ISU Petasys Co., Ltd. (KRX: 007660) had the best quarter in its history in Q2 2026. Revenue of ₩379.9 billion, up 57.4% year over year. Operating profit of ₩77.1 billion, up 83.3%, for a 20.3% operating margin. Net income of ₩59.7 billion. This is a company that lost ₩91.1 billion in FY2019 and was still posting a net loss in FY2021, so the transformation is not subtle.
The interesting part is what's underneath. The products everyone buys this stock for, the very high layer count boards for AI accelerator systems, were 11% of revenue in the quarter. Up from 7% in Q1, which is fast growth, but 11%. Most of what ISU Petasys sells is still ordinary high-layer-count networking board. My argument is that the AI demand is real and mostly hasn't arrived yet, that it hasn't arrived because the plant physically can't make it, and that at 41.2x trailing earnings the market has already paid for the capacity that shows up in 2027.
The clearest evidence sits in the parent company numbers. Average monthly revenue at the parent went from ₩99.1 billion in Q1 2026 to ₩102.7 billion in Q2. That's 3.6%. Consolidated revenue over the same two quarters went from ₩340.3 billion to ₩379.9 billion, or 11.6%.
So the main plant was roughly flat quarter to quarter while the group grew double digits. That's not a demand signal. A factory that adds 3.6% of output while its highest-value product mix jumps from 7% to 11% of sales is a factory running full, substituting better work in place of worse work because it can't do both.
Very high layer count boards are hard for a specific and unglamorous reason: drilling and lamination. Getting thirty or forty layers to register within tolerance across a large panel is a yield problem, and adding capacity means adding drills and press capacity and then qualifying them with customers who don't qualify quickly. The company has been doing exactly that, building a new facility of roughly 34,900 square meters in Daegu with initial mass production from January 2026 and staged expansion after.
Meritz Securities expects phase-two expansion to lift parent monthly capacity to around ₩120 billion from Q3, and a further plant-six investment to take it to roughly ₩150 billion a month by the second quarter of 2027. If that lands, parent output goes from about ₩1.2 trillion annualized today to roughly ₩1.8 trillion, a 50% increase over four quarters. That, not the order book, is the growth rate that matters.
Gross margin in Q2 was 25.1%, against 23.2% a year earlier. That's good for a board maker and it's been steadily climbing, from 5.1% in FY2019 to 25.9% in FY2025, which is what the mix shift into high-layer work looks like.
But 25.1% gross doesn't automatically make 20.3% operating. What does is SG&A of ₩18.3 billion, which is 4.8% of revenue. Most manufacturers give back ten to fifteen points between the gross and operating lines. ISU Petasys gives back five. There's no big sales organization, no meaningful brand spend, and the customer list is short enough to serve with a small team.
That's a genuine structural advantage and it's also a warning. When almost all of your cost sits above the gross line, in a plant, your operating margin is a direct function of utilization. There's no discretionary spending to cut when volume falls.
This is the part I'd want any buyer at this price to sit with. Over the trailing four quarters ISU Petasys earned ₩259.3 billion of operating profit and generated about ₩69.5 billion of operating cash flow. That's 27% conversion. Q1 2026 operating cash flow was negative ₩76.5 billion, in a quarter with ₩67.2 billion of operating profit.
Working capital ate it. Inventories reached ₩374.4 billion at the end of Q2, up 70% from ₩219.9 billion a year earlier, against revenue growth of 57%. Trade receivables went to ₩343.3 billion from ₩223.2 billion, up 54%. Together that's ₩717.7 billion tied up, close to two quarters of revenue.
Some of this is what growth costs. Copper-clad laminate and specialty prepreg for high-layer boards have long lead times, and building inventory ahead of a capacity ramp is the right call. Receivables scaling with revenue is arithmetic. Still, capex over the same four quarters was about ₩181.7 billion, so free cash flow ran roughly negative ₩112 billion.
The expansion isn't being funded from profits. It's being funded from the balance sheet, and the balance sheet was topped up by an equity raise. Total equity jumped from ₩357.1 billion at Q1 2025 to ₩665.8 billion at Q2 2025, with share capital rising from ₩63.2 billion to ₩73.4 billion, so existing holders took roughly 16% dilution to pay for the plant. That was a reasonable trade. It's also a reminder that the money for plant six has to come from somewhere.
At ₩112,800 on August 28, ISU Petasys is worth ₩8.28 trillion on 73.4 million shares. Trailing net income is about ₩201.0 billion, so 41.2x. Book equity is ₩859.8 billion, so 9.63x. Trailing revenue is ₩1.31 trillion, which puts the shares above 6x sales for a printed circuit board manufacturer.
Run the bull case forward and see where it gets you. Take the Meritz capacity path at face value: parent at ₩150 billion a month by Q2 2027, or ₩1.8 trillion annualized. Add subsidiaries at roughly their current share and you land somewhere near ₩2.2 to ₩2.3 trillion of consolidated revenue. Hold the 20% operating margin, which requires the new capacity to run as profitably as the old, and tax it at the rate the company actually pays. You get something in the region of ₩350 billion of net income, and the current market cap sits at about 24x that.
So on a full build-out, fully utilized, margin-holds basis, roughly two years out, the stock is at 24x. That's the optimistic case, and it isn't cheap. Every one of those assumptions has to work. Miss the capacity date, or fill it at a lower margin because the incremental customer negotiates harder, and the number moves quickly in the wrong direction.
The shares have already lived through that math both ways. The 52-week range runs ₩59,200 to ₩161,400, and the current price is about 30% below the high and 91% above the low.
There's a serious argument on the other side. The customer list reportedly includes Alphabet, Nvidia and Microsoft, and qualification at those accounts takes years, which is a moat that doesn't show up on the balance sheet. High layer count board supply is genuinely tight, the mix shift to Multi-Lam is accelerating rather than stalling, and pricing has been moving in the company's favor alongside the mix. A supplier that is capacity-constrained into structurally short demand is in the best position a component maker can occupy. Paying 41x for the year before the capacity arrives may look obvious later.
The chairman's stake becoming a billionaire-scale holding is a fair summary of how much this business changed, and I don't want to write off a company whose gross margin went from 5% to 26% in six years.
What tempers it is FY2023. Revenue that year grew 5%, from ₩642.9 billion to ₩675.3 billion, and operating margin fell from 18.1% to 9.2%. Operating profit halved on flat revenue. That's the same company, the same plant, the same broad customer base, in a year when networking orders paused. Nothing about the AI cycle repeals that arithmetic. When a fixed-cost fab loses utilization, margin doesn't erode, it drops.
And ISU Petasys will be carrying a much larger fixed cost base into the next pause than it was in 2023. Property, plant and equipment is ₩361.9 billion, up 80% from the end of FY2024.
The number that settles this is the Multi-Lam revenue share, and it should appear in the Q3 2026 disclosure. It went 7% to 11% between Q1 and Q2. If it clears the mid-teens in Q3 as phase-two capacity comes online, then the AI portion of this business is compounding fast enough to grow into the multiple, and the mix is doing the work rather than the cycle. If it stalls near 11%, the extra capacity is being filled with ordinary networking board and the AI story is smaller than the valuation says.
Two supporting checks in the same report. Watch parent monthly revenue against ₩102.7 billion, since ₩120 billion is what the phase-two ramp is supposed to deliver and it's the cleanest test of whether the expansion is on schedule. And watch operating cash flow, which has run about ₩69.5 billion over four quarters against ₩259.3 billion of operating profit. A quarter where cash conversion normalizes would remove the main thing that makes this look like a company outgrowing its own funding.
The semiannual report for the period ended June 2026 was filed with DART on August 14.
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.