On August 31 Samsung Heavy Industries Co., Ltd. (KRX:010140) filed a contract revision. The value went from ₩423.7bn to ₩639.9bn, a rise of 51%. The customer is Samsung Electronics. The work is the finishing of a fab building at the Pyeongtaek complex in Gyeonggi province, the P4 Phase 2 project. Nothing about it floats.
That is worth sitting with. Samsung Heavy is known as the offshore yard at Geoje, the company that has built more floating LNG units than anyone. Almost every article about it, including three we published last week, argues about the order book for ships and offshore platforms. Meanwhile the single largest contract revision the company filed this year is a construction job for its own group's chipmaker.
The shares closed at ₩21,650 on September 3, up 8.58%. That move followed shipbuilding news, not this. But the building business deserves more attention than it gets.
The filing carries a long history in its footnotes. The original award was signed on July 5, 2023, and work began on July 15. The first amendment came in December 2023, and that is when the value first crossed the threshold that forces a public filing. Three more amendments followed, in October 2024, November 2025 and March 2026. This August revision is the fifth.
The stated end date is December 31, 2027, measured to completion. The company notes that the date can move with construction progress.
So this is not a new order. It is a four-year job that keeps getting bigger. The March 2026 version was ₩423.7bn. Five months later it is ₩639.9bn. That is ₩216.1bn of extra scope inside a single year, on one building.
Samsung Electronics has been rebuilding its Pyeongtaek plan around memory demand. Fab shells that were paused have been restarted. A finishing contractor's scope grows when the owner decides the room needs to become a working cleanroom sooner. The filing does not say why the number moved, and I would not pretend to know beyond that.
The disclosure prints one ratio: the contract equals 10.8% of recent revenue. It then footnotes what "recent revenue" means. It is the FY2022 consolidated figure, ₩5,944.7bn.
Korean contract disclosures fix that denominator at the last audited annual revenue available when the original filing was made. Amendments keep the same base. So the ratio ages, and by 2026 it is describing a company that no longer exists.
Samsung Heavy's FY2025 revenue was ₩10,650.0bn. Against that, ₩639.9bn is about 6.0%. Against the ₩19.1 trillion market capitalisation at Thursday's close it is around 3%.
US readers should treat every percentage in a Korean contract filing this way. Check the footnote for which year the denominator came from. On a company whose revenue has nearly doubled in three years, the printed number can be off by a factor approaching two.
Two lines in the filing matter more than the headline amount. There is no advance payment. And the payment condition is collection according to construction progress.
Shipbuilding does not work that way. A yard collects staged instalments, usually with money up front at signing, and those receipts sit in current liabilities as customer advances until the vessel is delivered. That is why Samsung Heavy's current liabilities were ₩11,034.7bn at June 30 against total assets of ₩17,959.4bn. The balance sheet looks levered because customers have prepaid.
A progress-billed construction contract does the opposite. Samsung Heavy funds the work, invoices the milestone, then collects. It consumes working capital instead of supplying it. The amounts here are small enough that it will not show up as a swing in the cash flow statement. But the mechanics are worth knowing, because the two businesses inside this company have opposite cash shapes.
There is a compensation for that. Building for an affiliate on progress billing carries almost no counterparty risk and no currency risk. Samsung Heavy's ship contracts are dollar-denominated against owners in Greece, Africa and the Gulf. This one is won, paid by a company in the same group.
Look at what Samsung Heavy's earnings have done. Operating income was negative in every year from 2015 through 2022. FY2023 turned it positive at ₩233.3bn. FY2024 reached ₩502.7bn. FY2025 came in at ₩862.2bn. The first half of 2026 has produced ₩598.1bn on ₩6,133.0bn of revenue, a margin of 9.8%.
The recovery is real, and it is mostly ships and offshore. But the offshore order book is the definition of lumpy. A single floating LNG award can be a third of a year's target. When it slips a quarter, the whole plan slips.
Construction revenue is the opposite kind of income. It is small, it is domestic, and it books steadily against progress. On a yard that spent nine straight years losing money, having some revenue that does not depend on an oil major's investment committee is not nothing.
The company is also spending on itself in ways that have nothing to do with hulls. It announced an ₩86bn expansion of its Busan marine research centre this week. Purchases of property and equipment were ₩216.3bn in FY2025 against a net plant balance above ₩5.1tn, so capital spending remains light. The R&D building is a rounding error on the balance sheet and a signal about where management thinks the next decade of value sits.
Start with scale. Six percent of revenue spread across four and a half years is roughly 1.3% of annual sales. That will not move an earnings model. If the argument for owning Samsung Heavy is a diversified revenue base, this contract does not deliver one.
Then the related-party problem. The counterparty is disclosed as a group affiliate. Korea's Fair Trade Commission watches intra-group contracts closely, and large business groups file a separate quarterly disclosure listing them. Samsung Heavy filed one on August 30. Work awarded inside a group is not evidence that a company can win the same work outside it. Until Samsung Heavy shows a comparable building contract with a customer that is not Samsung, treat this as captive revenue.
Amendments cut both ways. A contract revised upward five times can be revised downward. The company says explicitly that the completion date may change with progress. Fab schedules follow memory pricing, and memory pricing is not stable.
And the contract expires. December 2027 is close enough that the replacement question is live now, not later. If nothing follows it, the construction line simply shrinks out of the numbers.
Finally, the accumulated deficit is still there. Retained earnings were negative ₩1,276.9bn at June 30. Profits are repairing that at roughly ₩450bn a year on the recent run rate. Nothing in this contract changes that arithmetic.
Watch for a new building contract filing with a non-affiliate customer. That single event would tell you whether Samsung Heavy has a construction business or a group assignment.
Second, watch the segment disclosure in the next annual report. Revenue split by segment will show how large the non-shipbuilding line has grown while nobody was looking.
Third, watch for further amendments to this contract before December 2027. Another upward revision would suggest Pyeongtaek P4 is accelerating. A downward one would say the opposite, and it would arrive as a two-line filing that most people ignore.
kstock reads DART every morning and writes up what moved — the contract, the buyback, the number that does not add up. The daily post and a Saturday roundup, by email.