Tokai Carbon Korea: inside the memory boom's least leveraged supplier

In the middle of the most profitable memory boom in years, a small South Korean maker of graphite crucibles and furnace rings will spend an afternoon trying to attract global fund managers. Tokai Carbon Korea, listed on the country's tech-heavy KOSDAQ index, is booked to brief investors at HSBC's Korea Corporate Day, the sort of event usually reserved for companies that have outgrown a thin, domestic shareholder base. The likely pitch is the obvious one: the AI memory supercycle, the strongest tailwind a semiconductor supplier has seen in a decade, runs straight through this firm's order book. The obvious pitch is mostly right, and mostly beside the point. The wind reaches this company late, and through a narrow crack.

What Tokai Carbon Korea sells is not memory, and not the machines that make it. It sells the consumable parts that line the inside of chipmaking furnaces — silicon-carbide-coated susceptors, heater rings, crucibles, dummy wafers and shower heads — which wear down with every batch of wafers and must be replaced. Demand therefore tracks "wafer starts" and the number of process steps a fab performs, rather than the spot price of DRAM. When Samsung and SK Hynix run more wafers and stack more layers onto high-bandwidth memory, each wafer consumes more deposition cycles and more components. The firm is a toll collector on the bit, not on the boom.

That structure explains a telling puzzle in the accounts. In 2025 revenue rose 9%, to about 301bn won, while net profit fell roughly 3%, to about 70bn won. The decline was booked in the very quarters when the memory boom was gathering force; SK Hynix would later print a single-quarter operating margin of 72%. Even deep into the supercycle, Tokai Carbon Korea's quarterly profit has bumped along near 24bn won, and the first real signs of acceleration appear only in forward estimates, which put next-quarter revenue up about 10%. The pricing power that has memory makers locking in multi-year contracts barely reaches this link in the chain.

A toll road with few lanes

Why so little flows through, when the traffic is so heavy, comes down to who holds the pricing power and who merely holds capacity. The moat is real: coating furnace parts in silicon carbide demands chemical-vapour-deposition know-how, and a part that sheds impurities can wreck an expensive wafer, so chipmakers are slow and reluctant to qualify a new supplier. That gives incumbents high switching costs and their margins a sturdy floor. Yet the same structure caps the ceiling. The customer list is essentially two Korean memory giants, a concentration that hands the buyer the leverage; and capable domestic rivals, notably Hana Materials in silicon carbide, press on price and capacity from the other side. In a boom, an incumbent with spare CVD lines should benefit; one without them simply loses share to the factory that added lines.

This is not a weak business. It carries no debt and a large net cash pile, and it runs roughly a third of revenue down to EBITDA — the financial posture of a quality toll road. The question is who enjoys the proceeds. The controlling shareholder is the Japanese parent, Tokai Carbon, a Tokyo-listed carbon specialist that holds just under half of the Korean unit and has been consolidating its grip by buying out a long-time Korean co-owner. The parent's motive is strategic integration and a supply of components for its global customers, not a rising cash return to outsiders. Consistent with that, the dividend is a token: a yield of about 0.6% and a payout of under a quarter of profit. For a minority shareholder, value arrives through the share price rerating, not through income.

Priced as if it has already arrived

Which brings the valuation into focus. The company is worth about 1.9trn won against last year's 70bn won of profit — roughly 27 times trailing earnings, and still a mid-teens multiple only if the boom's delayed benefits land in full. Sell-side analysts carry it as a Strong Buy, with a target price implying about a quarter of upside; in other words, the consensus has already adopted the supercycle thesis that the company will now present to HSBC's guests. The risk runs the other way: a volume-driven, two-customer business being bought at price-cycle valuations, with the earnings rise on which the multiple depends still an estimate rather than a result.

The corporate-day charm offensive is, in that light, a reveal about the share register as much as about the company. A stock so dependent on two customers, one parent and a cyclical lever arrives at the top of the cycle with its holder base thin, retail and Korean, and its parent content to let outsiders earn little. The investment case is real but unusually narrow. The memory boom will almost certainly fill this firm's order book eventually; whether it fills the gap between a flat profit and an optimistic multiple is a second-order bet on wafer volumes and process intensity, layered on top of the first-order bet on memory itself. Buyers of the obvious pitch should be sure they have bought the right thing.