whether the quarterly operating margin holds near the 12.44 percent of Q1 2026, against 6.84 percent for full-year 2025; one quarter at 1.8 times the annual rate does not settle it. Second, whether Dukovany revenue recognition keeps pace with the sell-side models, since 1.6242 trillion won spread from 2026 to 2038 can arrive faster or slower than the 66.3 to 74.4 billion won 2026 estimates assume.

If both check out and the multiple is still where it is today, then the ruler was not broken and the market was measuring correctly without one. I will have paid a higher price for waiting. That condition is not written in my favour, which is the only kind worth writing down.

I have watched the same shape from a different angle before. HD Hyundai Marine Engine saw operating profit jump while the share price fell 45 percent, and I put it on a watchlist pending one more quarter of numbers. With HD Hyundai Electric I did buy a 40 percent correction, because there the backlog was long-dated and the selling was macro. The difference in this case is that I cannot yet tell what the correction is measured against.

Questions and notes for readers outside Korea

How would I even own this?

KEPCO E&C trades on the KOSPI, the senior board of the Korea Exchange in Seoul, where the country’s larger listed companies sit; the KOSDAQ is the separate junior venture board. I could not confirm an American depositary receipt for this company. Practical access means a broker with direct Korea Exchange market access, settlement in won, and Korean dividend withholding before anything reaches you. The parent, Korea Electric Power Corporation, holds 53.06 percent alongside one related party, so free float is a good deal smaller than the market capitalisation suggests, and foreign ownership stands at 15.44 percent.

Why would my screener disagree with this article?

This is the caveat specific to this piece, and it is the one that would have cost me money. International screeners pull Korean trailing earnings from the completed fiscal year, so a foreign investor screening on price to earnings sees 45.89 and reads a nuclear design house at a growth-stock discount to BWXT. The disposal gain is disclosed, but it is disclosed in Korean, in a broker note, in a single line. The screener does not know the line exists. Nothing about this is a data error, and no vendor has done anything wrong. It simply means the number that reaches an overseas desk is the one furthest from the operating reality, and it will stay that way until the fiscal year rolls.

What did I deliberately leave out?

I discarded the vendor’s total equity, operating cash flow, free cash flow, EBITDA and three year revenue growth fields. Total equity of 562.5 billion won conflicts with book value per share of 16,621 won times the share count, which gives 635.2 billion. The EBITDA field carries a value well above fiscal 2025 revenue of 518.8 billion won, which cannot be true by definition. The earnings per share field is computed from price divided by the ratio, so using it to rebuild the ratio goes in a circle. I did verify price to book directly (102,500 divided by 16,621 gives 6.1669 against a published 6.17), which confirms the book value figure is struck on this close, and then set the rest of the equity family aside.

What is the one habit this changed?

When a multiple looks wrong I have always interrogated the price first, asking whether the market is right about the company. This time I opened the earnings underneath and the problem was sitting there in a single line item. It has taken me a good many of these entries to try that order, and in hindsight it should have been obvious.

Prices and multiples reflect the August 5, 2026 (Wed) close as checked at the time of writing. This entry publishes later than it was written, so figures can differ from live quotes and should be read with that lag in mind. The Korean won is the reference currency throughout; dollar figures are approximate, at roughly 1,424.5 won per dollar on that same date per Seoul Finance, which puts the market capitalisation near 2.75 billion dollars and fiscal 2025 revenue near 364 million dollars. AtkinsRealis figures are quoted in Canadian dollars as published and were not converted. Second quarter 2026 results had not been disclosed when this was written; once they are, every trailing four-quarter figure here has to be rebuilt. Primary sources: Bloter on the May 8 disclosure, the LS Securities review, the Mirae Asset initiation, Newsquest on the Kiwoom review, Daum News on the Samsung Securities note, and Hankyung on Dukovany and cash conversion. All are Korean-language sources and the wording here is my rendering, not a quotation of their English.