A Heavy-Machinery Conglomerate Buys Into Silicon

On July 31, Doosan Group closed its purchase of a 70.6% stake in SK Siltron, the wafer-making arm SK Group carved out of its chip empire, for roughly 2.3 trillion won, or about $1.6 billion 1. It is the largest acquisition Doosan has made since it bought construction-equipment maker Bobcat two decades ago 5, and it hands a group best known for excavators, turbines and circuit-board materials control of one of the small number of manufacturers that supply the overwhelming majority of the world's silicon wafers 4.

SK Group chairman Chey Tae-won kept his own 29.4% personal stake out of the transaction for now, a detail that lets SK retain an option to re-enter later even as it books cash from a business that no longer fit its post-restructuring focus on SK Hynix 1. For Doosan, the wafers complete a chain it has been assembling piece by piece: copper-clad laminate substrates through its Electro-Materials business, back-end chip testing through Doosan Tesna, and now the raw silicon itself 4 6.

The Financing Is the Real Story

The headline price understates the leverage behind it. In May, Korea Development Bank and Woori Bank arranged a 2.5 trillion won syndicated loan package for the deal, split between 1 trillion won to fund the acquisition and 1.5 trillion won to refinance SK Siltron debt that comes due automatically once ownership changes 2. That is state-adjacent credit, not Doosan's own balance sheet, doing most of the work. Korea Eximbank has separately signaled it may extend support to other Doosan units tied to the same buildout, including Doosan Tesna and Doosan Robotics, which points to financing here running wider than one transaction 6.

The contract also carries an earn-out clause that could push the final price well above the headline figure: if SK Siltron's earnings beat agreed milestones, SK collects an amount equal to 40% of the excess EBITDA multiplied by its 70.61% former stake 3. Enterprise value on the deal has been put at roughly $3.7 billion once that structure and existing debt are counted 8, well above the $1.6 billion equity check alone.

Why Wafers, Why Now

SK Siltron is not a market leader. Japan's Shin-Etsu Chemical and SUMCO dominate global wafer output by volume, and SK Siltron sits alongside GlobalWafers and Siltronic as one of a handful of manufacturers whose combined capacity accounts for nearly all of the world's supply 4. Its specialty is thin-film and power-device wafers, and it reached mass production of 300mm silicon-carbide wafers for power chips in 2024, a niche that matters more as AI data centers pull in more power electronics alongside logic chips.

Doosan has set a target of 3 trillion won in SK Siltron sales by 2031 and says it wants to become the world's second-largest supplier of memory-chip wafers, up from its current standing outside the top tier 4. That is an aggressive climb: TSMC, Samsung Foundry and the major memory makers requalify wafer suppliers only after multi-year validation cycles, and incumbents have decades of process know-how that a new owner cannot simply buy alongside the plant.

Completing an AI Infrastructure Chain, Not Just a Chip Bet

The Siltron purchase reads differently next to Doosan's other 2026 moves. In June, Nvidia and Doosan expanded a partnership spanning Doosan Robotics, which is applying physical-AI simulation-to-real workflows to industrial robots; Doosan Enerbility, which is pitching turbines, fuel cells and small modular reactors at AI-factory power demand; and Doosan Corporation Electro-Materials, which supplies copper-clad laminate for AI server boards 7. Wafers were the missing input. Doosan is not trying to out-execute TSMC on leading-edge logic; it is assembling a portfolio that touches power generation, substrates, testing and now base materials for the AI buildout across Asia, wherever that capacity ultimately gets built 6.

Korean brokerages have read the deal as earnings-accretive for Doosan's corporate entity once SK Siltron's cash flow consolidates, which is a milder claim than a bet on wafer-market share alone; it works even if the 2031 target slips.

What It Means for Investors

For Asia-focused portfolios, the deal is a signal about where second-order AI capital is flowing: not only into fabs and GPU makers, but into diversified industrials willing to lever up on state-bank credit to buy scarce materials capacity. That is a lower-multiple, higher-leverage way to gain AI exposure than buying TSMC or SK Hynix directly, and it carries real risk if AI capex growth cools before 2031 or if SK Siltron's biggest customers, including SK Hynix itself, diversify their supplier base now that a former sister company controls the plant.

Two signals are worth tracking from here. Whether Chey's retained 29.4% stake eventually sells into the same buyer would confirm SK views the wafer business as fully non-core rather than an asset it still wants a claim on. And whether Korea Eximbank follows through on financing Doosan Tesna and Doosan Robotics would confirm Seoul is treating Doosan's roll-up as adjacent to national industrial policy, not just one conglomerate's balance-sheet bet on where AI capital wants to sit next.