In this AI cycle, the market has learned how to price memory makers — we covered that in Korea’s Memory Industry: One Country, Two Cycles. How to price the suppliers behind the chipmakers is a different question, and the answer comes in layers. This piece looks at the least visible layer of the supply chain: test consumables.
Leeno Industrial (리노공업, 058470.KQ on KOSDAQ, hereafter Leeno) makes semiconductor test probes (the LEENO pin) and IC test sockets, and it is the global hidden champion of this layer: roughly 70% global share in test probes, more than 1,000 customers, and operating margins between 44.8% and 47.5% for three consecutive years from FY2023 to FY2025 on DART filings. FY2025 revenue reached 372.5 billion won, up about 34% year over year; operating profit reached 177.0 billion won, up about 42%. For this company, rising AI chip test intensity is not a narrative. It is a number already booked in the accounts.
The market has paid for it: from the start of 2026 through June 12, the stock rose 61.0%, against 8.9% for the KOSDAQ over the same period. On DART-basis FY2025 EPS, the trailing PE is about 52x, above the roughly 30x historical band cited in brokerage research.
The tension arrived in late April. Founder and chairman 이채윤 (Lee Chae-yoon) announced a block deal to sell about 7 million shares, roughly 860 billion won, taking his stake from 34.66% to 25.48%. The stock closed down 11.7% on the first trading day after the announcement, and as of June 12 it remained about 16% below the pre-announcement close and about 18% below its March high. The market did not refuse to charge a governance discount — it charged one, on the order of 15%, once, and has not widened it since. Through the decline, domestic institutions kept selling while foreign investors kept buying; but flows are zero-sum, and the information is in the price: what foreign buyers are buying is “the discount has been taken,” not “there is no discount.” This piece examines what that pricing rests on, and under what conditions it breaks.
On Friday, April 24, 2026, Leeno filed an executive securities trading plan report with DART: the largest shareholder, founder and chairman 이채윤, planned to sell about 7 million shares via block deal — roughly 9.18% of shares outstanding, worth about 860 billion won at the prevailing price. After completion, his stake would fall from 34.66% (about 26.41 million shares) to 25.48%, still the largest single holding.
On the next trading day, Monday, April 27, the stock fell from 124,400 won to 109,800 won, down 11.7% on the day, with an intraday low of 105,900 won. This was a stock that had started the year in the 60,000-won range and had touched its 52-week high of 128,000 won on March 3. After the price had roughly doubled, the founder announced a sale — some investors read it as a top signal. The other reading: 이채윤 was not resigning, 25.48% remained a controlling stake, and nothing about the business had changed.
What followed looked like a stress test. The stock had been correcting since late March, hit an interim low of 85,100 won on June 8 — a drawdown of more than 33% from the March high — then rebounded 16.3% in a single session on June 9 — the same day Busan Ilbo reported the deal might be delayed or reassessed amid the price decline. The 11.7% drop on announcement day and the 16.3% rebound say the same thing: the price is sensitive to governance news in both directions. The stock closed at 104,500 won on June 12, for a market value of about 7.96 trillion won. Year to date it is up 61.0%, an excess return of 52.1 percentage points over the KOSDAQ, with a gap of about 18% to the high still unclosed.
Put these numbers together and the question becomes clear: the market is willing to pay a premium for Leeno’s quality, but what the founder’s exit tests is not quality — it is governance. Both variables now sit inside the same valuation — and the governance slice has already been clipped by roughly 15%: even after the strong Q1 print, the stock has not reclaimed the pre-announcement close of 124,400 won.
Start with the business itself. Leeno’s revenue comes in roughly three parts: IC test sockets at about 57.3%, test probes (the LEENO pin) at about 33.6%, and medical device components (precision parts for ultrasound diagnostic equipment and similar) at about 8% to 11%. Exports account for about 77% of revenue (2024, DART annual report basis), and publicly named customers include Samsung Electronics and TSMC.
Probes and sockets are not equipment. They are consumables plus custom products. A test probe must be replaced after hundreds of thousands to millions of contact cycles, and every new chip design needs its own newly developed test socket — custom per SKU. Those two attributes combine into a revenue structure that equipment makers do not have: the more chips produced and the more new designs taped out, the more repeat purchasing follows. An equipment maker’s revenue is tied to customers’ expansion decisions — orders are one-off, and results swing with the capex cycle. A consumables maker’s revenue is tied to customers’ utilization rates and new-product cadence. Same supply chain, two entirely different businesses.
The margin can be decomposed into three layers. The first is dominant share: roughly 70% of the global test probe market (as reported by Asia Economic Daily in 2024), with fine-pitch probes depending on in-house CNC and tooling technology, so that precision plus delivery time form a compound moat. The second is customer dispersion: more than 1,000 customers means no single buyer has the bargaining power to push prices down — the mirror image of equipment makers that depend on two or three large customers. The third is the cost structure: FY2025 selling, general and administrative expenses were 17.4 billion won, less than 5% of revenue; gross margin of about 52% falls almost directly through to operating profit.
The cost elasticity shows most clearly in a downturn. From FY2022 to FY2023, revenue contracted from 322.4 billion won to 255.6 billion won, down about 21%, while the operating margin rose from 42.4% to 44.8%. Across the four years FY2022 through FY2025 on DART filings, the operating margin stayed above 42% in every year. A business that can lose a fifth of its revenue and hold its margin, and a business whose margin goes to zero in an order gap, normally deserve two different valuation frameworks — that is what “quality premium” means here.
A word on the size of the pond. Third-party industry research puts the global IC socket market at roughly $0.94 billion to $1.14 billion in 2025 (estimates vary meaningfully by source), growing at about 8% to 9% a year. Competitors in the socket segment include ISC of Korea, TSE and Yokowo of Japan, plus Enplas, Sensata, and TE Connectivity — competition there is tougher than in probes. This market is too small for giants to attack with heavy capital, and deep enough for a specialist consumables maker. Classic hidden-champion terrain.
The logic of AI demand for test consumables runs as follows: AI chips such as GPUs and HBM are highly integrated with high pin counts, so test times are longer than for ordinary chips and sockets are replaced more often; meanwhile AI chips iterate quickly, with many new designs, so demand for custom socket development rises in step. For now this logic is qualitative — there is no public quantitative data on the price of an AI socket or its replacement cycle, as the blind-spots section discusses — but the result is already written into the financial statements.
FY2025 (DART separate financial statement basis): revenue of 372.5 billion won, up about 34% year over year; operating profit of 177.0 billion won, up about 42%; an operating margin of 47.5%, the highest on a comparable basis since FY2022 — brokerage research has called this level a roughly ten-year high; net profit of 152.0 billion won.
Q1 2026 extended the slope: revenue of 78.4 billion won, up 43% year over year; operating profit of 34.9 billion won, up 50%; an operating margin of 44.5%, still above 44%. The company published a preliminary results disclosure on May 6 and confirmed it in the quarterly report on May 14.
The four-year sequence makes the picture clearer: revenue ran 322.4, 255.6, 278.2, and 372.5 billion won from FY2022 through FY2025. FY2023 was a semiconductor down-year; FY2025 was the year AI test demand converted into revenue. Sales returned to and exceeded the prior peak while the margin set a new high — the growth was not bought with price cuts.
Using the June 12 close of 104,500 won and DART-basis FY2025 EPS of 2,002 won, the trailing PE is about 52x. The reference point cited in brokerage research is an average of about 30.2x in the first half of 2024 — the current multiple stands at roughly a 70% premium to the stock’s own historical band. On consensus estimates, the forward PE is about 40x (StockAnalysis, as of 2026-06-12). Capital-return metrics support the quality label: ROE of 21.4% and ROIC of 37.8% (same source). One confounder must be stated: the entire Korean semiconductor chain has been re-rated in the AI rally over the same period — KOSDAQ-listed equipment peer HPSP trades at about 40x on 2025 expected earnings (a third-party estimate) — so part of the roughly 70% premium is sectoral, and how much belongs to “quality” itself cannot be cleanly attributed.
One accounting note is needed: FY2025 EPS of 2,002 won sits far below FY2024’s 7,463 won, which points to a likely stock split or other change in share count — dividing net profit of 152.0 billion won by the current 76,211,850 shares outstanding gives an EPS of about 1,994 won, close to the DART figure. This does not affect the 52x calculation, but it means EPS is not directly comparable across years; the anomalously high multiples shown by some data providers (for example 299x) come precisely from this basis mismatch.
More informative is the split in flows. In the decline from the late-April announcement into early June, institutions were persistent net sellers (with large selling on May 15, May 27 to 29, and June 1 to 3) while foreign investors were persistent net buyers: the foreign ownership ratio rose from 21.63% on May 14 to 23.17% on June 12 (Naver basis; the June 12 endpoint carries a data anomaly — reported foreign net buying that day exceeded the day’s total volume — so the precise level awaits verification against official KRX data). In the post-sale shareholder structure, Wasatch Advisors of the US is the de facto second-largest holder after the founder — 6.93% as of its latest disclosure (July 2025), after peaking at 9.08% in mid-2024 — the National Pension Service holds 5%, and Mirae Asset holds a smaller position. In other words, domestic institutions treated the founder’s sale as a reason to leave, and foreign long-only money treated the drawdown as an entry window. One announcement, two prices.
The governance question itself has not gone away. According to a January 2025 report by The Bell, chairman 이채윤’s children have no intention of succeeding him, and the company has no public succession plan. On March 26 the company filed a corporate value-up plan as a voluntary disclosure (joining Korea’s value-up framework), but this piece was not able to review the details. The market’s current pricing amounts to a claim: a 45%-margin consumables business does not depend on the founder personally, so a one-time discount of roughly 15% is enough to cover the absence of a succession plan. That is a falsifiable claim.
First, the valuation level itself. At 52x trailing against a 30x historical band, the premium is pricing the continuation of AI growth. If the semiconductor cycle slows or AI capex decelerates, a consumables maker’s revenue may prove more resilient than an equipment maker’s, but multiple compression from a 52x starting point does not require earnings to turn negative.
Second, the overhang is not over — “charged once” only holds if there is no second time. 이채윤’s remaining 25.48% stake is still very large; as of this writing there is no follow-up disclosure on the completion of the block deal or the composition of buyers; Busan Ilbo reported on June 9 that the sale may be delayed or reassessed amid the share-price decline. The pace of absorption is unknown. With no succession in place, nothing rules out further selling.
Third, customer concentration is invisible. The company does not disclose the share of its top customers; exposure to AI chip customers beyond Samsung and TSMC cannot be verified, and the risk of large customers bringing socket development in-house cannot be examined either.
Fourth, the 70% share figure covers probes only. Probes are about a third of revenue; the bulk of revenue — sockets at about 57% — faces merchant competition from ISC of Korea, TSE of Japan, Enplas, and others, where competition is tougher than in probes. The monopoly applies to a third of the revenue and does not extrapolate to the rest.
Fifth, the AI thesis lacks its quantitative link. “AI chips require higher test intensity” is currently qualitative reasoning plus circumstantial evidence from the income statement, with no public data on rising ASPs or shortening replacement cycles. If part of the FY2025 growth came from one-off demand during capacity build-outs, extrapolating the slope will mislead.
Sixth, the beta is asymmetric. This is a stock up 61% year to date — and the same stock that fell 11.7% in one session on April 27 and traded more than 33% below its high on June 8. Quality does not exempt a high-beta KOSDAQ name from downside volatility.
The public data behind this piece has nine gaps; five matter most for judgment. First, a precise FY2022-FY2023 historical PE band is missing, so the “52x versus 30x” premium measurement rests on a single reference point from brokerage research. Second, customer concentration is entirely invisible; AI customer exposure can only be inferred backward from reported growth. Third, comparable PEs for peers (ISC, HPSP, Hanmi Semiconductor, and others) are missing, so relative valuation is not possible. Fourth, there is no quantitative data on AI socket pricing or replacement cycles — the weakest link in the growth-durability argument. Fifth, the basis break between FY2025 and FY2024 EPS (a likely split) has not been confirmed from filing notes; in addition, Naver data for June 12 shows single-day foreign net buying of about 6.33 million shares, larger than the day’s total volume of about 4.53 million shares — a clear anomaly, so the precise scale of foreign buying needs to be re-verified against official KRX data.
The final completion status and buyer information for the block deal, and the specific contents of the corporate value-up plan, also could not be reviewed in the original filings as of this writing.
This piece therefore has the same standing as our earlier industry notes: it is a structural observation on how the market prices the consumables layer of the supply chain, not a conclusion about a single stock. Its core judgment is this — Leeno’s financial statements have already proven the quality of the consumables business; the market has also charged its governance discount, on the order of 15%, taken once and not widened since — what foreign money bought was never “there is no discount,” but “the discount has been taken.” The premises behind that pricing (the fundamental slope continuing, no further selling added) need to be re-verified every quarter.
This report is an independent KSINQ market observation for informational purposes only. It is not investment advice. Price and ownership data snapshot: June 12, 2026; written June 13, 2026.