On 31 July 2026, Doosan Corporation (KRX:000150) signed to buy 70.61% of SK Siltron from SK Inc. for ₩2,300bn. Four independent directors attended the board meeting and none were absent. Closing is scheduled for 31 January 2027, subject to regulatory approvals.
SK Siltron makes silicon wafers — the polished discs of single-crystal silicon that every semiconductor is built on. It is one of a small handful of companies worldwide that can do this at volume, alongside Japan's Shin-Etsu and SUMCO, Taiwan's GlobalWafers and Germany's Siltronic. For a conglomerate that has spent the last five years rebuilding after a near-death restructuring, buying into that club is a significant statement.
It is also, on the numbers in the filing, an expensive statement.
Doosan attached SK Siltron's summary financials, and they are the most useful part of the disclosure.
FY2023: revenue ₩2,025.6bn, net income ₩234.0bn. FY2024: revenue ₩2,126.8bn, net income ₩248.0bn. FY2025: revenue ₩2,057.5bn, net loss ₩293.6bn. EY Han Young signed unqualified opinions on all three.
So revenue has been flat within a 5% band for three years, and profit swung by ₩542bn between FY2024 and FY2025. Equity fell from ₩2,243.2bn to ₩1,933.4bn across that year, and total liabilities stood at ₩3,661.4bn at the end of FY2025 against assets of ₩5,594.8bn.
Price the deal against that. ₩2,300bn for 70.61% implies an equity value near ₩3,257bn for the whole company, or roughly 1.7 times FY2025 book and 1.6 times revenue. On last year's earnings there is no multiple, because there were no earnings.
Deloitte Anjin ran a valuation between 23 January and 31 July 2026 and put the 70.61% stake between ₩2,089.9bn and ₩2,473.6bn as of 31 March. The agreed ₩2,300bn sits comfortably inside that. Worth noting the range is only 18% wide, which is unusually disciplined for a Korean fairness opinion and suggests the appraiser had a genuine view rather than a corridor.
The most revealing item is buried in the restructuring section. SK Siltron's own board resolved on 16 July 2026 — two weeks before the sale agreement — to liquidate SK Siltron USA, Inc. and the companies it directly or indirectly controls. Those are the silicon carbide operations. After the share purchase agreement, the liquidation proceeds through a committee that Doosan participates in.
Silicon carbide wafers were the great growth story of 2021-23, driven by electric vehicle power electronics. SK acquired that business in the United States and invested in it. It is now being wound up rather than sold, which is a harsher outcome than a discounted disposal and tells you what the assets were judged to be worth.
Doosan is therefore buying the conventional silicon wafer business with the SiC problem being removed by the seller on the way out. That is the right structure for a buyer. It also means the ₩293.6bn FY2025 loss includes whatever the SiC operations contributed, so the continuing business is better than the headline loss implies — by how much, the filing does not say.
Beyond the ₩2,300bn base price, SK Inc. can collect more. The terms are laid out in detail.
First, if SK Siltron exceeds annual EBITDA thresholds from 2027 through 2034, SK receives 40% of the excess, applied at the 70.61% stake. The thresholds: ₩0.89tn in 2027, ₩1.0tn in 2028, ₩1.1tn in 2029, ₩1.2tn in 2030, ₩1.3tn in 2031, ₩1.4tn in 2032, ₩1.55tn in 2033, ₩1.7tn in 2034.
Sit with the 2027 figure. SK Siltron produced ₩2,057.5bn of revenue in FY2025 and lost money. An EBITDA of ₩890bn on revenue of that order would be a margin above 40%. Wafer makers do reach those margins at cycle peaks — depreciation is enormous in this industry, so EBITDA and net income diverge sharply — but it requires the market to be nothing like it was last year.
Second, ₩25bn per item, up to four items, if specific product qualifications with specific customers complete between 2026 and the end of June 2029. Qualification at a wafer customer is a multi-year technical process and passing it is worth far more than ₩25bn, so this is a modest and sensible incentive.
Third, if the assets of the liquidating US subsidiaries and other earmarked assets sell above book value, SK shares the gain.
Read together, the earn-out is SK saying it does not want to sell the recovery, and Doosan saying it will not pay for a recovery it has not seen. That is a reasonable meeting point. It also means Doosan's upside is capped in exactly the scenario that would justify the price.
Doosan disclosed the deal at 6.98% of its total assets and 18.8% of its equity, using FY2025 figures of ₩32,927.8bn and ₩12,234.3bn. Funding will come from cash on hand and borrowings, with the borrowing terms not yet settled — the filing leaves that table blank and promises an amendment.
There is more to come. Doosan says it will separately negotiate for the remaining 29.39% of SK Siltron, and that with that stake and the earn-out included, the total consideration could exceed 10% of its total assets.
Doosan is not entering this from a position of financial ease. Total liabilities were ₩22,432.6bn at the end of June against equity of ₩13,907.6bn, a ratio of 1.61. Operating cash flow was negative ₩732.3bn across the first half of 2026 on a cumulative basis, and financing activities brought in ₩1,047.1bn over the same period. The group has strong seasonality — FY2025 operating cash flow was positive ₩981.9bn for the full year — but it is currently a borrower, and it has just committed to another ₩2.3tn.
The bull argument is the obvious one and it is not weak. Silicon wafers are a genuine oligopoly with enormous barriers to entry: capital intensity, decade-long customer qualifications and process knowledge that cannot be hired. Companies like this are almost never for sale. When one is, it is because the seller needs money or the cycle is at a trough, and both appear to be true here.
Buying a top-five global position in a structurally short industry, at 1.7 times book, in the year it lost money, with the problem division being liquidated by the seller, is a textbook cyclical acquisition. If wafer demand recovers with the semiconductor capacity buildout everyone is forecasting, ₩2.3tn will look cheap and the earn-out will be a happy problem.
Doosan also has a plausible claim to knowing this market. Its electronics business, which makes copper clad laminate for printed circuit boards, has been growing fast on AI server demand — first-quarter 2026 revenue of ₩702.3bn, up 53% year on year. That is a materials business selling to the same customers. Not the same technology, but not a blind purchase either.
SK Siltron's FY2026 result, which Doosan will have to disclose in connection with the closing. If the loss narrows sharply or reverses, the cycle has turned and the ₩2,300bn was well timed. Another year of losses at this scale would mean Doosan bought a business consuming equity, financed with debt, at a moment when its own operating cash flow was negative.
The second thing is the borrowing table Doosan promised to amend once the financing is fixed. Amount, tenor, and any collateral. A ₩2.3tn acquisition funded largely with short-term debt, at a base rate the Bank of Korea has just raised to 3.00%, is a materially different proposition from one funded with long bonds or an equity raise.
Third, whether the remaining 29.39% gets bought and at what price. Doosan has said it intends to try. The price it pays for the tail tells you what it really thinks the whole thing is worth.
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.