Full Report

Industry — Semiconductor Test Handlers & Back-End Equipment

Mirae sells test handlers — robotic machines that load, thermally condition, and sort packaged chips while a separate "tester" decides which die pass and which get scrapped. The handler does not test the chip; it presents the chip to the tester at the right temperature, in the right position, hundreds of times per minute. Handler revenue is a small (~US$2–4B) but strategically critical slice of the US$130B+ global semiconductor equipment market, sitting downstream of the front-end fabs and just before final assembly and shipment. Two facts shape everything below: handlers are back-end equipment, so they track the memory and packaging cycle — not the front-end logic capex headlines — and the customer base is small, concentrated, and capital-driven, so each multi-million-dollar order can move a quarter.

1. Industry in One Page

Front-end wafer fabrication equipment (WFE) — ASML's lithography, Applied Materials' deposition, Lam's etch — is dominated by five firms with ~80% share and 25–35% operating margins. Mirae plays in a different room: back-end test and assembly, where the chip has already been made and now needs to be tested, sorted, and packaged before shipment. The back-end market is more fragmented, smaller, more cyclical, and more exposed to memory and to Chinese capex than the front-end.

ASML's order book and Mirae's order book respond to different stimuli on different timelines: ASML follows the leading-edge logic race (TSMC, Samsung Foundry, Intel), while Mirae follows memory bit growth, OSAT capacity adds, and — increasingly — Chinese domestic memory build-out. The same "AI chip super-cycle" headline can mean a strong year for both, but for different reasons and at different lead times.

Sources: SEMI Worldwide Semiconductor Equipment Market Statistics (WWSEMS, May 2025 forecast); Valuates Reports QYRE-Auto-31N12786; dataintelo Semiconductor Test Handler Market 2025; Yahoo Finance industry market caps; Mirae FY2025 DART Business Report. All sizes in source currency (USD).

2. How This Industry Makes Money

Test handlers are sold as custom-configured capital equipment at six-figure unit prices, against multi-month engineering lead times, with the cash flow profile of a bespoke industrial project. Gross margins are decent (35–45% for the better players) because each handler is engineering-intensive, but operating margins are thin because R&D, sales engineering, customer-site support, and a small installed base produce high fixed costs. The pricing unit is the handler set — Mirae's FY2025 ATE division produced 90 units (capacity 96/year) at an average production value of approximately ₩403 million per unit (~US$266k), with selling prices higher once aftermarket parts and service are layered on.

Cash terms are Mirae's stated handler terms from its FY2025 DART Business Report (II.4): 30% on contract / 60% on delivery / 10% on acceptance.

Where the bargaining power sits. Memory IDMs (Samsung, SK Hynix, Micron) and large OSATs (ASE, Amkor, JCET) buy in bulk, qualify a small set of approved vendors, and squeeze price aggressively because their own end markets are commodity-priced. Handler vendors enjoy lock-in only at the device-type / package level — once a customer has tooled a line and trained operators, switching cost is real, but the next greenfield line is genuinely contestable. Result: handler vendors are price-takers on standard memory packages and price-makers only where the package is exotic (HBM stacks, system-in-package, automotive-qualified parts).

Illustrative ranges drawn from FY2024-FY2025 reported results for Applied Materials, Lam, KLA, ASML, Advantest, Teradyne, Besi, ASMPT, Hanmi, Cohu, and Techwing. Test handler industry average is depressed by Korean and Chinese small-cap players running closer to break-even outside the upswing.

The investor takeaway. Profits compound at the front-end and at choke points (EUV, HBM-test, advanced packaging). Test handlers sit closer to the commodity end of the equipment chain. The handler segment is cyclically leveraged: when memory capex turns and OSATs add capacity, even sub-scale handler players can earn 15–25% operating margins for a few years (Mirae's FY2024 reported ATE operating margin was 51% on tooling-deposit accounting, before settling to FY2025's still-strong 19.5%). The same leverage works in reverse — Mirae's FY2023 ATE segment posted a ₩15.7 billion operating loss on ₩18.4 billion of segment revenue.

3. Demand, Supply, and the Cycle

Test handler demand is the second derivative of the chip market: it moves when chipmakers decide to add or retire capacity, not when end-device demand changes. The cycle is violent, and the leading indicators are memory bit shipments, OSAT utilization, foundry capex announcements, and government subsidy disbursements (China's Big Fund, US CHIPS Act, Korea's K-CHIPS).

Where the cycle hits first. When memory prices fall, IDMs cut equipment orders before they cut chip output. The signal sequence: spot DRAM/NAND price → memory-maker quarterly capex guidance → equipment-maker book-to-bill → handler-maker bookings → handler-maker revenue (4–9 months later). On the way down, the order book empties fastest because new lines are deferred while existing lines keep running. On the way up, the order book rebuilds with a lag because customers wait for price recovery to confirm before committing. The 2022H2–2023H1 downturn was textbook: Mirae's FY2023 revenue fell to ₩21.7B with a segment operating loss, FY2024 rebounded to ₩27.0B, and FY2025 surged to ₩50.8B (+87.8%) as China memory orders flooded in.

A sub-scale handler vendor running near full plant utilization (Mirae was at 97.5% machine utilization in FY2025) amplifies cyclical swings into nearly +88% revenue jumps and double-digit margin moves.

4. Competitive Structure

The global test handler market is moderately concentrated but not consolidated: third-party research lists the top five — Cohu (Xcerra brand), Changchuan Technology (China), Advantest, Hon Precision, and Techwing — at a combined ~57% share, with Cohu the leader at ~21%. The remaining ~43% splits among 25+ vendors, mostly Asian small- and mid-caps. Mirae sits in the long tail.

Top-5 shares from Valuates Reports "Global Semiconductor IC Test Handler Market Research Report 2025". Combined top-5 = 57%. Mirae's share is sub-1% (FY2025 revenue ~US$33M against a global market of US$2.3-3.8B).

Three competitive layers matter.

Geographic concentration. China is the single largest buyer of test handlers — about 47% of global handler demand by consumption, per Valuates — while Japan, Korea, and Taiwan host most of the production. Korea accounts for ~14% of handler demand and is dominated by SK Hynix, Samsung, and their captive equipment arms (notably SEMES, a Samsung subsidiary that does not compete in the open market). The Korean back-end ecosystem includes Hanmi Semiconductor (dominant HBM thermal-compression bonder, ₩31.5T market cap), Techwing (handlers + cube probers, ₩2.2T), ISC (test sockets, ₩5.1T), and Mirae at ₩123B — three to four orders of magnitude apart in scale.

5. Regulation, Technology, and Rules of the Game

The semiconductor equipment industry is shaped less by traditional regulators than by export controls. The US Bureau of Industry and Security (BIS) sets which equipment can be sold to which Chinese entities; Japan and the Netherlands have aligned with most US measures since 2023; Korea has selectively cooperated. Front-end advanced-node equipment is heavily restricted; back-end test and packaging equipment is mostly not on the Entity List directly, but end-customers are — most importantly YMTC (added Dec 2022). This bifurcates the market.

6. The Metrics Professionals Watch

Generic financial ratios are distorted by the cycle. The metrics below are what analysts and customers actually track.

The North Star is the SEMI book-to-bill ratio. It has called every back-end equipment cycle since the 1990s with a 1–2 quarter lead. Per SEMI's mid-2025 forecast, back-end test equipment sales are projected to grow 48.1% to US$11.2B in 2025, then 12.0% in 2026 and 7.1% in 2027 — the strongest sustained back-end up-cycle since 2017–2018. Whether Mirae captures more or less than that industry growth rate is the central operating question for the next two years.

Source: SEMI WWSEMS — May 2025 press release. 2023 baseline normalized to zero for visual clarity; actual 2023 was a down year for back-end equipment.

7. Where Mirae Corporation Fits

Mirae is a niche Korean handler specialist with outsized China exposure, operating in the long tail of a moderately concentrated global market. FY2025 revenue of ₩50.8 billion against a global handler market of US$2.3-3.8B implies under 1% global share, but its position in the Korean handler ecosystem is more meaningful: Mirae is named in its own FY2025 Business Report alongside Advantest, JT Corp, Techwing, and AMT as one of five test handler suppliers active in Korea.

Market caps as of 2026-05-21 from peer_valuations.json; non-KRW peers converted at 2026-05-21 FX. Mirae is roughly two orders of magnitude smaller than Techwing and three orders smaller than Hanmi Semi.

8. What to Watch First

---

Know the Business

Mirae is a sub-scale Korean test-handler vendor whose FY2025 financials are the by-product of one Chinese customer's capex decision. The economic engine is custom-built capital equipment paid for in stage deposits; the moat is thin; the cycle is violent; and the headline P/E disguises a business that ran negative operating cash flow last year because the same revenue surge that produced reported profit also blew up working capital. After the FY2025-low-to-now rally the market is now pricing the equity at ~1.0× book — the question is not whether that is cheap, but whether the equity is durable.

1. How This Business Actually Works

Mirae sells two products, but one matters: ATE test handlers were 89% of FY2025 revenue and SMT chip mounters / linear motors were 11%. Each handler is a six-figure custom-built machine that loads packaged chips, thermally conditions them, presents them to a separately-purchased "tester" and sorts the survivors. The plant in Cheonan ran at 97.5% utilization in FY2025, producing 90 ATE units and 13 SMT units — the entire business is, physically, about a hundred handlers a year.

Per-unit figures are segment revenue (new equipment plus spares and service) divided by units shipped — they overstate the change in a pure ASP because year-over-year spares/service mix moves with the installed base. The directional read still holds: FY2025 added volume at a markedly lower per-unit revenue.

FY2025 ATE units jumped from 23 → 90 (4×) but segment revenue per unit fell from ₩978M to ₩502M. This is what a single mega-order from one customer looks like — Mirae traded ASP for volume to land YILINING PRECISION. The capacity is small enough that one buyer can absorb most of it; the pricing power is small enough that one buyer can dictate the deal.

The 30/60/10 cash structure is why Mirae's FY2025 net income (₩10.1B) bears no resemblance to its operating cash flow (negative ₩5.2B). When orders triple in a year, deposits cover only part of the build, inventory and receivables swell, and the cash impact lags revenue by six months or more. Mirae is a working-capital-heavy industrial: incremental profit shows up on the income statement long before it shows up in the bank.

Where margin actually comes from. Two levers: utilization and customer mix. The fixed-cost base is roughly ₩12B/year of operating expenses (SG&A + R&D), barely changed across the FY2021–FY2025 swings. Gross margin moved from 34.7% (FY2021) to -33.9% (FY2023 trough) to 41.8% (FY2025). FY2024's reported 83% gross margin is an accounting anomaly — likely tooling deposits and recovery items recognised as revenue without matched COGS, per the segment disclosure where ATE alone reported ₩11.5B operating income on ₩22.5B revenue. The clean read: at ~₩50B revenue and 90+ handlers a year, this business earns 18-20% operating margins. Below ₩30B revenue, it loses money.

2. The Playing Field

Mirae sits at the long-tail bottom of the test-handler universe — two to three orders of magnitude smaller than the global benchmarks, and roughly a third the revenue of Techwing, the next-up Korean handler peer. The peer set splits into three economic tiers: a global leader (Advantest), a Korean back-end ecosystem champion (Hanmi Semi), and a cluster of sub-scale cyclicals (Cohu, Techwing, ISC, Mirae) where margins live or die on the cycle.

Latest FY for each peer; FX at period-end. Mirae market cap and multiples computed at the 2026-05-21 close of ₩27,600 (4.47M shares = ₩123.4B / ~US$82M); P/B uses FY2025-end equity of ₩122.6B; P/E uses FY2025 net income of ₩10.1B; EV/EBITDA uses FY2025 EBITDA of ₩10.2B and FY2025-end net cash of ₩4.4B. Cohu and Techwing margins reflect the down-leg of the back-end cycle; Mirae, Hanmi, and Advantest the up-leg.

Scale and product mix decide margins. Advantest (global tester duopolist) and Hanmi (Korean HBM bonder near-monopoly) sit at the top — both make products with real switching costs. Cohu, despite owning the largest revenue share of the handler sub-segment, is loss-making at $453M revenue because handlers alone do not generate enough recurring service revenue to absorb fixed engineering cost in a normal year. Techwing, the closest Korean handler peer, runs sub-scale at $110M and a 10% margin. Mirae posted 18% only because FY2025 was an exceptional volume year courtesy of one customer.

What "good" looks like in this industry is product adjacency to HBM or being the tester, not handler scale. Hanmi Semi proves the point: ₩600B revenue, 44% op margins, ₩31T market cap. Mirae's standard test handlers are not on that road, and the valuation reflects it: ~1.0× P/B at current price versus Hanmi at 30×+.

3. Is This Business Cyclical?

Mirae is deeply cyclical — beyond what "cyclical" means for a typical industrial. The last four years swung revenue from ₩56B → ₩22B → ₩27B → ₩51B (-61%, +24%, +88%) and operating margin from +15% → -84% → +39% → +18%. There is no normalized year in this dataset because the company never spent two years in the same state.

Three places the cycle hits, in order of investor relevance:

The cycle pattern matters because it puts the FY2024 print (83% gross margin, 39% operating margin) in context: an accounting catch-up year as tooling deposits and prior-period recognitions normalized, sandwiched between a wipeout (FY2023) and a volume surge (FY2025). The next downturn — when Chinese memory capex slows or YILINING reduces orders — would look like FY2023 again: gross profit goes negative, operating losses approach ₩15-20B, and equity is consumed at roughly ₩30B/year of run-rate burn until new capital arrives.

4. The Metrics That Actually Matter

Standard ratios (P/E ~12 at current price, ROE 8.9%, current ratio 3.04) flatter this business. Below are the five readings that actually tell you whether the next twelve months are heading up or down.

The metric that does the most work is the OCF/NI gap. A handler vendor in a true up-cycle should convert ~70–80% of net income to operating cash over a two-year window. Mirae converted roughly −29% over FY2024-FY2025 combined (₩10.1B + ₩6.6B = ₩16.7B NI; -₩5.2B + ₩0.3B = -₩4.9B OCF). The reported earnings did not produce cash, and the missing cash had to be raised through equity (financing CF +₩9.4B in FY2025) and ₩6.5B of new long-term debt — the first long-term debt this company has carried in the data window.

5. What Is This Business Worth?

The right lens is through-cycle EV/revenue floored by book value, not a current P/E. The two segments (ATE 89%, SMT 11%) are economically similar (custom industrial equipment, same plant, same customers); there is no listed subsidiary or hidden stake to unpick. The single segmentation question is "what is normalized revenue?" — and the answer depends almost entirely on whether the one Chinese customer stays.

The honest framing. Two reasonable values bracket the equity:

• Floor: 0.4–0.6× book = ₩49–73B. What the market pays when it does not believe through-cycle earnings cover the cost of capital — this is where Mirae traded at year-end 2025 (₩61.7B / 0.50× book).
• Up-cycle fair value: 8–10× normalized EBITDA of ₩5B = ₩40–50B EV → ₩45–55B equity after net cash. The current market cap of ₩123B sits well above this band, paying a premium for the FY2025 P&L to extend.

At the current ₩27,600 spot, the market has moved through book value and is now pricing in cycle continuation. The investment case turns on whether the YILINING revenue stream is durable enough to lift normalized EBITDA above ₩5B sustainably. If it is, ₩100B+ equity is defensible. If not, ₩50–75B is the right number. No projection of segment economics or DCF will tighten that range; only customer evidence will.

6. What I'd Tell a Young Analyst

Read this company as a levered bet on one Chinese customer's capex plus a Korean handler vendor's equity floor. Everything else is noise.

Watch four things. First, every quarterly DART disclosure for the major-customer note — when YILINING's share moves below 40%, the business has diversified; when it stays above 55%, the concentration is structural. Second, the single-supply contract notices (단일판매공급계약) — the only forward-looking signal Mirae publishes; track customer name, currency, and amount over rolling four quarters. Third, operating cash flow versus net income — if the gap widens further in FY2026, the up-cycle was a working-capital illusion and the next downturn will be financed by equity. Fourth, share count drift — the company has been a serial diluter; a flat share count for two consecutive years would itself be a thesis-changing event.

What the market may be missing: the FY2025 surge looks like a top-line breakthrough but came at a 50% per-unit revenue cut and burned cash. What it may be over-rewarding: at ~1.0× book the margin of safety that was real at year-end 2025 (P/B 0.50) has compressed, and FY2023-style downturns can vaporise ₩30-40B of equity in eighteen months. The thesis-killer is not weak guidance — it is one Korean exchange filing showing the top customer changed.

Do not over-think this name. It is a small, levered, customer-concentrated cyclical with a 100-handler-a-year ceiling, owned by a controlling shareholder that changed in 2023, with a 16:1 reverse split last year as evidence of past balance-sheet stress. Value it at book, demand a discount for capital-structure history, and own it only if you have an independent reason to believe the China customer relationship is good for two more years.

---

Long-Term Thesis

Mirae compounds only if three independent variables break the right way over the next five-to-ten years — the Korean non-US, non-Chinese back-end supplier position survives the next BIS export-control cycle, customer concentration broadens beyond YILINING PRECISION into either an HBM-adjacent SK Hynix line or a second China memory account, and the new ownership runs a full down-cycle without another rights offering or convertible bond. None of those three is on a credible track today. The historical record — five-year cumulative net income of ₩0.7B against cumulative free cash flow of −₩49.8B, share count up roughly 169% before a 16:1 reverse split, a 49% ASP cut to land the FY2025 volume surge, a 72/100 forensic risk score, and no disclosed HBM handler product — argues the equity is a cycle-leveraged trade dressed in a 42-year heritage brand, not a long-duration compounder. The 1.0× book floor is genuine and the order cadence into Q1 FY2026 is real, but the durable case rests on customer diversification and cash conversion that have not yet printed. Read this page as the underwriting check on whether Mirae is a multi-year asset or a multi-quarter rental.

The 5-to-10-Year Underwriting Map

Every variable that has to clear for a 5-10 year compounding case to be real. Each row is testable in DART filings, KRX disclosures, or industry data — none requires earnings-call colour or management commentary.

The driver that matters most is customer diversification. Every other variable in the table assumes a baseline of revenue stability that only a second large customer can deliver. The Korean regulatory window is binary and outside management's control. Cash conversion is downstream of customer mix because YILINING's payment cadence dominates the receivables build. Reinvestment economics are downstream of revenue durability. The HBM product gap is real but addressable on a 3-5 year horizon. If Mirae cannot widen the customer book — by either landing a second China memory account or growing SK Hynix into a ₩15B+ line — the long-term case has no spine.

Compounding Path

A long-duration case requires plausible math on how reported revenue turns into per-share owner value. The honest answer: the math only works in the upper-bound scenario, and the upper bound requires the customer-diversification milestone above.

The headline chart is not a compounding chart. Revenue moved ₩48B → ₩56B → ₩22B → ₩27B → ₩51B in five years — amplitude, not growth. Net income summed to ₩0.7B. Free cash flow summed to -₩49.8B. ROIC averaged near zero. Share count tripled before a 16:1 reverse split reset the optics. Whatever happens in the next five years, it has to break that pattern decisively, not just print one more strong year.

Even in the bull case Mirae is a 1.5-2.0× book outcome on a five-to-ten-year view, not a 5-10× multi-bagger. The difference between the base case (₩120-160B equity) and the bull case (₩220-290B) is mostly customer diversification, not heroic operating leverage. The bear case (₩40-60B) is what FY2023 looked like in a single year — and the same setup is structurally repeatable because the customer concentration has not changed. A reader who buys the long-duration thesis is paying for a 50-100% upside scenario with a 50-60% downside scenario; the asymmetry depends entirely on which variable in the underwriting map prints first.

Durability and Moat Tests

A 5-10 year thesis only survives stress the business has not faced yet. The four tests below are observable, the signals are specific, and at least one is competitive and one is financial.

Three of the five tests have a refutation signal that has already shown signs of arriving — Changchuan's share has grown from sub-5% to ~12% over five years, the FY2025 OCF print was already negative against record NI, and the HBM product gap was unchanged in the FY2025 AR. The two tests still genuinely open are regulatory (binary, outside management's control) and governance (where the January 2026 Nextern injection is the one piece of validating evidence). A long-term reader should weight the regulatory test highest — it is the single variable that can take revenue back to FY2023 levels in one quarter, and no operating improvement can offset that shock.

Management and Capital Allocation Over a Cycle

The post-2023 Nextern-installed management cohort has been in place for under three years, which is too short a sample to underwrite capital-allocation discipline through a full down-cycle. The track record on what they directly control is mixed-to-good: contracts shipped, capital raises closed, side ventures cleanly disclosed and walked back. The track record on what defines a long-term steward — preserving per-share value through a cycle without dilution — is structurally negative because the cohort's only completed cycle so far was the FY2024 capacity build funded by rights offering and CB conversions.

The capital structure history is the most important single fact for the 5-10 year case. Pre-split shares went from 1.66M (FY2021) to 58.2M (FY2024), then 16:1 reverse-split to 4.48M in July 2025. The Nextern stake of 32.31% was assembled largely through CB conversions and bonus issues rather than market purchase. The 2026-01-04 Nextern injection of ₩6.5B into Mirae is the first clear signal that the controlling chain is willing to fund the operating subsidiary rather than extract from it — but the chain itself (Nextern Roll Korea -₩13.0B, SSC -₩10.6B FY2025 net losses) has structural cash needs that have not gone away. A 5-10 year underwriter must price the probability that the next downturn is again funded by dilution, because the only completed precedent in the data window is that one.

Three patterns from the History tab are load-bearing. First, management is unusually candid about admitting side-venture failures in print — DeFi, AI healthcare, mining, EV parts all added and then deleted with the same boilerplate "no synergy" footnote. This is the strongest signal in favor of process honesty. Second, the same management has never voluntarily restored per-share economics after dilution — the reverse split reset the optics, not the economic value. Third, no buyback, no dividend, and no meaningful direct insider ownership exists; alignment runs through the Nextern/SSC chain, not Mirae shares.

Failure Modes

Five thesis-killers, each specific enough to be observable in DART or KRX disclosure within a defined window.

Two of the six failure modes are severe and binary (YILINING substitution, BIS expansion); three are high and observable on a 1-3 year cadence; one is medium and follows the industry cycle. A 5-10 year underwriter should weight the binary risks highest because they are uncorrelated with management quality — no operating improvement and no governance reform offsets a single Entity List notice or one Chinese customer's decision to switch suppliers.

What To Watch Over Years, Not Just Quarters

Five observable milestones that would update the 5-to-10-year thesis. Each is a disclosure that the company or a regulator must publish; none requires management commentary.

The long-term thesis changes most if the top-customer revenue share falls below 40% with a second named customer above 10% by FY2028 — that single disclosure would simultaneously validate the cash-conversion case (because diversification implies sustainable receivables collection), the regulatory durability case (because YILINING substitution risk shrinks), and the reinvestment case (because plant economics work at multiple customer endpoints). Until that happens, every other improvement in Mirae's reported financials is a single-cycle artifact of one Chinese customer's capex schedule, not a 5-to-10-year compounding story.

---

Competition — Mirae Corporation (025560)

Competitive Bottom Line

Mirae has no durable moat. It is the smallest listed name among the five test-handler vendors active in Korea (Mirae, JT, Techwing, AMT, Advantest — all named in Mirae's own FY2025 DART business report), and across the broader back-end test ecosystem it sits two to three orders of magnitude below the relevant scale benchmarks. The single competitive fact that explains FY2025's 18% operating margin, four-fold ATE shipment jump, and ₩50.8B revenue print is that one Chinese customer (YILINING PRECISION, 56.5% of revenue) chose Mirae over a Chinese domestic handler maker like Changchuan. The competitor that matters most to the next 24 months is therefore not Advantest, Cohu, or even Techwing — it is Changchuan Technology (Shanghai-listed) and the wider Chinese domestic handler industry, subsidised, sanctions-protected, and gaining on the same wave that floated Mirae in FY2025.

The Right Peer Set

The five comparators below were chosen because they share customers (memory IDMs + OSATs), share the back-end cycle, or are explicitly named as Mirae's competitors in its own FY2025 DART business report. The Korean back-end ecosystem peers (Hanmi, ISC) compete for the same customer capex wallet but with different products — they are not "handler competitors" but are the reference points investors use to judge whether a Korean back-end equipment vendor is winning.

Two named competitors are missing from the public peer set: JT Corporation (KOSDAQ 089790) is a direct test-handler peer named alongside Techwing in Mirae's AR; we held it as a documented alternative rather than swap it for any of the five comparators that have stronger Korean-back-end signal value. AMT is private/unlisted and has no comparable financials. Changchuan Technology (300604.SZ) is the Chinese champion that matters most for Mirae's China revenue but did not pass through the staged-data pipeline; it is flagged in the Threat Map and in the unresolved-question queue.

Market caps and EV from data/competition/peer_valuations.json (USD values converted at 2026-05-13 FX). Mirae market cap = 4.47M shares × ₩27,600 = ₩123.4B (~US$82.8M). Hanmi, ISC, Techwing, Mirae values in millions of KRW; Advantest in millions of JPY; Cohu in millions of USD. Source filings: Mirae FY2025 DART 사업보고서, Cohu FY2025 10-K, Advantest IAR2025, Hanmi/Techwing/ISC FY2025 DART 사업보고서.

The map shows Mirae bottom-left: the smallest revenue, a respectable cycle-peak margin, and the smallest market cap by an order of magnitude. The two profit-pool winners (Hanmi, Advantest) are top-right. Cohu is the cautionary case — biggest pure-play handler revenue but loss-making at the bottom of the cycle, proving that handler scale alone does not bring durable profits.

Where The Company Wins

Mirae's wins are narrow, real, and almost entirely commercial, not technological. There is no proprietary node, no patent moat, no captive customer base. What Mirae has are four operational advantages that matter in specific deals.

Where Competitors Are Better

Every comparator in the peer set has at least one structural attribute that Mirae cannot match without years of investment.

Cohu R&D ₩137B figure converts US$92.2M at ~1,490 KRW/USD. Advantest R&D is estimated from segmental reporting and IAR2025 disclosure (FY2026 figure converted from JPY). Mirae and Korean peer R&D figures from each company's FY2025 DART 사업보고서 연구개발활동 section.

The intensity chart explains why Mirae's product roadmap is shallow: Hanmi spends 8× Mirae's R&D in absolute KRW, and they spend it on the part of the market (HBM) that is winning. Cohu and Techwing each have R&D budgets 5-7× Mirae's despite being smaller-cap than Hanmi or Advantest, which means even the down-cycle players continue to invest at multiples of what Mirae can fund.

Threat Map

The threats below are ranked by severity over the next 24 months. The top two are the only ones that can move Mirae's equity value materially.

Moat Watchpoints

Five measurable signals tell the investor whether Mirae's already-thin competitive position is improving or deteriorating. None of them require management commentary — every one is observable in DART filings, industry data, or competitor disclosures.

---

Current Setup & Catalysts

The stock is trading around ₩27,600 (21 May 2026), up roughly 94% in a single month and 154% over the trailing year, and the market is mostly watching whether the China-driven order book disclosed since November 2025 actually converts to cash in the Q2 and Q3 FY2026 quarterly filings. The recent setup is bullish on tape and bookings, bearish on the underlying thesis variables — six months of strong DART supply contracts and a ₩6.5B controlling-shareholder injection have already been priced, while the FY2025 cash conversion gap, the FY2024 PP&E reclassification, and the 56.5% YILINING customer concentration remain unresolved. The near-term calendar is unusually rich for a microcap: a 5-for-1 forward stock split on 2026-07-27 is hard-dated 66 days away, a Q2 FY2026 quarterly business report follows in mid-August, and the prospective CXMT IPO sits inside the window as the single biggest external lever on the YILINING/Yiling Trading order chain. This page is the bridge from the durable thesis page to the events that actually update it over the next two quarters.

What Changed in the Last 3-6 Months

The post-November 2025 stretch is the spine of the current setup. Eight events deserve the PM's attention because they either changed the order book, the capital structure, or the market's read of the controlling-shareholder behaviour.

The recent narrative arc. Six months ago the market worried about whether Mirae could survive another FY2023-class trough — the share count had nearly tripled to 4.47M (post the July 2025 16:1 reverse split), cash was at a five-year low, and there was no third-party validation of the customer story. Two things changed that: the November 2025 ₩14B contract published in DART crystallised the China order book as a discrete event chain, and the January 2026 Nextern Roll Korea ₩6.5B injection removed the acute funding question. What investors are now arguing about is no longer survival — it is whether the FY2025 print is a single-cycle artifact priced at peer multiples or the start of a structural earnings step-up. The unresolved item across the whole post-November stretch is cash conversion: every single forward variable (capex normalization, dividend optionality, lendability of the controller, ability to absorb another customer pause without dilution) flows through whether reported earnings of ₩10B can pay out as ₩7-8B of operating cash in any rolling twelve-month window.

What the Market Is Watching Now

The five debates rank in declining order of decision value. Cash conversion in Q2 sits at the top because every other variable — capital allocation, dilution risk, sustainability of the China relationship, ability to weather a customer pause — eventually flows through it. CXMT IPO is the highest-value external variable because Mirae's name on the Korean supplier list is the closest thing to third-party validation of the durable revenue case. BIS is binary and asymmetric — the highest-severity tail risk, but the least useful to track day-to-day.

Ranked Catalyst Timeline

Eight events with enough date precision and thesis linkage to belong on a PM's near-term board. Ranked by decision value, not chronology.

The ranking deliberately puts the Q2 FY2026 cash flow print ahead of CXMT because the cash test is observable on a definite date and resolves the most contested bear point in the file. CXMT is higher impact in magnitude but lower confidence — a delayed IPO is a non-event for Mirae's FY2026 revenue line, since shipments are already booked. BIS sits at #3 for severity but with a Low confidence on timing; a PM should monitor it as a tail rather than time it. Items #5-8 update the long-term thesis but are second-order to the cash and customer questions.

Impact Matrix

The catalysts that genuinely resolve underwriting questions, not just add information.

Four of the five rows in the matrix update durable thesis variables, not just the near-term print. That is unusual for a microcap — most catalyst pages are stuffed with single-quarter events. For Mirae, the events that matter on a six-month horizon are also the events that matter on a five-year horizon, because the company is small enough that any single one of these resolutions can rewrite the entire underwriting model. The Q2 cash flow turn is the closest to an "atomic" thesis test in the file: a clean print refutes three forensic flags simultaneously (B4 capitalising opex, C2 op outflows in investing, C4 working-capital lifelines), and a bad print confirms all three at once.

Next 90 Days

The window from today (2026-05-22) through late August 2026.

The 90-day calendar is dense, not thin. Two hard dates inside the window (split + Q2 print) and three rolling watchpoints (DART cadence, CXMT, BIS) is more catalyst content than most KOSDAQ microcaps carry in a quarter. The asymmetry sits in the Q2 print: a clean OCF turn could justify a starter long even at the current parabolic tape; a missed print would invalidate the bull case before any other variable has a chance to update.

What Would Change the View

Three observable signals over the next six months would change the investment debate more than any other. First, a positive Q2 FY2026 OCF print in the August quarterly filing — that single number would refute the cleanest forensic short thesis (B4 capitalising opex / C2 op outflows in investing / C4 working-capital lifelines all at once) and convert the FY2025 P&L from a single-cycle artifact into the start of a structural earnings base. Second, a CXMT IPO prospectus that names Mirae or Yiling Trading as a contracted equipment supplier — the long-term thesis on customer durability rests on a relationship the company itself has only ever confirmed through one DART supply contract; external validation removes the "one customer, one year" framing the bear case depends on. Third, any BIS Federal Register notice or Korean MOTIE statement that expands controls to back-end test equipment shipped to YMTC / CXMT / Yiling-adjacent customers — this is the binary, uncorrelated tail risk that nothing else in the file offsets, and a single ruling can reverse 56.5% of FY2025 revenue inside one quarter. The first two would force a re-rating toward the bull case (₩40-45K target band); the third would force a re-rating toward the bear case (₩12K target band). The current price of ₩27,600 sits between them because none of the three has yet printed — but all three have observable evidence windows inside the next six months, which makes the position size question a function of which signal a PM is willing to wait for.

---

Bull and Bear

Verdict: Watchlist — the order book is real, but the cash conversion, forensic, and customer-concentration questions cannot be resolved at today's price after a +169% twelve-month move. Bull and Bear agree on most underlying facts and disagree on what they mean. The decisive question is whether the FY2025 earnings recovery converts to operating cash flow in H1 FY2026 — both sides anchor their exit/entry triggers to the same disclosure, which makes this a wait-for-the-print situation. The forensic profile (FY2023 restatement, ₩30.8B PP&E movement, no audit committee, loss-making holding chain) and parabolic tape (RSI 78, 30-day vol 120%) tilt the immediate risk asymmetry against ownership; the real order book and 1.0× P/B floor keep the long thesis alive. The condition that moves this off Watchlist is one printed quarter of positive OCF with revenue ≥₩18B and YILINING share trending down — until then, paying for the recovery requires faith the cash statement has not yet extended.

Bull Case

Bull commits to ₩42,000 per share (market cap ~₩188B, ~52% upside from ₩27,600), set at 12× FY2026E net income of ₩15–16B and cross-checked against 1.4× P/B on FY2026-end equity of ~₩135B. Timeline 12–18 months, anchored on the Q2/Q3 FY2026 quarterly reports and the H2 FY2026 cash conversion turn. Disconfirming signal: YILINING/Yiling Trading revenue dropping below ₩5B in a single quarter without a ≥₩10B replacement contract surfacing in the single-supply disclosures within 90 days.

Bear Case

Bear commits to ₩12,000 per share (~₩53.6B market cap, −57% from ₩27,600), set on a blended floor of 0.4× book on ₩122.6B equity (= ₩10,960) and 8× a normalized through-cycle EBITDA of ~₩5B; both methods converge at ₩11–13K. Timeline 12–18 months. Primary trigger: a DART disclosure that YILINING's order share is dropping — through pause, BIS sweep, or Korean alignment with US back-end controls. Cover signal: two consecutive quarters of positive OCF with YILINING below 30% of revenue and no new equity or CB filing in the same window.

The Real Debate

Verdict

Watchlist. The Bear carries slightly more weight today, not on the durable thesis but on timing: the forensic profile and the parabolic tape (RSI 78, 30-day vol 120%, +110% above the 200-day SMA after +169% twelve-month gains) raise the standard of proof above what the FY2025 financials currently meet. The most important tension is cash conversion — timing or structural? — both sides have anchored their exit/entry signals to the same future disclosure (H1 FY2026 OCF), and it has not yet arrived. The Bull could still be right: the order book is real, contracts are signed and dated in DART, the geopolitical wedge favours Mirae, and 1.0× P/B is a genuine floor on a recapitalised balance sheet. The condition that would move this from Watchlist to Lean Long is one quarterly report showing OCF turning positive while revenue holds ≥₩18B, with no new equity issuance or convertible filing in the same window. The marker that would tip the other way is any YILINING share decline in the next quarterly customer disclosure without a ≥₩10B replacement contract surfacing within 90 days; if that prints, Bear's downside target becomes the operative anchor.

---

Moat — What Protects This Business, If Anything

1. Moat in One Page

Conclusion: No moat. Mirae Corporation is a price-taking, sub-scale Korean test-handler vendor whose FY2025 earnings recovery is the by-product of two transient advantages — being a non-US, non-sanctioned supplier of legacy memory handlers to one Chinese DRAM customer (YILINING PRECISION, 56.5% of revenue) and being able to deliver a six-month build window into 97.5% plant utilization. Neither survives the questions an institutional reader should ask of an alleged moat.

A moat is a durable, company-specific advantage that protects pricing power, margins, share, or customer relationships better than competitors. Mirae's behavior in FY2025 is the cleanest possible refutation: ATE units shipped jumped 4× from 23 to 90, but average revenue per unit was cut nearly in half (₩978M → ₩502M) to land the order. A vendor with a moat does not concede 50% of price to move volume; a vendor without one does. R&D at ₩3.1B is one-seventh of the closest Korean handler peer (Techwing ₩21.0B) and one-thirtieth of Cohu's. Patent stock is one-fifth of Techwing's. There is no disclosed HBM-specific handler product while every comparable peer has one. The FY2023 P&L (operating margin minus 84%, equity nearly halved, ₩31.5B net loss) shows what happens when the only "advantage" — one customer's order — pauses.

The single thing that might be a narrow, segment-specific advantage is the durable ₩5–10B/year SK Hynix relationship (~12.5% of FY2025) — once a fab line is tooled on Mirae handlers, operator training and recipe-development cost makes the next greenfield re-tool slightly painful. That is the entire moat case, and it covers roughly one-eighth of the business.

2. Sources of Advantage

The table below catalogues every plausible source of a moat in Mirae's business. Each is named, defined, and tested against the company's actual evidence. The pattern is consistent: alleged sources either reduce to industry attractiveness, depend on factors Mirae does not control, or fail the durability test.

Definitions for the beginner. Switching costs are the cost, risk, or workflow disruption a customer faces when changing suppliers — retraining operators, re-qualifying products, rebuilding software recipes, or replacing certified spare-parts inventory. Network effects exist when each additional user makes the product more valuable to other users. Cost advantage means producing the same output at structurally lower cost than competitors. Intangibles include brands, patents, regulatory licenses, and accumulated data that competitors cannot replicate. Embedded workflow means a customer's operating processes are built around the supplier's product.

Of the seven candidate sources, only two carry any measurable evidence — partial switching costs at SK Hynix and the regulatory positioning that won the YILINING contract — and both are weak. The switching-cost advantage protects a single customer worth roughly 12% of revenue. The regulatory advantage is structurally borrowed time. Everything else either does not exist (data, scale, true cost advantage) or describes the industry rather than the company (operating leverage, brand heritage).

3. Evidence the Moat Works

A moat that exists shows up in the data. The evidence below tests whether Mirae's outcomes — pricing, returns, retention, share, margin durability — look like a protected business or a price-taker. Most lines refute the moat case; one is mixed.

Eight of ten dimensions score 1–2. The only two that reach 3 ("narrow") are SK Hynix switching costs (covering 12.5% of revenue) and Korean-vs-US-vs-China regulatory positioning (which the Competition tab and Industry tab both flag as 2–4 year shelf life). The scorecard is not a moat — it is the absence of one.

4. Where the Moat Is Weak or Unproven

The previous section showed the evidence against. This section explains why — what an institutional skeptic should hold the case to. Each weakness is named, explained, and tied to the specific dynamic that makes Mirae's position fragile.

5. Moat vs Competitors

The peer table compares Mirae's moat candidates against the five most relevant comparators. The takeaway is uniform: every peer has at least one structural advantage Mirae cannot match, and the closest Korean handler comparator (Techwing) is widening the gap on the dimensions that matter most.

The comparison is not flattering. The two profit-pool winners in this peer set (Hanmi, Advantest) earn 40%-plus operating margins because they sit in product niches with genuine pricing power and IP protection. ISC (test sockets) shows the consumable adjacency that delivers a real moat. Even Cohu — currently loss-making — has a structural scale and customer-diversification position that Mirae does not. Mirae's moat ranks last not by a small margin but by an order of magnitude on R&D, by 3-4× on customer concentration, and by 2-3× on margin durability.

6. Durability Under Stress

A moat only matters if it survives stress. Mirae has limited test data because the current ownership/business model is only two-and-a-half years old, but FY2023 and the customer-concentration setup give us most of what we need.

Every stress test in the table either confirms the absence of a moat or shows the limited evidence base. The cleanest test — the FY2023 downturn — produced a P&L outcome that no moated business would have suffered: revenue down 61% in a year, operating margin minus 84%, equity halved, change-of-control. A wide-moat business in the same industry (Hanmi, Advantest) earned positive double-digit operating margins through the same year.

7. Where Mirae Corporation Fits

Mirae's competitive position is bifurcated, and the bifurcation matters more than the consolidated picture suggests. There is a small protected segment and a large unprotected one — and the protected segment is too small to anchor a moat thesis.

The chart makes the moat case obvious. Roughly ₩8B of FY2025 revenue (16%) sits behind something that looks like a narrow advantage at the SK Hynix relationship. The other ₩43B (84%) sits in segments where Mirae is competing on price, regulation, or capacity availability — none of which are durable, company-specific protections. The valuation case for Mirae cannot be a moat case; it must be a cycle-and-capital-allocation case (which is what the Numbers and Business tabs already make).

8. What to Watch

The watchlist below distills the moat thesis into measurable signals. None require management commentary or earnings calls — every line is observable in DART filings, KRX disclosures, competitor reports, or industry data.

---

Financial Shenanigans

Forensic Risk Score: 72 / 100 — High. Five years of reported net income sum to roughly ₩727M, yet free cash flow over the same window is −₩49.8B and operating cash flow has been propped up by financing of +₩43.6B. FY2024 carries the single most concerning artifact in this filing set — cost of revenue collapsing from ₩29.1B to ₩4.5B while revenue grew 24%, lifting gross margin to 83%, alongside ₩36B of capex that briefly tripled PP and E before ₩31B of that PP and E dropped off the balance sheet again in FY2025. The board already approved a restatement of FY2023 consolidated financials in early 2025, and the company is controlled by a loss-making holding entity (Nextern Roll Korea) inside a multi-layer chain rising to ROA and Co. The single fact that would most change the grade is independent confirmation of where the FY2024 PP and E went in FY2025 — a sale-leaseback, impairment recycle, or reclassification footnote would either resolve the cost-shift signal or harden it.

Shenanigans scorecard

The 13-category coverage map below uses the standard taxonomy. Six tests fail with material evidence, five flash yellow, and only two come back clean.

Breeding Ground

The control structure and the audit oversight are the weakest links in the file. This is a small-cap KOSDAQ semiconductor equipment maker with a 113-employee headcount, controlled through three layers of holding companies up to ROA and Co — a film/media-adjacent group with no operating connection to semiconductor equipment. The CEO Lee Chang-jae and the rest of the executive bench were installed on the same day Nextern Bioscience took control (2023-07-13). One outside director resigned mid-year FY2025 with 60% attendance. There is no audit committee — Korean small-cap law lets a standing auditor (Moon Hyung-suk, Hyundai Accounting Firm) stand in. Compensation is modest (no single director crossed the ₩500M individual-disclosure threshold), so monetary incentive to manipulate is low, but stock options were granted with an exercise price of ₩12,992 to the CEO and President at the bottom of the cycle (2025-03-26), now well in the money.

The breeding-ground map is decisively red. A new controller, weak independent oversight, a small-firm statutory auditor and a confirmed prior-period restatement together amplify the accounting flags below. The single soft spot in management's favor is that none of the named executives clears the disclosure threshold for individual compensation — direct cash-incentive pressure to bend numbers is modest. The mechanism worth worrying about is equity-linked: convertible bond #8 and #9 converted into 23.2M shares (pre-split) for Nextern between November 2024 and May 2025, immediately followed by the 16:1 reverse capital reduction on 2025-07-01. Stock options for the CEO and President were granted between those two events at ₩12,992 (post-split).

Earnings Quality

FY2024 reported earnings are not a faithful representation of underlying operations and FY2023 reported earnings were already restated. The cleanest forensic signal in this file is the FY2024 cost-of-revenue collapse alongside a one-time PP and E surge. COGS dropped 84% YoY while revenue grew 24%. Gross margin jumped from −33.9% to 83.3%. In the same year capex was 36.6x depreciation and PP and E nearly tripled. In FY2025 PP and E unwound by ₩30.8B while non-current assets only grew ₩3.3B — meaning ₩27.5B of the FY2024 PP and E build was reclassified into other non-current accounts or disposed without a matching cash receipt. The economic substance behind that flow has not been disclosed in the management discussion excerpt available.

The FY2024 COGS print is structurally implausible for a semiconductor test-handler maker that shipped 23 ATE units that year. Per-unit ATE revenue in FY2024 was ₩518M (=11,924/23). At a 60% gross margin assumption — which is the long-run norm for premium test handler peers — implied per-unit COGS would be roughly ₩207M, or ₩4.76B across 23 units. So the reported COGS figure of ₩4.52B is internally consistent with a 60% gross margin only if every shipped ATE handler had zero cost of materials and labor outside the equipment itself — implausible. The likelier explanation is that material, labor, and overhead were either capitalized into the ₩30.98B PP and E expansion, or held as inventory (FY2024 inventory turnover dropped to 0.36, implying DIO of roughly 1,014 days). FY2025's inventory turnover normalising to 1.13 and gross margin returning to 41.8% supports the cost-shift interpretation.

The capex-to-depreciation ratio averaged 0.5x in FY22–FY23 and then hit 36.6x in FY2024 and 16.1x in FY2025 — orders of magnitude above what an equipment maker with 97.5% factory utilisation should need to spend to support 87% revenue growth. PP and E peaked at ₩47.8B at FY24 year-end and then dropped to ₩17.0B one year later, but non-current assets only grew ₩3.3B and there is no separately disclosed gain/loss-on-sale matching the magnitude. The line moved off PP and E and into another non-current bucket, which is the textbook shape of either a sale-leaseback that flips PP and E into a right-of-use asset or operating-lease receivable, or a reclassification into "investment property" / "intangibles" / "long-term financial assets". Either is permissible under K-IFRS but materially changes whether the earlier capex represented core-business investment.

Q4 FY2023 is the single most distorted quarter in the file. Revenue fell to ₩2.62B while operating loss landed at −₩19.89B — implying the company recognised roughly ₩17B of one-shot costs against ₩2.62B of revenue (gross margin −586%). The big-bath quarter coincides with the first full quarter after the change of control. Q3 FY2024 then prints a 161% gross margin and operating income of ₩7.41B on revenue of ₩6.24B — economically impossible without one-time gains classified inside operating costs or inventory reversals. These two quarters bookend the FY2024 cost-shift pattern: bad-news pulled forward into late FY23, good-news pulled forward into FY24 once Nextern's controlling stake was fully fixed.

Cash Flow Quality

Operating cash flow does not support the reported income line. Cumulative net income over FY2021–FY2025 is ₩727M; cumulative free cash flow is −₩49.8B. The gap was closed by financing — ₩43.6B in net financing inflows from convertible bonds (CB #8 and #9, both converted), rights offerings, and bank borrowings (₩13.95B drawn at FY25 against a ₩14.10B credit line). FY2025 operating cash flow swung to negative ₩5.18B even as reported net income reached ₩10.11B — a 5-year-low cash conversion print despite a 5-year-high earnings print.

The FY2022 figure is the cleanest illustration of the cash-flow problem in the file: net income ₩8.02B, operating cash flow ₩22M. The company reported strong profit while collecting essentially nothing from operations. FY2024 repeated the pattern (NI ₩6.60B, CFO ₩329M). FY2025 broke even further — net income ₩10.11B, CFO negative ₩5.18B — meaning the year that looks most prosperous on the income statement is the year operating cash actually went into reverse.

The FY2024 financing draw of ₩18.18B was the largest single-year cash injection in five years. The bulk of it funded the ₩36.26B capex line that produced the unexplained PP and E spike. By FY2025 the company drew ₩9.38B in financing while CFO went negative and ₩16.91B more capex landed — meaning even the latest year of strong reported revenue did not produce enough cash to fund the business. Without the bond conversions and bank lines, the cash balance would have fallen from ₩38.17B at FY21 to below zero rather than the ₩10.87B reported.

The shape is decisive. Over a 5-year period the business raised ₩43.6B from financing and burned through almost all of it — cash declined ₩27.3B (from ₩38.17B to ₩10.87B). On a steady-state basis, the company has consumed approximately ₩70B of cash from operations and investing combined. Whatever the reported earnings show, the cash engine is upside-down.

Metric Hygiene

Per-share figures are dominated by capital actions, not operations. Because Mirae files under K-IFRS as a small/mid-cap, no non-GAAP / adjusted-EBITDA reconciliation is published — eliminating one type of metric-distortion risk. The replacement risk is the capital-structure churn. From the start of FY2022 to the end of FY2025, shares outstanding moved 4.45M → 30.43M → 58.23M, then collapsed to 4.48M after a 16:1 reverse capital reduction on 2025-07-01. EPS comparisons therefore look orderly only because the reverse reduction backed-out earlier dilution.

The most surprising hygiene point is the customer concentration disclosure inside the FY2025 business report: a single Chinese counterparty named YILINING PRECISION delivered ₩28.7B (56.5%) of reported revenue, with two additional China-linked counterparties (CXMT, YMTC) adding meaningful volume. CXMT is on the US Entity List as of December 2024 and YMTC has been since December 2022. The customer book that drove FY2025's reported growth from ₩27.0B → ₩50.8B is highly concentrated, sanctioned-entity adjacent, and difficult to verify externally — exactly the profile where audit and revenue-recognition scrutiny should be at its sharpest.

What to Underwrite Next

Treat the accounting risk as thesis-relevant, not a footnote. Forensic risk of 72/100 is not an automatic short call — but it is enough to require a materially wider margin of safety than the headline ₩50B revenue and ₩10B net income suggest. Five practical diligence items to track before sizing this position:

1. Footnote the FY24 PP and E unwind. Identify in the FY2025 audited footnotes exactly what happened to the ₩30.8B of PP and E that vanished. Three candidate buckets: sale-leaseback (right-of-use reclassification), reclassification into investment property/intangibles/other non-current assets, or impairment with offsetting capitalised value reversed. Any answer other than a normal disposal with cash receipt validates the cost-shift signal.
2. Reconcile the FY24 cost-of-revenue print. Reconstruct unit-level COGS for the 23 ATE units delivered in FY2024. If reconstructed COGS materially exceeds the ₩4.52B reported figure, the gap was either capitalised into inventory or PP and E.
3. Trace Yilining Precision. Confirm whether Yilining Precision / Yiling Trading is the same entity, identify its ultimate ownership and operating profile, and check for any sanctions or export-control exposure. A single Chinese counterparty supplying 56.5% of revenue is a thesis-grade risk in its own right.
4. Watch FY26 working capital signals. If Q1 FY2026 (already published — revenue ₩20.7B, net income ₩5.91B) is followed by another negative CFO print at H1, the cash-flow problem is structural, not a working-capital timing artifact.
5. Monitor any further restatement or "(정정)" disclosures. FY2023 was already restated; the disclosure index shows multiple amendments to single-customer supply contracts in 2025. A second prior-period restatement or an audit-firm change would push this file from High to Critical.

Bottom line. Mirae's reported FY2024 and FY2025 income statement is not a clean read of operating economics. The cash flow tells the truer story — the company has consumed cash for five straight years and relied on convertible bonds and bank lines to keep the cash balance above zero. The FY2023 restatement, the FY2024 cost-shift, the unexplained PP and E reversal in FY2025, the holding-company control chain and the 56.5% single-customer concentration are converging signals. This is a position-sizing limiter and a valuation haircut — not a footnote, and not yet a thesis-breaker. The upgrade path is full clean footnote disclosure on the PP and E unwind plus two consecutive quarters of positive operating cash flow. The downgrade path is a second restatement, an auditor change, or any export-control action affecting the Chinese customer book.

---

Governance: D — control beats alignment

Mirae is not a founder-led semiconductor equipment maker anymore. Since mid-2023 it has been operated by appointees of an investment vehicle (Nextern Roll Korea) that sits inside a stacked, loss-making holding chain — and the board, pay, and capital structure all reflect that. The profitable industrial business is intact; the governance scaffolding around it is the weakest part of the case.

The People Running This Company

The registered board is five people — four executive directors and exactly one outside director. The two highest-paid individuals at the company are not even on the board.

Three things to take away from this roster. First, the only person at the registered table with any operating semiconductor pedigree is President Kim Kyung-tae — a Nextern-era holdover who ran MPT before joining Mirae in 2018 and is the only insider with a direct stake (and only 0.21%). Second, the CEO and the three other inside directors all arrived together in July 2023 alongside the Nextern takeover; the CEO's prior role was VP at Nextern Biosciences, so he is the controlling shareholder's representative inside the operating company, not an independently recruited operator. Third, the two unregistered executives — Chairman On (from ROA Holdings) and VP Nam — collect the highest pay in the building yet sit outside the registered structure that any shareholder vote can touch.

What They Get Paid

FY2025 disclosed compensation totals ₩1,132M (~$781K) across all directors, the auditor, and the two unregistered executives. Every single individual stayed below the ₩500M individual-disclosure threshold Korea requires, so no specific person's pay is reported.

The shape of pay tells the story better than the level. Two unregistered executives — appointed by the board with no shareholder vote — earned roughly ₩251M each, nearly 1.8× the average registered executive director (₩138M). Sitting at the top of that pile is Chairman On Seong-jun, a former director of ROA Holdings and ROA & Co (the entities at the top of Mirae's ownership chain). The structure pays the controlling group's representative more than the registered CEO, and keeps both names off the individual-disclosure table by staying just below ₩500M. That is a deliberate design choice, not a coincidence of pay levels.

Stock-option grants on 2025-03-26 covered the CEO (18,750 post-split shares), President (6,250), unregistered VP (6,250), and 78 employees (70,813 collectively) at a strike of ₩12,992 — barely below the current ₩13,780 close. Total option fair value was just ₩200M (~$138K), so even on a fully realized basis, options would not bridge management's near-zero direct equity stake.

Are They Aligned?

This is where governance breaks. The aligned party is not the operating CEO or the registered board — it is the holding chain.

The 32.31% controlling stake sits in Nextern Roll Korea, which booked a ₩13.0B net loss in FY2025 on its own balance sheet, and is itself controlled by Studio Santa Claus Entertainment (a separately listed media company with a ₩10.6B FY2025 net loss), which is in turn controlled by ROA & Co. Two of the three layers above Mirae need cash. Mirae's profitable equipment business is the only positive earnings node in the disclosed control stack, which creates a structural incentive for capital extraction — through dividends, related-party services, or further securities issuance — that did not exist when Mirae was independently held.

That structural incentive has already shown up in the share register. Since the Nextern takeover in mid-2023, Mirae's count of outstanding shares (split-adjusted) has more than doubled, driven by a rights offering, two convertible bonds (Series #8 and #9), and then masked by a 16:1 reverse split in July 2025.

Read top-to-bottom, that table is the alignment problem in one page. Nextern walked from a ~10.9% standing stake to 32.31% control over 22 months without paying market for most of it — the bulk came through bonus issues and convertible-bond conversions while existing minority holders were diluted from ~89% down to ~68%. Then a 16:1 reverse split compressed the optics in July 2025. The minority base is still substantial — 23,908 small shareholders own 65.09% between them — but they have one outside director and no audit committee to defend them.

Disclosed insider direct stakes are vanishingly small: President Kim Kyung-tae owns 9,375 shares (~0.21%) worth roughly ₩259M (~$170K) at current prices, and CEO Lee Chang-jae has no direct holdings disclosed in the Business Report. Whatever economic alignment management has runs through their positions inside the Nextern/SSC chain rather than through Mirae stock — and those positions are not disclosed at the operating-company level.

Board Quality

The board met 23 times in FY2025 and passed 100% of resolutions — typically a tell that the body is rubber-stamping management, not challenging it. Half the agenda items (12 of 23) were "exercise voting rights at subsidiary" decisions, with a further block of real-estate-backed loans and capital-structure resolutions (16:1 reverse split, stock-option grants, rights offering paperwork). No committee meetings took place outside the board itself.

The scorecard surfaces two structural gaps. Independence is formal, not real — four of five directors arrived together with the Nextern takeover, two of the four executive directors are non-standing "Management Advisors" rather than line operators, and the sole outside director has the lowest attendance rate of any director on the board. Operating expertise lives in one person — President Kim Kyung-tae, the only board member with a deep semiconductor-equipment operating history; remove him and the registered board has zero industry depth. The financial-skill column scores well only at the standing auditor (Moon, CPA), and even there the only formal external audit oversight is one person without a committee structure around him.

The pattern is hard to argue with: the controlling-group's representatives (Lee Chang-jae, Lee Jin-gyu) show up 100%; the lone semiconductor operator (Kim Kyung-tae) is at 83%; and the two outside directors — the only people meant to push back — are the two least present, with the resignation of the second mid-year leaving the board structurally captured. No outside-director training was provided in FY2025 ("not yet deemed necessary"), and no audit-committee training either.

The Verdict

Grade: D. This is a competent industrial business sitting inside a governance structure built to serve the controlling vehicle, not the minority shareholders who own two-thirds of the float.

The strongest positives are narrow and operational: President Kim Kyung-tae is a credible semiconductor-equipment operator with seven years of continuity, attendance at registered board meetings is generally strong, all FY2025 resolutions cleared without legal incident, and the standing auditor has accounting credentials. The board did discipline itself enough to disclose the full ownership chain, the convertible-bond conversion history, and the option-grant terms in unambiguous Korean disclosure — there is no obvious concealment.

The real concerns are structural, not behavioral. Mirae's profitable cash-generating business is owned 32.31% by a vehicle that lost ₩13B in FY2025, which is in turn owned by an entity that lost ₩10.6B; the only operating-cash-positive node in the disclosed chain is Mirae itself. Share count has roughly tripled in four years driven by rights offerings and convertible-bond conversions, then been visually compressed by a 16:1 reverse split. The board has one outside director who attends 74% of meetings and no audit committee. The two highest-paid executives sit outside the registered board entirely. None of the registered directors holds a meaningful direct stake — alignment runs through the holding chain, not through Mirae shares.

The one thing that would upgrade this: A second working outside director with semiconductor-industry expertise (not another lawyer), a real audit committee, and either a meaningful direct equity grant to the CEO with multi-year vesting or a disclosed buyback that absorbs some of the dilution. A clear cap on related-party advance / lending to the Nextern–SSC parents would help even more. Absent any of those moves, the grade should not improve.

The one thing that would downgrade this further to F: A loan, guarantee, or asset transfer from Mirae upstream to Nextern Roll Korea, Studio Santa Claus Entertainment, or ROA & Co — i.e., the operating cash actually exits to support the parent's losses. The structural incentive is there; whether the board enforces against it is the next thing to watch.

---

History

Mirae Corporation pioneered Korean semiconductor test handlers in 1983 and quietly compounded as a small specialist for forty years — then in mid-2023 collapsed (revenue down 61%, a ₩31.5B net loss), was sold to a new controlling group (ROA-linked Nextern Roll Korea), recapitalised through dilution and a 16:1 reverse split, and re-emerged in FY2025 as a ₩50.8B-revenue China supplier with 56.5% of sales going to a single Chinese DRAM customer. The story management tells is "back to focus on our core ATE business after side ventures didn't pan out." The story the filings actually show is a near-death cycle, an ownership transfer, an aggressive capital structure rebuild, and a customer concentration so extreme it is now the dominant fact about the business. Credibility has improved versus the pre-2023 governance regime — the new team is unusually candid in print about side-venture failures — but the present-day equity story is a single-customer China bet wearing a 42-year-old Korean test-handler brand.

1 — The Narrative Arc

The current chapter of this business began in July 2023, when Nexturn Biosciences took control from prior owner Kwang Lim and Lee Chang-jae was installed as Representative Director on the same day. Everything in the post-2023 ownership chain, capital structure, customer base, and reported strategy resets from that moment. The legacy 1983–2022 chapter is a separate company in most respects that matter for valuation.

2 — What Management Emphasised — and Then Stopped Emphasising

Mirae does not host earnings calls or issue forward guidance the way US issuers do. The cleanest evidence of what management is willing to bet on lives in the articles of incorporation — specifically, in the list of officially declared business purposes that the board added and shareholders ratified. Each addition was sold as future-growth optionality; the subsequent deletion (or the explicit footnote that "we judged this has no synergy with our core business") is the recantation.

The pattern is unusually clear: every two years Mirae's board adds a basket of trendy adjacent businesses — masks during COVID, DeFi during the 2022 crypto top, AI-healthcare and EV-battery materials during the 2023 EV/AI cycle — and then quietly deletes the ones that "do not have synergy" once the cycle rolls over. None of these additions has ever produced disclosed revenue. The fact that the issuer itself prints the recantation in the next year's articles section is the most credibility-supportive aspect of the whole disclosure regime — it would be easy to leave the dead purposes on the books indefinitely. Each cycle, however, is an opportunity-cost statement: while management was adding mask manufacturing and DeFi, competitor Hanmi Semiconductor was building the high-bandwidth-memory test handler business that drove its multi-billion-dollar re-rating.

3 — Risk Evolution

The risk language in the annual reports themselves is shorter than US-style 10-K risk-factors sections and changes less. What changes is which risks are now empirically dominant in the underlying financials and contracts, regardless of how they are described. The heatmap below scores each risk on a 0–3 scale based on disclosure intensity and the financial weight of the underlying exposure.

The single most important shift: the company has traded operational risk (memory cycle, customer capex) for political and counterparty risk (China concentration, export control, single-customer receivables). Both classes of risk are real, but they are managed differently and price differently. A reader who was comfortable with Mirae in 2021 because "Korean ATE specialist with SK Hynix exposure" should not be comfortable in 2025 for the same reason — it is now closer to "Chinese DRAM-equipment subcontractor with a Korean listing".

4 — How They Handled Bad News

Mirae's disclosure regime is light on press releases and earnings commentary, so "handling bad news" mostly means how items are presented in the formal Business Report. There are three episodes worth comparing.

"신사업의 수익성, 기술성, 경제성 등 주 사업과의 시너지효과가 없다고 판단되어 [추진계획은 없습니다 / 삭제하였습니다]."

"We judged the new business has no profitability, technology or economic synergy with our core business, so [we have no plans to pursue it / we have removed it]." — recurring footnote across the 2021, 2022, and 2023 article-of-incorporation additions.

That sentence is the most informative thing the company says about itself. It tells you the additions were either pure governance theatre or genuinely tried and failed; it tells you the next basket of additions (whatever it is) should be discounted in the same way; and it tells you the board is willing to admit it in print. None of those facts is good news on its own, but the candour is genuine.

5 — Guidance Track Record

Mirae does not issue numeric forward guidance. What it does issue are (a) business-purpose additions that imply intended entry into a market, (b) capital-raise commitments (CBs, rights offerings), and (c) contract-disclosure announcements that show order book conversion. Treating each of those as an implicit promise allows a track record to be built.

Credibility: 5 / 10. Current management (Lee Chang-jae and the Nextern-Roll Korea-installed board, in place since July 2023) has a perfect record on the things they actually do — capital raises that close, contracts that ship, an operational turnaround that arrived on schedule. They are also unusually willing to admit, in print, that their side-venture additions did not work. Against that, every diversification ambition the board has ever asked shareholders to approve has been quietly walked back; the headline financial recovery rests on one Chinese customer; and the dilution-then-reverse-split path destroyed per-share economics that the reverse split does not actually rebuild. A 5 reflects high process honesty paired with a strategy whose durability rests on something — Chinese DRAM capex — that management cannot itself control or even publicly defend.

6 — What the Story Is Now

The current Mirae story is simpler than at any point since 2021, and more concentrated than at any point in its 42-year history. It is a Korean-listed Chinese-DRAM-equipment supplier with a single dominant customer, a recovered but fragile operating profile, and a clean — if recently rebuilt — capital structure.

The cleanest summary: management has done well at what they directly control, and the business is now exposed to things they do not control. That is a different equity story than the one Mirae told for the first 40 years of its existence, and the reader who is underwriting the FY2025 P&L as a run-rate should be sure they are underwriting the China-DRAM-capex cycle, not the Korean-test-handler franchise. They are no longer the same thing.

---

Financials — What the Numbers Say

Mirae Corporation is a micro-cap Korean semiconductor test-handler maker (113 employees, ATE plus SMT plus linear motors) whose financials swing violently with one customer cycle. FY2025 revenue more than doubled year-on-year to ₩50.8 billion, operating margin recovered to 18.0% and the company posted ₩10.1 billion of net income — yet operating cash flow was negative ₩5.2 billion and free cash flow was negative ₩22.1 billion. The balance sheet still has more cash than debt, but the cash pile has shrunk from ₩38 billion (FY2021) to ₩10.9 billion, and the share count has nearly tripled since FY2022 to fund a heavy capex cycle. The stock trades at 6.1× earnings and 0.5× book, which looks cheap until you see that 56.5% of FY2025 revenue came from a single Chinese customer (YILINING PRECISION) and that reported earnings keep diverging from cash.

1. Financials in One Page

Definitions for beginners. Revenue is sales. Operating margin is operating profit divided by revenue — a measure of profitability before interest and tax. Free cash flow (FCF) is operating cash flow minus capital expenditure — the cash a company actually has left after running the business and reinvesting. Net cash is cash minus debt; positive means the company could pay off all its borrowings and still have money left. ROIC (return on invested capital) measures the cash return earned on the capital invested in the business. P/E and P/Book are the share price divided by earnings per share and book value per share respectively — lower usually means cheaper.

The single financial metric that matters most right now is operating cash flow conversion. Reported earnings say recovery; the cash statement says the business is consuming cash. Whether Q1 FY2026's ₩4.9 billion of operating income translates into actual cash over the rest of FY2026 will decide whether this stock is a re-rating story or an accounting mirage.

2. Revenue, Margins, and Earnings Power

Mirae's income statement is a textbook semiconductor-cycle whiplash chart. FY2021 and FY2022 were two consecutive strong years driven by memory test handler shipments. FY2023 was a disaster — revenue more than halved and a Q4 inventory writedown drove gross profit negative for the full year. FY2024 was a partial rebound. FY2025 returned to roughly FY2021 revenue scale.

The FY2024 gross margin of 83.3% is not a sign of pricing power — it is an accounting mirror image of FY2023's writedown. In Q4 FY2023, Mirae took ₩15.3 billion of negative gross profit on just ₩2.6 billion of revenue, almost entirely an inventory and asset impairment tied to the customer cycle collapse. Much of that was reversed when shipments resumed in FY2024, which mechanically inflated reported gross margin. The 41.8% gross margin in FY2025 is closer to the underlying through-cycle reality.

Recent quarterly trajectory

The quarterly chart shows why the stock has run hard since mid-2025: Q3 FY2025 was the standout, with ₩22.6 billion of revenue (more than all of H1 FY2025 combined) and a 26.6% operating margin. Q1 FY2026 came in at ₩20.7 billion — confirming that the ramp is sustaining.

Q3 FY2024's apparent 119% operating margin is the FY2023 writedown reversal mechanically flowing through one quarter; ignore it as a margin signal. The clean read is: from Q1 FY2025 onward, operating margin has held positive in every quarter — that is the first stretch of clean profitability since 2022.

3. Cash Flow and Earnings Quality

This is the section that should make a careful investor pause.

Free cash flow is cash generated after operating needs and capital expenditures. If reported net income exceeds cash earnings persistently, the gap usually shows up later as a writedown, an inventory build, or an equity raise to fund the gap. Mirae has all three patterns.

The pattern is uncomfortable. In four of the last five fiscal years, free cash flow has been either negative or modest. The headline ₩10.1 billion of FY2025 net income converted to negative ₩5.2 billion of operating cash flow, a gap of roughly ₩15 billion that almost certainly sits in receivables (the customer ramp shipped late in the year) and in inventory build for FY2026. FY2024's near-zero OCF combined with ₩36 billion of capex required ₩18 billion of financing — most of which came from issuing new shares.

Major cash-flow distortions

Item	FY2023 (₩B)	FY2024 (₩B)	FY2025 (₩B)	What it reveals
Operating cash flow	6.69	0.33	-5.18	Real cash collection has lagged reported profits for two years
Capex	-0.19	-36.26	-16.91	A capacity build cycle ~10× revenue in FY2024 alone
Free cash flow	6.50	-35.94	-22.09	Two consecutive years of cash-burn even as earnings turned positive
Financing cash flow	1.83	18.18	9.38	Funded by share issuance and modest new long-term debt (₩6.5B added in FY2025)
Stock-based comp	—	—	0.24	Small relative to dilution from issuance

The cash flow statement says Mirae has been a cash-consuming, capital-raising story for the last two years, not a cash-compounding one — despite headline profits.

4. Balance Sheet and Financial Resilience

Mirae's balance sheet is still defensible — equity is 80% of assets, total debt is small, and the company holds more cash than it owes. But the direction of travel is the wrong one: cash has fallen by 72% in four years and long-term debt appeared for the first time in FY2025.

Liquidity and working capital snapshot (FY2025)

Line item	FY2024 (₩B)	FY2025 (₩B)	Read
Cash and equivalents	13.5	10.9	Down again, despite a "profitable" year
Current assets	42.0	59.4	Bulked up — likely receivables and inventory
Current liabilities	20.3	19.6	Stable
Long-term debt	0.0	6.5	First time on the balance sheet in this 5-year window
Shareholders' equity	106.0	122.6	Grew via retained earnings plus share issuance
Net cash (cash − debt)	13.5	4.4	Net cash buffer has thinned by two-thirds

The current ratio of 3.04× looks comfortable, but the quick ratio of 1.17× shows that almost half of current assets are inventory — the riskiest asset on a cyclical handler maker's balance sheet because it depends on the next customer order to convert. Equity-to-assets of 80.5% is genuinely strong; this is not a leveraged company. But the cash position and the appearance of long-term debt warn that the financing slack is being used.

5. Returns, Reinvestment, and Capital Allocation

This is where the small print becomes the big story. Mirae has earned a positive ROE in four of the last five years — but the share count has nearly tripled to fund the capex cycle, so per-share returns have lagged.

Share count and dilution

Shares outstanding went from 1.66 million (FY2021) to 4.47 million (FY2025), a 169% increase. That is the cost of funding the FY2024 capex blow-out plus refilling the cash buffer after FY2023's loss year. There have been no buybacks and no meaningful dividends in this window — every won of capital returned has been injected, not returned.

Capital allocation (cumulative FY2021–FY2025, ₩ billions)

Use of cash	Cumulative ₩B	Comment
Capex	-56.6	Heavy capacity expansion concentrated in FY2024 (₩36B)
Net debt change	+6.5	First long-term debt taken on in FY2025
Equity issuance (net financing)	+43.6	Funded the capex; explains the dilution
Dividends / buybacks	~0	No capital returned to shareholders
Operating cash generated	+7.1	Five-year cumulative operating cash flow

The cumulative scorecard says: Mirae has raised more outside capital in five years (~₩50 billion via equity and debt) than the business has generated organically (~₩7 billion). On a per-share basis, EPS in FY2025 (₩2,396) is half of FY2021's ₩4,914 even though net income is higher. That gap is the dilution tax.

6. Segment and Unit Economics

Mirae operates in two segments — ATE (test handlers, burn-in sorters) and SMT (chip mounters, linear motors). ATE has always carried the economics and FY2025 made that even more extreme.

Units shipped — the operational tell

Metric	FY2023	FY2024	FY2025
ATE units	14	23	90
ATE revenue (₩B)	5.89	11.92	36.31
Implied ATE ASP (₩M/unit)	420	518	403
SMT units	8	28	13
SMT revenue (₩B)	0.87	3.64	1.59
ATE factory utilization	n.a.	n.a.	97.5%

ATE unit volume grew ~6× year-on-year in FY2025; SMT shrank to less than half of FY2024 units. The story is overwhelmingly an ATE volume story — and within ATE, overwhelmingly one customer.

Geographic and customer concentration (FY2025)

Customer	Country	Revenue (₩B)	% of total
YILINING PRECISION	China	28.7	56.5%
SK Hynix	South Korea	6.33	12.5%
YMTC (Yangtze Memory)	China	1.78	3.5%
MSV Systems	Singapore	0.22	0.4%
All others	—	13.75	27.1%
Total		50.78	100.0%

Export revenue is 80.1% of the total; China-linked customers (YILINING plus YMTC) alone are ~60% of revenue. Domestic Korea is only 19.9%.

7. Valuation and Market Expectations

The headline multiples on FY2025 close numbers look cheap. The current multiples, after a +169% one-year rally in the stock (price ₩27,600 on 21 May 2026), are still modest — but the implied earnings the market is now paying for is roughly double last year's reported number.

Multiple history

Current snapshot (post-rally)

Metric	FY2025 close (year-end mkt cap ₩61.8B)	Current (21 May 2026, mkt cap ~₩123B)
P/E (on FY2025 NI ₩10.1B)	6.1×	~12.2×
P/Book (on equity ₩122.6B)	0.50×	~1.00×
EV/EBITDA (on FY2025 EBITDA ₩10.2B)	5.8×	~11.4×
P/Sales (on FY2025 revenue ₩50.8B)	1.22×	~2.4×

Choice of metric. For a cyclical capital-equipment business that has just turned the cycle, EV/EBITDA and P/Book are more honest than P/E (since trough-year P/E is meaningless). Current EV/EBITDA in the low double digits is no longer a deep-value multiple — it is closer to peer-group average — and current P/B at ~1.0× is roughly fair for a business earning a 7.9% ROIC.

Quick bear / base / bull frame

Scenario	Assumption	Implied FY2026 revenue (₩B)	Implied FY2026 net income (₩B)	Fair value vs today
Bear	YILINING orders halve, no offsetting customer	~30	-5 to 0 (loss)	down 40–60%
Base	FY2026 holds FY2025 scale plus modest growth; cash conversion improves	60–70	12–15	flat to up 15%
Bull	YILINING ramp continues, second large customer added, HBM exposure broadens via SK Hynix	90+	20+	up 50% or more

The bull case requires customer diversification and clean operating cash flow conversion. Without both, the stock is asking the buyer to trust a single-customer story at near-peer multiples.

8. Peer Financial Comparison

The natural peer set is Korean test-equipment names (Techwing, Hanmi Semiconductor, ISC) plus the two western/Japanese benchmarks (Advantest, Cohu). Numbers are FY2025 close (Advantest's fiscal year ends 31 March; FY2026 figure shown for its just-reported year).

Note on currency. Reporting currency varies — peers are shown in their own reporting currency. The USD version of this page restates everything in dollars. Advantest's fiscal year ends in March; FY26 figures are its year ended 31 March 2026.

The peer gap that matters. Mirae's operating margin (18.0%) and ROIC (7.9%) sit between the cyclical loss-making Cohu and the cash-printing duo of HANMI Semiconductor and Advantest. The two profitable Korean peers (HANMI, ISC) earn 27–44% operating margins and 25–45% ROIC because they sit in structurally favourable niches (HBM TC bonders, test sockets) with pricing power. Mirae sells a more commoditised handler product, and its returns reflect that. The valuation discount to those peers (Mirae trades roughly in line with Techwing on P/E, but at a fraction of HANMI's revenue multiple) is deserved, not a market mistake — the question for the bull is whether the FY2025–FY2026 customer cycle changes the structural earnings profile or just produces one strong cycle.

9. What to Watch in the Financials

Metric	Why it matters	Latest value	Better	Worse	Where to check
Operating cash flow	Tests whether reported earnings are real	-₩5.2B (FY2025)	turns positive in FY2026	stays negative two more quarters	cash flow statement
Receivables and inventory days	Where the ₩15B earnings/cash gap is sitting	not disclosed quarterly	both drop in Q2–Q3 FY2026	both keep climbing	balance sheet schedules
YILINING PRECISION revenue share	Single largest concentration risk	56.5% of FY2025	drops toward 30–40% as other customers ramp	climbs above 60%	DART business report top-customer disclosure
Capex run-rate	Tells whether the FY2024 build cycle is done	₩16.9B (FY2025)	falls to ₩5–8B in FY2026	another ₩30B+ year	cash flow statement
Share count	Dilution risk if FCF stays negative	4.47M (FY2025)	flat through FY2026	new equity raise	quarterly filings
Operating margin	Earnings quality through the cycle	18.0% (FY2025)	holds at 15–20% across all FY2026 quarters	drops to 10% as mix shifts	quarterly income statement
Long-term debt	Liquidity slack	₩6.5B (FY2025)	flat or repaid	grows past ₩20B	balance sheet

What the financials confirm. Mirae is profitable again, the FY2023 disaster has been worked through, and order momentum is strong enough to sustain FY2026.

What the financials contradict. The headline earnings strength is not yet showing up in cash. Two consecutive years of free-cash-flow burn while the company raises equity is the opposite of what a high-quality compounder looks like. The stock's +169% rally already prices in a clean cycle.

The first financial metric to watch is operating cash flow conversion in the Q2 FY2026 filing. If reported earnings of roughly ₩4–6 billion per quarter convert to positive operating cash flow without further equity issuance, this is a re-rating story. If the cash gap widens — receivables or inventory keep building — the earnings recovery will look more like the FY2024 mirror image of FY2023's writedown than a sustainable compounding business.

---

Web Research — What the Internet Knows

The Bottom Line from the Web

The filings show a Korean semiconductor test-handler maker whose FY2025 revenue nearly doubled. The web reveals the part the filings underplay: control of Mirae Corporation flipped in July 2023 to a holding chain anchored by Nexturn Bioscience / Nexturn Roll Korea, with intermediate stakes routed through KOSDAQ-listed talent agency Studio Santa Claus Entertainment and ROA Holdings (chairman On Seong-jun, listed as the highest-paid "non-operational" executive), and that on 2026-01-04 Nexturn Roll Korea announced it would inject ₩6.499993696 billion of additional funding into Mirae. Pair that with the issuer's own disclosure that 56.5% of FY2025 revenue came from a single Chinese counterparty (YILINING / Yiling Trading) — and the bottom-line picture is a small-cap precision-equipment business riding a Chinese-memory order surge under the control of an entertainment-and-biotech holding pyramid.

What Matters Most

Ten findings, ranked by how much they would change an investor's view of Mirae Corporation today.

1. Controlling shareholder is a non-industrial holding chain — and just injected more capital

On 2026-01-04 Mirae disclosed that it expects to receive ₩6,499,993,696 in funding from Nexturn Roll Korea Co., Ltd. (per Marketscreener news feed for 025560 — https://www.marketscreener.com/quote/stock/MIRAE-CORPORATION-6494937/). Control of Mirae passed from Gwangrim to Nexturn Bioscience in July 2023; specialist queries trace the chain as Nexturn Roll Korea (~32.31%, mostly via CB Series 8 and 9 conversions) → Studio Santa Claus Entertainment (KOSDAQ:204630, formerly Huayi Brothers Korea / Fleet Entertainment, ~18%) → ROA Holdings (~27%, chairman On Seong-jun / 온성준). That a loss-making entertainment-and-biotech holding pyramid is now the controlling shareholder of a 1983-vintage semiconductor-equipment maker is the single most consequential governance fact external sources surface.

2. Customer concentration is severe and geographically/politically loaded

Forensic, Warren, Industry and Historian specialist queries all flag the same fact: a single Chinese counterparty accounted for 56.5% of FY2025 revenue. The counterparty is identified in two single-supply DART contract disclosures as Yiling Trading (YILINING PRECISION), announced on the Marketscreener feed for 2025-09-08 as "Mirae Signs Two Semiconductor Inspection Equipment Supply Contracts with China's Yiling Trading." English-language sources could not establish Yilining's ultimate beneficial owner, leaving open whether the equipment is flowing downstream to a US Entity-Listed counterparty such as CXMT (added Dec 2024) or YMTC (added Dec 2022).

3. Top-line surged 97.68% but margins compressed — operating leverage went the wrong way

EMIS (https://www.emis.com/php/company-profile/KR/Mirae_Corporation_en_1651085.html) summarises the FY2025 YoY change set:

Net sales almost doubled, yet EBITDA shrank −5.37% and operating-profit margin fell 16.18 percentage points. The classic interpretation — Mirae is shipping more equipment at thinner mix and worse pricing — is not contradicted by anything else in the search results.

4. Sustained DART-filed order book, but average ticket size is small

The Marketscreener news ticker (https://www.marketscreener.com/quote/stock/MIRAE-CORPORATION-6494937/) records a heavy 2025–26 cadence of single-supply contract DART filings, summarised in the timeline table further down. Total disclosed order flow across the last twelve months is in the ₩50–60 billion range, against FY2025 revenue of ₩50.78 billion — a credible forward-bookings proxy that suggests revenue momentum is not yet rolling over.

5. Lagging peer set despite the recent rally

Yahoo Finance places Mirae's market cap at ₩123.272 billion (close 2026-05-21, ₩27,600). Marketscreener's peer table shows 5-day +26.32%, 1-year +169.11%, 3-year −61.52%. The peer ATE/handler group (KLA, Advantest, Cohu, Teradyne) rallied +130–700% over the same 3-year window. The +94% one-month surge in April–May 2026 that the technicals specialist flagged is closer to peer catch-up than to a fundamental re-rating; no English source attributes a specific catalyst.

6. Equity-restructuring activity that disqualifies historical share-count comparisons

Investing.com split history (https://www.investing.com/equities/mirae-historical-data-splits) records a 16:1 reverse stock split on 2025-06-27 (0.0625-for-1) and a scheduled 5-for-1 forward split on 2026-07-27. The reverse split is the standard tell of either KOSDAQ minimum-price compliance or a clean-up after heavy CB-driven dilution. Any per-share or share-count series sourced from before 2025-06-27 must be adjusted before it can be compared to post-split data; few third-party feeds do this correctly for KOSDAQ small-caps.

7. FY2023 was a near-collapse — and the financial statements were restated

Specialist queries flag a FY2023 revenue collapse of roughly −61% and a net loss in the ₩31.5 billion range under former CEO Sun Jong-eob, coinciding with the broader memory-test downturn and with the July-2023 control flip. Forensic specialist queries (Korean-language "재무제표 정정") indicate the FY2023 financial statements were restated, but no English-source reconciliation was found. The combination of an inflection-low FY, a CEO change, a controlling-shareholder change, and a restatement in a single 12-month window is the kind of event cluster that warrants reading the Korean DART filings directly.

8. The FY2024 ₩36.26 billion PP&E spend has no disclosed counterparty in English sources

Forensic specialist query "FY2024 capex 36 billion PP&E land building facility purchase counterparty" returned no English-source identification of who Mirae bought the land/building/facility from. With control resting in a related-party holding chain, the natural follow-up question is whether the seller was an arm's-length party or another entity within the Nexturn / ROA / Studio Santa Claus pyramid.

9. No analyst coverage; no governance score; thin insider-transaction data

Investing.com, AdvFN and Yahoo all surface placeholder pages for "analyst ratings and price targets for 025560" but no actual coverage is published. ISS Governance QualityScore is N/A. Simply Wall St ownership page reads: "Insufficient data to determine if insiders have bought more shares than they have sold in the past 3 months" (https://simplywall.st/stocks/kr/semiconductors/kose-a025560/mirae-shares/ownership). For a stock that has rallied +169% over a year, the absence of any sell-side or governance-watchdog scrutiny is itself a finding — the rally is happening in a coverage vacuum.

10. Heavy noise from unrelated "Mirae" entities — most lawsuit/regulatory hits do not apply

Web searches for "Mirae" return dominant results for Mirae Asset Securities (KRX:006800), Mirae Asset Global Investments, Mirae Industry (civil-engineering, Siheung-based), Mirae IT, Mirae Eng (oil & gas) and a US 485BPOS SEC filing for a Mirae mutual fund. All flagged regulatory/litigation hits — Ryze v. Mirae Asset Securities arbitration (July 2023), PS Financial ₩200bn won fraud (Jan 2025), SEC/FINRA BrokerCheck items — belong to the financial-services group, not to Mirae Corporation 025560. A reader who relies on a name-only news scrape will be badly misled on litigation exposure.

Recent News Timeline

The cluster of high-significance items in late 2025 / early 2026 — the Yiling Trading contracts, the ₩14 bn and ₩10 bn orders, then the Nexturn funding round — is the spine of the post-control-flip recovery narrative. The cadence of small DART-filed single-supply contracts in March 2026 is the secondary tape that the market appears to be trading.

What the Specialists Asked

Thirty-three specialist questions were issued across nine specialist queues. Below: the questions, the synthesised answer, the source signal and a confidence note. Questions where English-language sources returned no primary evidence are flagged explicitly — those are the gaps the Korean-language DART filings would need to close.

Governance and People Signals

The official roster from Yahoo Finance (https://finance.yahoo.com/quote/025560.KS/profile/) and the DART entity record:

Three patterns stand out:

• ISS Governance QualityScore is N/A. No independent governance scoring exists. A reader cannot triangulate board quality through the usual screens.
• Highest-paid executive is "non-operational." Specialist queries flag On Seong-jun (chairman of ROA Holdings, up the control chain) as receiving the largest compensation despite a non-operational title. The DART executive-pay disclosure is the primary source — English coverage does not reproduce the number.
• No insider-transaction signal. Simply Wall St explicitly reports "Insufficient data to determine if insiders have bought more shares than they have sold in the past 3 months." For a stock up +169% in twelve months, the inability to verify whether insiders are taking liquidity is the relevant gap, not the absence of a flagged sale.

The Ryze v. Mirae Asset Securities arbitration (July 2023), PS Financial ₩200bn won fraud (January 2025), and various SEC/FINRA BrokerCheck items returned by name-only searches refer to Mirae Asset Securities (KRX:006800) — a separate Park Hyeon-joo group company — and do not apply to Mirae Corporation 025560.

Industry Context

The Industry tab already covers the structural primer. The external evidence that newly bears on Mirae specifically:

• Global semiconductor-equipment market is sized at USD 115–138 billion in 2025–26, growing 9–10% CAGR to USD 220–330 billion by 2031–35 per Coherent Market Insights and Mordor Intelligence — but these forecasts are dominated by front-end and lithography (ASML, Applied, Lam, KLA, Tokyo Electron). Mirae's back-end test-handler and SMT mounter segment is a much smaller slice and is not separately sized in the returned pages.
• HBM is the dominant industry catalyst through 2027 — 2 nm mass production starting late 2025, 1.6 nm starting late 2026, HBM4 ramping — but the named HBM-test winners are Hanmi Semiconductor and Techwing, not Mirae. No source ties Mirae's product roadmap to HBM known-good-die test flows. This is the largest structural risk if it remains true.
• SEAJ forecasts Japanese semiconductor equipment sales at ¥5.35 trillion in fiscal 2026 (kantenna.com). Korean small-cap equipment makers are positioned as second-tier suppliers into both Korean memory (SK Hynix, Samsung) and Chinese memory; recent disclosed orders suggest Mirae's recent revenue mix tilts more Chinese than Korean.
• US BIS back-end export-control expansion is the most-watched 2026 rule change. Back-end test equipment is largely exempt today; if controls extend, Mirae's 56.5%-of-revenue Chinese counterparty is the direct exposure.

---

Web Watch in One Page

Mirae Corporation is a sub-scale Korean test-handler vendor whose FY2025 P&L (revenue ₩50.8B, net income ₩10.1B) was driven by one Chinese customer — YILINING PRECISION / Yiling Trading at 56.5% of revenue — at a 49% ASP concession. The verdict is Watchlist: the order book is real but cash conversion (FY2025 OCF -₩5.2B, FCF -₩22.1B), an unresolved ₩30.8B PP&E reclassification, and parabolic tape (RSI 77, +110% over the 200-day SMA) block ownership at ₩27,600. Every variable that decides whether this is a 5-to-10-year compounder or a one-cycle rental sits in a disclosure that has not yet arrived. The five monitors below cover those disclosures: the Q2 FY2026 mid-August cash print, US/Korean back-end export-control posture, the CXMT IPO supplier slate, capital-structure discipline around the 2026-07-27 forward split, and the HBM/Changchuan competitive flank that decides the multi-year case.

Active Monitors

Rank	Watch item	Cadence	Why it matters	What would be detected
1	DART quarterly cash conversion + YILINING customer concentration	Daily	Bull and bear cases pivot on the Q2 FY2026 OCF print. The cumulative five-year scorecard is net income ₩0.7B against FCF -₩49.8B, plugged with ₩43.6B of financing — one printed quarter of positive OCF on ≥₩18B revenue is the only thing that moves this off Watchlist.	New quarterly business reports (분기보고서); single-supply contracts (단일판매공급계약), counterparty name, ticket size, and any (정정) amendment for delayed payment; first appearance of a non-Yiling customer (SK Hynix, Samsung, ASE, Amkor, JCET) above 10% of revenue.
2	US BIS / Korean MOTIE export-control actions on back-end test equipment	Twice daily	This is the binary tail risk. CXMT and YMTC are already on the US Entity List; back-end test handlers are currently exempt. A single Federal Register notice covering Korean back-end shipments to Chinese memory customers — or the addition of Yiling Trading to a US designation list — would reverse 56.5% of FY2025 revenue inside one quarter.	Federal Register notices on Entity List, Unverified List, or foreign-direct-product coverage of back-end test; OFAC/SDN/50%-rule additions of Yiling Trading or YILINING PRECISION; Korean MOTIE alignment statements or new strategic-items license requirements; Chinese reciprocal countermeasures on Korean back-end imports.
3	CXMT IPO docketing and Korean supplier-slate confirmation	Daily	Korean trade press has named Mirae alongside Jusung and Nextin as a CXMT-tier supplier, but no primary CXMT filing has confirmed it. A prospectus that explicitly names Mirae or Yiling Trading converts the FY2025 P&L from a single-cycle artifact into a multi-year revenue franchise; a delay or alternative slate disconfirms the bull case.	CXMT prospectus filings at HKEx or Shanghai STAR Market; CSRC review status; the equipment-supplier list and named purchase volumes; wafer-capacity targets vs the ~200k wafers/month indication; IPO size revisions from US$4.13B; any Changchuan substitution into the supplier slate.
4	Capital-structure discipline + governance signals around the forward split	Daily	History is share count +169% before a 16:1 reverse split, ₩43.6B of cumulative financing plugging the cash burn, and the controlling chain booking ₩23.6B FY25 combined net loss. A new CB / rights / secondary inside ±90 days of the 2026-07-27 5-for-1 forward split repeats the FY2024 dilution pattern. Any auditor change or footnote on the unresolved ₩30.8B PP&E reclassification is thesis-grade.	DART 주요사항보고서 for CB, rights, or secondary placements; related-party loans, advances, asset transfers, or guarantees from Mirae to Nexturn Roll Korea, Studio Santa Claus Entertainment, or ROA & Co; auditor change away from Moon Hyung-suk / Hyundai Accounting; (정정) restatement of FY2024 or FY2025 financials; outside-director changes or audit-committee formation; insider share sales by the controlling chain.
5	HBM product extension + Changchuan substitution at Chinese memory customers	Weekly	The two slow-moving variables that decide the multi-year case. Mirae has no HBM-specific handler while Hanmi, Techwing, and Advantest already participate; Changchuan has grown from sub-5% to ~12% global handler share and qualifies new package types yearly — each qualification removes one Mirae-addressable order at YILINING / CXMT / YMTC.	Mirae announcement of a named HBM handler model; disclosed SK Hynix or Samsung HBM-line contract; R&D meaningfully above the historical ₩3B/year line; Changchuan or other Chinese domestic vendors (Hwatsing, Shanghai Precision) winning qualifications, single-supply orders, or named contracts at YILINING, Yiling Trading, CXMT, or YMTC.

Why These Five

The report's central tension is that every variable in the long-term underwriting map — cash conversion, customer diversification, capital-allocation discipline, regulatory durability, HBM optionality — resolves through a disclosure that has not yet arrived. Monitors 1 and 4 cover the disclosures Mirae itself controls (DART filings, capital actions, governance). Monitor 2 covers the single binary risk that no operating improvement can offset. Monitor 3 covers the one external confirmation — the CXMT IPO supplier slate — that would convert the FY2025 ramp into a multi-year revenue base. Monitor 5 covers the two slow-moving competitive flanks (HBM extension and Changchuan substitution) that decide whether Mirae is a 1.5-2× book outcome or a 0.4× book outcome on a five-year view. Together they target every "what would change the view" item flagged in the verdict, catalysts, and long-term-thesis sections; nothing on this watchlist is generic news collection.

---

Where We Disagree With the Market

The market is paying ~1.0× book and ~12× FY2025 P/E for an earnings recovery the cash statement explicitly refused to confirm — and our evidence says that gap is structural, not a timing artifact. Consensus, visible in a +154% twelve-month rally, trade-press coverage naming Mirae as a "CXMT-tier" Korean supplier, and a controlling-shareholder ₩6.5B injection read as commitment, has effectively underwritten the FY2025 P&L as the start of a structural earnings step-up. We disagree on three measurable points: cash conversion is a five-year structural pattern (cumulative net income ₩0.7B vs FCF −₩49.8B), the YILINING win was a price concession (ATE ASP cut 49% from ₩978M to ₩502M per unit), and the unexplained ₩30.8B PP&E reclassification sits unpriced inside a coverage vacuum. The single most important resolving signal is the Q2 FY2026 operating cash-flow line in the mid-August DART quarterly business report.

Variant Perception Scorecard

Variant strength is high but not maximal because the three disagreements share one underlying resolution — the Q2 and Q3 FY2026 cash flow line — and a single clean print could refute all three at once. Consensus clarity is only moderate because there is no published sell-side coverage; we are reading consensus off price action, trade-press framing, and how the tape reacted to the ₩14B and ₩10B contracts disclosed in late 2025. Evidence strength is high because the five-year cumulative cash gap, the per-unit ASP series, and the FY2024 PP&E unwind are all observable in DART filings and would survive a hostile institutional review. Time to resolution is short — the Q2 FY2026 quarterly business report lands roughly 84 days from now and is the single observable that decides whether the FY2025 print was a working-capital timing artifact or the structural signature of a price-taker.

Consensus Map

The five issues above are not equally consensus. The earnings-durability and customer-durability assumptions are the strongest — they are visible in the price level, the trade-press framing, and the bull anchor. The forensic profile and parent-commitment readings are weaker; they look more like "absence of disconfirmation" than active conviction. We focus the disagreement ledger on the three issues where consensus is clear, observable, and material to a PM's underwriting decision.

The Disagreement Ledger

Disagreement #1 — cash conversion. Consensus, anchored in the +154% rally and Bull's ₩42,000 anchor at 12× FY2026E NI, treats the FY2025 cash flow statement as a quirk. Our evidence is that the five-year cumulative cash gap is structural — cumulative net income ₩0.7B sits against cumulative FCF of −₩49.8B, financed by ₩43.6B of equity and CB issuance — and FY2025 widened the gap rather than closing it. If we are right, the market is paying an earnings multiple on an income statement that has not produced cash in any full cycle, and the equity needs a book-floor anchor (0.4-0.6× book) rather than an earnings multiple. The cleanest disconfirming signal is one Q2 FY2026 OCF print of ≥₩2B against revenue ≥₩18B with no new financing — that would refute the structural-gap reading immediately.

Disagreement #2 — ASP and the CXMT-supplier narrative. Consensus, visible in the November-December rally on the ₩14B and ₩10B contracts and the February 2026 trade-press CXMT coverage, treats the YILINING relationship as a qualifying win that delivers a durable ₩50B+ revenue base. Our evidence is that ATE per-unit revenue collapsed from ₩978M to ₩502M (−49%) in a single year — the operational signature of a price-taker, not a moat-protected vendor. If we are right, the bull case has no support even if revenue holds, because incremental volume at lower price erodes the margin assumption embedded in the ₩42,000 anchor. The cleanest disconfirming signal is FY2026 ATE revenue ÷ units shipped: a print above ₩700M/unit with stable volume validates the franchise read; a print at or below ₩500M confirms the price-taker.

Disagreement #3 — the FY2024 PP&E reclassification is unpriced. Consensus has not formed because nobody is looking — there is no sell-side, no governance score, and the ₩30.8B PP&E movement requires side-by-side reading of two annual balance sheets to surface. Our evidence is that the line moved off PP&E in FY2025 without a matching cash receipt or disclosed disposal, leaving an audit-grade open question on the largest single asset movement in the five-year file. If we are right, the rally has been pricing earnings on a balance sheet whose composition is unverified, and any (정정) amendment or auditor change becomes the trigger for a forensic short. The cleanest disconfirming signal is a clean FY2025 audited footnote that names an arm's-length counterparty with cash receipts visible; the cleanest confirming signal is a related-party identification or a 정정 filing.

Disagreement #4 — the Nexturn injection direction. Consensus reads ₩6.5B flowing into Mirae as a parent vote of confidence; our evidence is that the chain above Mirae booked combined FY2025 losses of ₩23.6B, the injection landed exactly when Mirae's OCF turned negative, and the long-run incentive of a loss-making controlling chain is to harvest a cash-positive subsidiary, not to fund it. If we are right, the 2026-07-27 5-for-1 forward split is the optics-reset moment to watch for a follow-on capital action — exactly the pattern that surrounded the July-2025 16:1 reverse split. The cleanest disconfirming signal is two consecutive fiscal years with flat share count and no new CB issuance; the cleanest confirming signal is any new equity-linked filing inside ±90 days of the forward split.

Evidence That Changes the Odds

The eight items above are not equally load-bearing. The first three — five-year cash conversion, ASP collapse, and PP&E reclassification — are the spine of the variant view; each is independently observable and each has a defined resolution window inside the next two quarterly reports. The remaining five are corroborating; they harden the case but a clean Q2 print could refute the lead three regardless.

How This Gets Resolved

The six signals above are ranked by decision value. The Q2 FY2026 OCF print sits at the top because a single number from one disclosure can refute three of the four ledger items simultaneously: it tests the cash-conversion thesis (Disagreement #1), it constrains how much receivables build can be attributed to YILINING ramp timing (Disagreement #2), and the absence of a concurrent capital-action filing tests the parent-commitment thesis (Disagreement #4). The PP&E footnote signal sits at #3 because the audit review is the only forum where the FY2024 movement gets independently sourced, but the resolution window is longer. The BIS signal is binary and uncorrelated with operating evidence; it sits at #6 because no operating improvement offsets it and no operating deterioration accelerates it.

What Would Make Us Wrong

The cleanest way we are wrong is if the FY2025 cash gap really was a working-capital timing artifact of the late-year China shipment schedule, and Q2 FY2026 prints positive OCF on stable revenue. That outcome would not refute the five-year cumulative cash gap, but it would shift the question from "structural price-taker" to "cycle-leveraged industrial in an up-leg," and the multiple the market is paying would have an economic basis. If the same Q2 also shows YILINING revenue share drifting toward 45-50% with a non-Yiling counterparty appearing in the disclosure stream, two of the four ledger items break at once and the variant view collapses to the forensic profile alone — which is interesting but not investable on its own.

The second way we are wrong is on ASP. A 49% per-unit decline in one year is consistent with both a price-taker concession and a one-time mix shift toward simpler legacy package types as Mirae enters a new buyer's footprint. If FY2026 segment disclosure shows the ASP normalising back toward ₩700M as the mix re-balances, the price-taker reading weakens materially. That is a real path — Mirae is a small vendor at the start of a new customer relationship, and entry-level package types do command lower per-unit pricing across the industry — and we do not have enough years of YILINING-specific disclosure to rule it out.

The third way we are wrong is the PP&E footnote. A clean K-IFRS sale-leaseback with an arm's-length counterparty and visible cash receipts in the FY25 audited statements would resolve the cost-shift signal entirely. Sale-leasebacks are common in capital-intensive industrials and the right-of-use reclassification under K-IFRS 16 is mechanically what the FY24/FY25 PP&E shape looks like. We are willing to be wrong on this one signal because the alternative explanations are routine and the audit window for verification is short.

The fourth way we are wrong is on the controlling chain. Nextern Roll Korea has now injected capital into Mirae once, and there is a real (if narrow) scenario where the chain is funded externally and continues to support the operating subsidiary through the next cycle. In that scenario, Mirae's capital-allocation history is a closed chapter as of January 2026 and the next downturn is the first one underwritten by a committed parent rather than the market.

The first thing to watch is the Q2 FY2026 operating cash flow line in the DART 분기보고서 expected around 2026-08-14 — a positive OCF print on revenue at or above ₩18B with no concurrent capital filing refutes three of the four disagreements at once; another quarter of NI-positive, OCF-negative confirms the structural reading we are arguing here.

---

Liquidity & Technical

Mirae Corporation is a thinly-traded KOSDAQ microcap whose daily liquidity has briefly surged on a violent rally, but in steady state cannot absorb meaningful institutional size — twenty-day average daily traded value is roughly ₩7.8 billion (about $5.2M USD), and the sixty-day baseline is less than half of that. Technically the tape is in a parabolic, late-stage breakout: price sits 110% above the 200-day moving average, RSI is at 77.75, and 30-day realized volatility has jumped to 120% — well into the stressed band — making fresh entries at the spot price a momentum-chase rather than an institutional position trade.

1 — Portfolio implementation verdict

2 — Price snapshot

The one-year double sits inside a three-year drawdown — investors who held since 2023 are still underwater roughly 36%. The current price is therefore a sharp counter-trend rip against a long secular decline from the 2016 high near ₩215,000.

3 — Critical chart — 10-year price with 50/200 SMA

Price is above the 200-day moving average — and by an extreme margin. At ₩27,600 versus an SMA200 of ₩13,136, the spot price is roughly 2.1× the 200-day. That is not a benign uptrend; it is a vertical move. The full ten-year picture remains a secular downtrend from the 2016 ₩200,000-area highs to the 2025 ₩10,000 floor, with the current spike representing the first decisive break of a multi-year base.

4 — Relative strength

A benchmark and sector comparator were not available in the pipeline for this Korean ticker, so the chart shows only the company's own three-year return path on a rebased scale. The story it tells is unambiguous: from May 2023 through April 2026 the stock lost roughly four-fifths of its value before doubling in a single month to recover to 62 on the rebased scale. Even now, after the 154% one-year rally, holders from the 2023 peak are still down a third — this is not "outperformance," it is a partial recovery from severe drawdown.

5 — Momentum panel

Near-term momentum is unambiguously bullish but at the edge of its useful range. RSI(14) at 77.75 is the highest reading in the entire eighteen-month window — above the 70 overbought line for the first time in two years. The MACD histogram is exploding: the last reading at +998 is more than three times the prior eighteen-month peak (+290 on the late-April spike), which says the 12-day EMA has decoupled violently above the 26-day. Buyers are in control on a one-to-three-month view; the issue is degree, not direction.

6 — Volume, volatility, sponsorship

The recent uptick is being confirmed by volume — emphatically. The 50-day average daily volume has roughly tripled from ~50,000 shares in early 2026 to nearly 160,000 by 21 May, and the past two weeks have averaged 540,000–840,000 shares per day. That is real participation, not just price drift on no flow. But realized volatility at 120% sits well above the five-year 80th-percentile band of 78% — the market is demanding a much wider risk premium than normal, and that is what happens at trend peaks as often as trend births. The top three historical volume spikes were all single-day fireworks in past cycles, two of which (2022-04-05, 2020-04-17 not shown) marked local tops with -21% and -24% same-day reversals.

7 — Institutional liquidity panel

Fund-capacity table

Read the table from right to left: at a typical 10% participation rate, a 5% position is implementable inside five trading days for funds up to roughly ₩95B (~$63M USD). The 20% ADV figures double that — but they assume the operator is willing to be one-fifth of the tape every day, which leaves a noticeable footprint and forfeits the ability to be patient on bad fills. The 60-day ADV (₩3.0B/day) is less than half the 20-day (₩7.8B/day), so the capacity at "normal-regime" volume is closer to ₩37B (~$24M) AUM for a 5% position at 10% participation — the recent ADV is a function of the rally itself, not durable depth.

Daily-range proxy

Median intraday range over the last sixty sessions is 5.05% — well above the 2% threshold that flags elevated implementation cost. Combined with the small ADV, this means transaction-cost analysis on any block of more than ~₩500M will show meaningful slippage. Patient block execution via the KRX block-trade window (장중·장후·시간외 대량매매) is the realistic path; sweeping the order book intraday will pay 50–150 bps in impact at minimum.

Bottom line: the largest position a typical institutional book can build inside five trading days at a 10% ADV constraint is roughly ₩24B (~$16M) on the durable 60-day baseline. At 20% ADV that doubles, but exit symmetry requires the same patience on the way out, in markets that may be a lot less friendly than today's tape.

8 — Technical scorecard + stance

Three-to-six month stance: neutral, with a near-term bearish tilt for any new long entered at the spot price. The trend, momentum, and volume signals are all genuinely bullish, but RSI at 77.75 and realized vol at 120% are exactly the conditions under which sharp mean-reverting drawdowns happen — and the company has no fundamental cushion (revenue ₩50.8B FY2025, market cap roughly ₩123B at current shares, so the entire equity value can swing 20–30% in a week without changing the business). Two levels to anchor on: a decisive close above ₩30,800 (the 52-week high) with realized volatility cooling toward 60% would validate a genuine regime change and merit a starter long; a break below ₩19,075 (the 20-day moving average) would break the parabolic structure and bring the ₩13,136 200-day back into play as the next test. Liquidity is the binding constraint — even with a constructive thesis, this is a watchlist name for any fund larger than roughly $50M AUM at 5% position weight, and the correct action for everyone else is to build slowly over multiple weeks and only after the volatility cools.

---

Short Interest & Thesis

Short positioning is not decision-useful as a quantitative input for Mirae Corporation in this pipeline: the run staged zero rows of official Korean Exchange (KRX) reported short-interest balance, zero short-sale volume rows, no borrow data, and no public net-short disclosures. The qualitative read still matters, though — the Korea-wide short-sale ban was fully lifted on 31 March 2025, so short selling has been legally permitted in 025560 for roughly fourteen months, and no public short-seller report on the company has surfaced in prior web research. The genuine downside risk is not a crowded short or an activist report; it is an unwritten short thesis already visible in the staged forensic, governance, and concentration evidence — combined with an extremely illiquid, parabolic tape that would be very hard to exit in either direction without moving price.

1 — Bottom line

2 — Source classification — what is staged, what is not

The institutional read is straightforward: this run cannot quantify whether shorts have stepped in or stayed away since the ban lifted. It can only describe the regulatory regime, the underlying thesis risk, and what crowding would look like if shorts existed at scale.

3 — The Korean short-selling regime — short selling is now permitted

What this means for 025560 specifically:

• Short selling has been legally permitted on 025560 for roughly fourteen months as of 2026-05-22. Any short interest is the product of actual market participation, not regulatory artifact.
• 025560 is a small KOSDAQ industrial — it is not in the KOSPI 200 or KOSDAQ 150 universe that had partial short-selling permission during the 2021–2023 window. Pre-2025 short data points effectively do not exist for this name.
• Under the post-reform regime, corporate net-short positions of 0.01% or more of issued shares (excluding balances below ₩100M) must be disclosed to KRX within two business days. With 4.47 million shares outstanding, 0.01% is only ~447 shares — a low absolute trigger that would surface even modest discretionary shorts if they exist.
• The absence of any net-short threshold disclosures in this run is consistent with one of two equally likely realities: no professional short has crossed the disclosure threshold, or the run's collector did not query KRX 공매도잔고대량보유 entries. Either way, do not treat it as confirmed.

4 — Latent (unwritten) short thesis already in the staged evidence

No published short-seller report on Mirae Corporation (025560) has been identified. But the forensic, governance, and customer-concentration work already done on this name reads exactly like the raw material a short would write up. The institutional question for a long is: would the page below be hard or easy to dispute in front of a risk committee?

5 — Crowding analysis — could shorts actually cover?

Even hypothetical short positioning is hostage to liquidity. The exit math is brutal.

Two observations:

• On the recent inflated 20-day window (₩7.8B daily traded value), even 10% of shares outstanding clears in roughly 1.3 trading days. That sounds easy.
• On the steady-state 60-day window (₩3.0B daily traded value, ~144k shares), 10% of shares short would take 3.1 days at 20% ADV — and the move would be visible to anyone watching the tape. More importantly, the recent ADV is itself a product of the parabolic rally. If the rally reverses, ADV will collapse first and any actual short would be covering into shrinking depth.

The realistic crowding constraint is not "can the short cover" but "what happens to volume when retail momentum unwinds." Mirae's 60-day baseline of ~₩3B/day is a thin-trading microcap signature — the technical agent already flagged it as specialist / illiquid.

6 — Market setup — positioning interacts with a parabolic tape

A name that is +93% YTD, RSI 78, sitting 110% above its 200-day moving average, and running at 120% annualized vol is exactly the kind of tape where:

• A squeeze on a new short could be violent in the first 24-48 hours because depth is thin (₩3B/day steady state) and short interest disclosure has a 2-day lag, so the position would not be visible to retail until after the fact.
• A de-risking on a fundamental short could also be violent — if a forensic report were published on the FY2024 cost reclassification or a US export-control action hit the YILINING/CXMT/YMTC chain, there is no long-only bid deep enough to absorb the supply at current levels. The same 5% daily range that helped on the way up cuts both ways.
• A mean-reversion unwind without any short involvement is the modal risk. The name has tripled in dilution since FY2022 and the rally is concentrated in roughly five months; existing shareholders may sell before any new short steps in.

For a PM, the practical conclusion: short positioning is not the marginal driver. The marginal driver is whether the rally has buyers left.

7 — Borrow & shareholder structure context

The borrow pressure read is honest: we don't have it. A directional inference based on float structure says lendable supply on this name should be tight rather than abundant — a single 32.31% controller plus listed-affiliate holdings dominate the cap table — but tight borrow on a microcap is unsurprising and not by itself decision-useful.

8 — Evidence quality and limitations

9 — Verdict — what should a PM do with this?