Doosan Corporation (KRX:000150) is best known outside Korea for two things it does not directly make: the Bobcat skid-steer loaders parked on American construction sites, and the turbines and reactor components its Enerbility affiliate builds for power plants. Both sit under the holding company.
The business growing fastest is neither. It is a laminate.
Doosan's electronics unit reported first-quarter 2026 revenue of ₩702.3bn, up 53% year on year and 10.8% on the previous quarter — a record. The company attributed it to sustained demand centred on data centres and semiconductors, and said the second quarter would extend the growth through existing AI accelerator and memory-related products plus new applications including optical modules.
Against a group that grew revenue 10.6% in the first half, a division growing 53% changes the shape of the whole thing.
The product needs a sentence of explanation because it is invisible in the finished machine.
A printed circuit board starts as copper clad laminate: sheets of glass fibre bonded with resin, with copper foil pressed onto the surface. The circuit is then etched out of the copper. Ordinary laminate is a commodity made by dozens of suppliers. High-frequency, low-loss laminate is not.
The reason is physics. Signals moving between an AI accelerator and its memory run at very high frequencies, and ordinary resin absorbs and distorts those signals across the distances involved on a server board. Low-loss grades use different resin chemistry to keep the signal intact. They are difficult to make consistently, they take years to qualify at a customer, and they sell at a large multiple of standard laminate.
That is why the mix disclosure matters more than the revenue line. Doosan said high-value products would account for around 81% of electronics revenue in the first half, up roughly ten percentage points year on year. A ten-point mix shift toward premium grades does more for profit than the revenue growth does.
The leading suppliers in this segment have historically been Japanese and Taiwanese — Panasonic and Mitsubishi Gas Chemical are generally regarded as the technical benchmarks in the lowest-loss grades. Doosan competing there at scale is a genuine achievement and not one the group's history would have predicted.
One caution on figures circulating for the first half. Reporting around the first-quarter result cited an electronics business first-half revenue expectation of ₩1,277bn, up 45.2%. Set against the ₩702.3bn actually delivered in the first quarter, that implies second-quarter revenue near ₩575bn — a sequential decline of 18%, which contradicts the company's own guidance of continued growth.
One of those two things is wrong, and I cannot tell which from outside. Take the segment revenue from the half-year report filed on 14 August rather than from either figure quoted secondhand. The first-quarter number of ₩702.3bn is well attested; the half-year total is not.
Consolidated results support the idea that mix is doing work. First-half 2026 revenue was ₩10,646.4bn against ₩9,645.1bn, up 10.6%. Operating income was ₩829.3bn against ₩556.2bn, up 49.1%. Second-quarter operating income alone reached ₩488.4bn against ₩357.8bn a year earlier.
Gross margin in the second quarter was 20.4% against 17.5% a year before. For a group whose revenue is dominated by construction equipment and power plant engineering, three points of gross margin in a year is a large move, and a fast-growing high-margin materials business is the most plausible explanation available from the outside.
Group operating margin was 8.7% in the second quarter against 5.4% for FY2025. Still modest in absolute terms. Improving quickly.
On 31 July, Doosan agreed to buy 70.61% of SK Siltron from SK Inc. for ₩2,300bn, closing around 31 January 2027. SK Siltron makes silicon wafers — the substrate every chip is built on — and is one of five companies globally that does so at scale.
Viewed as a conglomerate buying an unrelated asset, that deal looks like empire building. Viewed alongside the laminate business it looks like something else: a company assembling positions in semiconductor materials, at two different points in the stack, selling to overlapping customers.
Whether the strategy is coherent depends on how much genuinely transfers. Wafer manufacturing and resin laminate manufacturing share almost no process technology. What they share is a customer set, a qualification culture, and an exposure to the same capital spending cycle. That is a real but limited synergy, and Doosan should be pressed on which specific capabilities it thinks move between the two.
It also concentrates the risk. A holding company with a construction equipment cycle, a power plant cycle and a semiconductor cycle at least had diversification. One with two large semiconductor materials businesses is making a call.
Take the sceptical view properly. AI server demand has been the single most crowded investment theme of the past three years, and every materials supplier touching it has reported record quarters. Fifty-three percent growth off a low base, in a category where the incumbent Japanese suppliers are capacity constrained, may say more about temporary shortage than about Doosan's position.
The tell will be what happens to the premium. Low-loss laminate commands its price because supply is tight. Panasonic, Mitsubishi Gas Chemical and the Taiwanese producers are all expanding. When that capacity lands, the mix benefit that took high-value products from 71% to 81% of revenue does not disappear, but the margin on those products compresses.
There is also the ordinary conglomerate objection. Doosan's electronics business could be worth a great deal as a standalone AI materials company and is instead one division inside a holding structure carrying ₩22,432.6bn of liabilities and a construction equipment subsidiary. Investors wanting exposure to AI circuit board material are buying a lot of other things alongside it.
Segment operating profit, not segment revenue. Doosan discloses divisional results in its quarterly materials and the half-year report carries the detail. What matters is whether the electronics division's operating margin is expanding along with its mix, or whether it is growing revenue at flat margin — which would mean it is winning volume on price rather than on technology.
Second, the high-value mix percentage in the second half. It went from roughly 71% to 81% in a year. Another ten points would be remarkable. A flat reading would suggest the premium grades have reached their ceiling in Doosan's current capacity.
Third, whether the company says anything concrete about the industrial logic connecting laminate and wafers once the SK Siltron acquisition closes. A specific answer — shared customers, shared qualification teams, shared capital planning — would make the strategy credible. Silence would suggest the two purchases are related only in a slide.
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.