If you want to know whether the AI semiconductor boom has actually spread into general memory capacity, or whether it is still confined to a narrow band of high-bandwidth memory and advanced packaging, Wonik IPS Co., Ltd. (KOSDAQ:240810) is the cleanest instrument on the Korean market. And right now it is reading cold.
Second-quarter revenue was ₩216.5bn, down 10.6% from a year earlier and 21% below the ₩275.0bn the company booked in the fourth quarter of 2025. Trailing twelve-month revenue is ₩924.8bn. In 2021 this company did ₩1,232.3bn. In 2020 it did ₩1,090.9bn. Six years and one AI buildout later, it is running roughly a quarter below its own peak.
Korean semiconductor equipment splits into two very different exposures, and they have been diverging for two years.
One group sells into the packaging step where HBM stacks get assembled. Thermal compression bonders are the obvious example. That equipment is bought specifically to make high-bandwidth memory, and demand for it has been extraordinary because HBM output has grown faster than almost anything in the industry's history. A company selling only into that step can post record numbers while the rest of memory capex sits still.
Wonik IPS is in the other group. It builds deposition and etch tools, the equipment that processes wafers before any of them get stacked. Chemical vapor deposition, atomic layer deposition, dry etch. Those tools get bought when a fab adds wafer starts or converts a line to a new node. They do not get bought because the packaging house downstream is busy.
That distinction is the whole point. Converting existing DRAM capacity to HBM-grade DDR5 or building out packaging lines does relatively little for a deposition supplier. Adding new wafer capacity does a great deal. So Wonik IPS reads out the second thing, and the second thing is what people mean when they say the cycle has broadened.
I should be careful about how much interpretation I am loading onto one revenue line. The company does not publish a breakdown that proves memory makers are converting rather than expanding. What the filings do show is that a broad-line deposition and etch supplier to Korean fabs is running below where it ran in 2020, which is a fact, and which is hard to square with the claim that a general capacity wave is already underway.
Wonik IPS has been through this before and the shape of the last recovery is instructive.
Fiscal 2019 was the trough of the previous memory downturn: revenue ₩669.2bn, operating income ₩41.1bn. Then the cycle turned. Fiscal 2020 revenue came in at ₩1,090.9bn, up 63% in a single year, with operating income of ₩140.6bn. Fiscal 2021 added more, ₩1,232.3bn and ₩164.1bn. When Korean memory capex genuinely broadens, this company's revenue does not creep. It jumps by more than half in twelve months.
Now run the current recovery. Fiscal 2023 was the trough: revenue ₩690.3bn and an operating loss of ₩18.1bn. Fiscal 2024 came in at ₩748.2bn, up 8.4%. Fiscal 2025 at ₩909.8bn, up 21.6%. Annualize the first half of 2026 and you get roughly ₩763bn, which is below 2025.
Three years off the bottom, the recovery has produced a slower climb than the first twelve months of the last one, and the most recent half year went backwards. That is the observation. It does not prove the cycle will not broaden. It does say that as of the June quarter, it had not.
The operating line makes the same point more sharply. Operating income for the first half of 2026 was ₩29.1bn, against ₩29.1bn in the first half of 2025. Identical, to within a rounding error, while everything around this company got louder.
At ₩111,100 the stock is close to three times its 52-week low of ₩38,150. The 52-week high is ₩183,300. Market cap is ₩5.45tn (~$3.9bn at roughly 1,400 won to the dollar, an approximate rate), which against ₩73.9bn of trailing twelve-month operating income is roughly 74x, and against ₩1,073.6bn of book equity is about 5.1x.
So the market is not confused about the trailing numbers. It has looked past them entirely and is paying for a 2027 in which memory makers add wafer capacity in volume. Local brokerage commentary carried in the Korean press points the same way, citing a rising order backlog and equipment revenue recognition ramping from the third quarter, including shipments to the US and to Chinese customers. Those are estimates and forward statements rather than reported results, and I am treating them as such.
The gap between a stock near triple its low and revenue below its 2020 level is the entire investment question here. If the broadening arrives, today's multiple resolves quickly, because this business has ₩300bn or so of fixed operating cost that turns incremental revenue into operating profit at a steep rate. If it arrives late, the multiple has to hold up for several more quarters against numbers that do not support it.
The strongest counterargument is a timing one, and it has evidence behind it. Equipment revenue recognizes on shipment and acceptance, and those are lumpy in a way that makes any single quarter a poor read on demand. The June balance sheet is genuinely suggestive: inventories jumped to ₩458.1bn from ₩272.7bn at the end of December, and current liabilities more than doubled to ₩414.8bn from ₩188.1bn. Tools appear to have been built and not yet recognized. If that is what happened, my cold reading of the second quarter is measuring the wrong thing, and the third quarter will say so loudly.
A second objection is that I am treating 2020-21 as the benchmark when that was an unusually synchronized capex wave across DRAM, NAND and display. Comparing any cycle to that one sets a high bar. Fair. But the point of a litmus test is the direction of travel, and a company whose revenue fell 10.6% year over year in the June quarter is not signaling broadening no matter what benchmark you use.
The third risk runs the other way and deserves stating. China. Wonik IPS has growing exposure to Chinese memory customers, and Chinese capacity additions are exactly the kind of greenfield wafer capex that lifts a deposition supplier. That demand is also the most politically exposed revenue in the company's mix. Export controls have moved several times in the last three years, and a change there hits the recovery case directly.
One number, in November. If third-quarter revenue comes in materially above ₩275.0bn, meaning above the best quarter of the last two years rather than merely above the weak second quarter, the inventory on the June balance sheet was pre-built shipments and the cycle is broadening on schedule. If it lands in the ₩220-260bn range, the tools shipped but the volume is a normal quarter, and the broadening is still ahead rather than underway.
Watch the fourth quarter guidance around the same date, and watch whether the order backlog commentary that brokers are citing shows up in the company's own disclosure rather than only in research notes.
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.