Hanmi Semiconductor Co., Ltd. (KRX:042700) earned an operating margin of 51.9% in the second quarter of 2026, on revenue of ₩251.1bn and operating income of ₩130.3bn. Both figures are records, up 39.5% and 51.0% respectively on the same quarter last year.
Two quarters earlier, in the fourth quarter of 2025, the company recorded revenue of ₩83,014,164,855 and operating income of ₩27,627,813,980 — a 33.3% margin on less than a third of the revenue.
The gap between those two quarters is the whole investment case, in both directions.
Hanmi makes thermocompression bonders. High bandwidth memory is built by stacking DRAM dies vertically and bonding them together under heat and pressure; the TC bonder is the machine that does the stacking and pressing. Get it wrong and the stack warps or the connections fail, which at eight or twelve layers destroys the part.
That is a narrow product with a narrow customer set — there are three memory manufacturers in the world that matter — and it is why Hanmi's revenue swings the way it does. Equipment orders arrive when a memory maker decides to add capacity or move to a new generation, and they stop when the capacity is installed.
The fourth quarter of 2025 was one of those pauses. The recovery since has a specific cause: memory makers began mass production of HBM4 during 2026, and a new generation requires new tools. Hanmi signed a ₩44.2bn contract with SK hynix for HBM4 TC bonders, and it supplies Micron, whose capital spending expansion should keep orders coming.
Set the number in context. Semiconductor capital equipment is one of the most profitable industries in manufacturing, and the best companies in it — ASML, Lam Research, Applied Materials — run operating margins in the 25% to 35% range. Hanmi earned 51.9% in a quarter and 43.6% across FY2025.
That is not a company earning a good return on a difficult product. That is a company with pricing power that comes from being the only qualified supplier of a tool the customer urgently needs.
Gross margin tells the same story: 57.6% in FY2025, on machines Hanmi builds itself. The cost of goods is 42% of the selling price.
Margins like that are wonderful and they are also a signal. In capital equipment, a 50% operating margin is an advertisement, and it has already had an effect. SK hynix — for years Hanmi's defining customer — qualified a rival supplier after a well-publicised disagreement, and Hanmi's dependence shifted toward Micron and others. The company remains the market leader in TC bonders. It is no longer the only choice.
The competitive threat that matters is not another TC bonder maker. It is the possibility that TC bonding stops being the right way to build HBM.
The successor technology is hybrid bonding, which joins dies directly at the copper level without the microbumps TC bonding relies on. It allows thinner stacks and more layers, which is exactly what HBM generations beyond HBM4 need. It also requires different equipment, made by different companies — the wafer-bonding specialists rather than the thermocompression ones.
If hybrid bonding becomes the standard for the generation after next, Hanmi's installed base becomes legacy equipment and its order book resets. If TC bonding persists, or if the two coexist for years while hybrid bonding yields improve, Hanmi keeps selling.
Nobody outside the memory makers' process engineering teams knows which. That uncertainty, more than anything in the accounts, explains why this stock traded between ₩81,500 and ₩409,500 over the past year.
At the ₩217,500 close on 25 August 2026, the market capitalisation was ₩20,730,403,500,000 on 95,312,200 shares.
Against FY2025 net income of ₩214.0bn, that is about 97 times earnings. Against total equity of ₩690.3bn at the end of FY2025, it is 30 times book.
Annualise the second quarter instead and the picture softens considerably — ₩130.3bn of quarterly operating income run-rated gives roughly ₩520bn a year, and the multiple drops to something in the forties on earnings. Which is the right basis depends entirely on whether the second quarter was a level or a peak, and the fourth quarter of 2025 is the reason that is not obvious.
The balance sheet, at least, carries no risk. Total liabilities were ₩123.0bn against ₩690.3bn of equity, with ₩276.2bn of cash. This company could not get into financial trouble if it tried.
The bull argument deserves stating properly, because it is not naive.
TC bonding equipment is qualified into a customer's process over years. A memory maker running HBM production on Hanmi tools does not switch suppliers casually, because requalification costs time in a market where being late to a generation costs billions. Incumbency in semiconductor equipment is genuinely sticky, and Hanmi has it at more than one customer now — the loss of exclusivity at SK hynix was offset by winning Micron, which is arguably a better position than depending on one buyer.
The HBM4 ramp is also real and large. Every AI accelerator shipped needs HBM beside it, and each generation requires more stacking steps than the last. Volume growth in HBM converts almost mechanically into demand for bonders.
And the company is not standing still. It filed a corporate value enhancement plan in July, disclosed acquisitions of property and of a stake in another company during the summer, and has held an unusual number of investor events — fifteen investor relations announcements filed in three months.
Whether the third quarter holds near ₩250bn of revenue. The fourth quarter of 2025 showed how fast this company's revenue can halve. Two consecutive quarters above ₩240bn would establish a level rather than a spike.
Second, any disclosure about hybrid bonding — whether Hanmi is developing its own, partnering, or betting that TC bonding persists. That is the single most consequential question about the business and it will not be answered by an earnings figure.
Third, customer concentration in the FY2026 annual report. Korean filers disclose customers exceeding 10% of revenue. Whether Hanmi's dependence has genuinely broadened beyond one buyer, or simply moved from one to another, determines how much of that 51.9% margin it gets to keep.
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.