Zeus Co., Ltd. (079370.KQ) - Deep Dive Research Report

Prepared 2026-06-06. KOSDAQ-listed semiconductor / display equipment and industrial-robot manufacturer, Hwaseong, South Korea.

A note on sources and the "four concalls" requirement. Zeus is a ~600-employee KOSDAQ mid-cap. Like the overwhelming majority of KOSDAQ companies, it does not host public quarterly earnings conference calls and publishes no concall transcripts. After searching the company IR site (globalzeus.com), DART (dart.fss.or.kr), KIND (kind.krx.co.kr) and Korean transcript aggregators, no quarterly concall transcript exists for this issuer. The substitute, used throughout and cited by period, is the four most recent statutory DART periodic disclosures - Q1 FY2026 분기보고서 (filed 2026-05-15), FY2025 사업보고서 (filed ~March 2026), Q3 FY2025 분기보고서 (filed 2025-11-14), and 1H/Q2 FY2025 반기보고서 (filed ~August 2025) - supplemented by management statements carried in Korean press releases and sell-side earnings reviews. Where I attribute a forward-looking statement to management, it comes from these filings or from on-record press quotes, not from a fabricated transcript. This is flagged again in Sections 7 and 9.

1. What the Company Does

Zeus makes the machines that clean, heat-treat, and move silicon wafers and display panels inside chip and screen factories - and, increasingly, the robots that do the moving. Strip away the catalogue and there is one core craft at the centre of the company: getting microscopic contamination off a wafer without damaging it. That is what a wet-cleaning tool does, and it is the single largest thing Zeus sells.

The founding arc explains why the company looks the way it does. The Zeus lineage traces to a trading and equipment business established in 1970 (originally ZEUS COMM. Corporation), rebranded to Zeus Co., Ltd. in December 1988 under founder Lee Dong-ak, and listed on KOSDAQ on 1 February 2006. The founder's eldest son, Lee Jong-woo - a University of Michigan electronics engineer with a KAIST MBA who spent his early career as an engineer in Silicon Valley at MACOM, Magma Design Automation and Cadence - joined in 2004 and became CEO in 2011. Since March 2025 he has run the company as co-CEO alongside Hwang Ha-seop, a former Samsung Electronics vice-president - a deliberate hire that signals where management wants the relationships and the revenue to come from. The founding family controls roughly 39-40% of the shares (Lee Jong-woo personally ~22.1%), so this is a founder-controlled, second-generation-run company.

The core value proposition is import substitution of Japanese and American process equipment for Korean chipmakers. Korea's two memory giants, Samsung Electronics and SK Hynix, historically bought wet-cleaning and thermal tools from Japan (SCREEN/DNS, Tokyo Electron) and the US (Lam). Zeus positions itself as the domestic alternative - close to the customer, faster to service, and able to localise a process step that the customer would rather not depend on a foreign supplier for. After Japan's 2019 export-control scare, this "탈일본" (de-Japanification) localisation thesis became a national-policy tailwind for Korea's 소부장 (materials/parts/equipment) sector, and Zeus sits squarely inside it.

What makes the product genuinely hard is not the metal box - it is the process recipe and the qualification. A wafer-cleaning tool has to remove particles, organic residue and metal ions down to single-digit-nanometre tolerances, on a wafer worth thousands of dollars, at high throughput, without scratching, etching or warping it. Each tool has to be qualified on the customer's specific process node before it ever runs production silicon, a process that can take many months. Once qualified, it is sticky.

A concrete example of Zeus in action: an HBM (high-bandwidth-memory) maker stacks many thin DRAM dies on top of each other. To grind those dies thin, the wafer is temporarily bonded to a carrier wafer, then later debonded (separated) and the ring frame that holds the warped, ultra-thin wafer flat has to be cleaned of adhesive residue before reuse. Zeus supplies the temporary-bonding/debonding tool (its "TBDB" platform and a photonic-debonding system co-developed with US firm PulseForge) and the ring-frame cleaning tool (branded "ATOM"). When a memory maker builds out HBM capacity, those are exactly the niche tools that get pulled in - which is why Zeus's 2025-2026 story is told almost entirely through the HBM lens.

2. Business Segments

Zeus reports four meaningful businesses. The FY2024 revenue mix (the cleanest recent full-year split) was: semiconductor manufacturing equipment ~66.3%, industrial robots & vacuum ~20.9%, display manufacturing equipment ~9.0%, and plug valves ~3.7%.

2.1 Semiconductor Manufacturing Equipment (~66% of revenue)

This is the company and everything else is an option on it. Zeus designs and builds wet-cleaning equipment in both major architectures: batch type (many wafers cleaned together in a tank, high throughput, cost-efficient) and single-wafer / 매엽 type (one wafer at a time, more precise, required at advanced nodes). It also builds rapid thermal-processing and heat-treatment equipment.

• Core capability: the wet-clean process recipe plus the chemistries, fluid dynamics and surface-tension drying that remove contamination without damaging advanced-node wafers. This is decades of accumulated process know-how, not something a new entrant replicates from a datasheet. Critically, Zeus owns its single-wafer competence partly through its Japanese subsidiary J.E.T. Co., Ltd., a single-wafer wet-processing specialist that is separately listed on the Tokyo Stock Exchange. The irony is sharp: a Korean "de-Japanification" champion sources part of its single-wafer platform from a Japanese subsidiary it controls.
• Why it exists as the anchor: it is the business with the highest customer intimacy (Samsung, SK Hynix), the deepest qualification moat, and the direct line into the HBM capex cycle via the new TBDB, photonic-debonding and ring-frame-cleaning (ATOM) tools.
• Competitive position: competes against SEMES (Samsung's captive equipment arm), SCREEN/DNS and Tokyo Electron (Japan), and Lam Research (US) on cleaning. Zeus wins on localisation, proximity and price; it loses where the customer's most advanced node is locked to an incumbent foreign tool that is already qualified.
• Strategic priority: this is both the cash engine and the growth bet. Management's entire 2025-2026 narrative - HBM tools, the US PulseForge debonding launch, the YES single-wafer distribution deal in the US - lives here.

2.2 Industrial Robots & Vacuum (~21% of revenue)

Zeus's growth diversification play. Under the "ZERO" robot brand, it builds autonomous mobile robots (AMRs), SCARA robots, and 6-axis vertical articulated robots, plus the vacuum-related equipment grouped with it. The initial beachhead is using its own factory-automation and display relationships to sell transfer/handling robots into manufacturing lines.

• Core capability: robot motion control plus, crucially, domain access - Zeus already sells into the same fabs and panel lines where these robots are deployed, so it can co-sell handling robots alongside process tools. Its marquee robot win is a display-process transfer-robot supply contract with Avaco worth a total of ~KRW 47.2 billion, plus supply relationships with Samsung Display.
• Why it exists separately: different technology stack (mechatronics/motion vs. wet-process chemistry), different competitive set, and a deliberate attempt to reduce the company's dependence on the brutally cyclical chip-equipment order book.
• Competitive position: a small challenger against Korean robot makers (Hyundai Robotics, Doosan Robotics, Rainbow Robotics, Yujin Robot) and global incumbents (Fanuc, Yaskawa, Mitsubishi). Zeus is not a robotics leader; it is a process-equipment company leveraging its factory access to sell robots into adjacencies it already understands.
• Strategic priority: the long-dated growth option and the cyclicality hedge. Management talks about pushing robots into overseas markets and larger projects.

2.3 Display Manufacturing Equipment (~9% of revenue)

The legacy growth engine that built the company before chips took over. Zeus makes thermal-process (curing/heat-treatment) equipment, in-line transfer systems, and inspection systems for LCD and, increasingly, OLED panel production. The structural tailwind here is Chinese panel makers converting LCD lines to OLED, which pulls new process and inspection tooling.

• Core capability: panel-scale thermal uniformity and in-line handling at very large substrate sizes - a different engineering problem from wafer-scale work.
• Why it exists separately: different end market (display panels vs. silicon), different customers, and a historical foundation predating the semiconductor pivot.
• Competitive position: competes against domestic and Chinese display-equipment suppliers; commoditisation pressure is higher here than in semiconductor cleaning.
• Strategic priority: mature, cyclical, smaller. Useful when OLED capex is up, not the centre of the story.

2.4 Plug Valves (~4% of revenue)

A small, legacy industrial mechanical-valve business (plug valves for industrial fluid control). It is the oldest commercial DNA in the group, runs at low single-digit revenue share, and is essentially a stable cash-generative tail rather than a strategic priority.

Segment summary

3. Products and Business Detail

Semiconductor cleaning - the catalogue. Zeus carries both batch wet-cleaning stations (tank-based, high-throughput cleaning used broadly across memory and logic) and single-wafer (매엽) cleaning tools (one wafer at a time, needed where advanced nodes cannot tolerate batch-process variation). Alongside cleaning sit rapid thermal-processing / heat-treatment systems. The single-wafer platform is the basis of the "Single Wafer Processing Platform (SWPP)" that Zeus and its Japanese subsidiary J.E.T. market, and which US partner Yield Engineering Systems (YES) agreed to distribute and eventually build in the United States under a strategic partnership announced 10 July 2023.

The HBM-specific tools - the part the market actually cares about now:

• TBDB (Temporary Bonding / De-Bonding): bonds a device wafer to a carrier for thinning, then debonds it. Essential for the thin-die stacking inside HBM.
• Photonic debonding system (with PulseForge): uses high-intensity pulsed light to separate bonded wafers with minimal damage and residue. Announced 19 February 2025, targeted at US integrated-device manufacturers (IDMs) and shown at Semicon Korea 2025.
• Ring-frame cleaning ("ATOM"): cleans adhesive residue from the ring frames that hold warped ultra-thin HBM wafers flat, so they can be reused.

Robots. The ZERO brand spans autonomous mobile robots, SCARA (selective-compliance assembly) arms, and 6-axis vertical articulated arms, commercialised first as transfer robots for display and factory lines (the Avaco contract being the anchor reference).

Display tools. Thermal/curing ovens, in-line transfer systems, and optical inspection systems for LCD and OLED panel fabs.

Manufacturing, geography and milestones. Zeus is headquartered and manufactures in Hwaseong, Gyeonggi Province, at the heart of Korea's semiconductor corridor near Samsung and SK Hynix sites. It is building out capacity with a ~KRW 14.1 billion land investment in the Yongin semiconductor cluster (2024-2027), positioning it next to the next wave of Korean fab construction. Internationally, the Japanese subsidiary J.E.T. (TSE-listed) supplies single-wafer technology and serves Japanese and Chinese customers, and the US PulseForge and YES partnerships are the company's route into the American IDM market. Notable milestones: KOSDAQ IPO (2006); the 2023 YES US distribution partnership; the 2025 PulseForge photonic-debonding launch; and the build-out of the HBM tool set (TBDB / ATOM) that reframed Zeus from a cyclical cleaning supplier into an HBM-packaging name.

4. Customers

Who buys. The semiconductor segment sells to memory makers - principally Samsung Electronics and SK Hynix in Korea, and US IDMs for the new debonding tools. The robot segment sells to display and factory-automation customers, including Samsung Display and Avaco. The display segment sells to Korean and Chinese panel makers.

Concentration. Customer concentration is real: a single customer accounted for roughly 23.7% of revenue in the most recent annual disclosure, and the top handful of memory/display accounts dominate. For a sub-scale equipment vendor selling into an industry with only a few buyers on the planet, this is structural rather than a flaw - there are simply not many companies that buy advanced wafer-cleaning tools. The flip side is that one customer deferring capex can swing a year, which is exactly what happened in 2025.

Who makes the buying decision and how. The buyer is the customer's process-integration and equipment-engineering organisation. The criteria are technical first (does the tool hit particle, defect, throughput and damage specs on our node?) and commercial second. The sales cycle is long: a new tool must pass an on-site qualification on the customer's live process before it earns production orders, which can run many months to over a year.

Why they choose Zeus. Localisation and proximity (same time zone, same corridor, fast service), price relative to Japanese/US incumbents, a national-policy preference for domestic 소부장 suppliers, and - for the HBM niche tools - the fact that Zeus actually built tools for steps (ring-frame cleaning, debonding) that the incumbents under-serve.

Switching costs. High once qualified. Re-qualifying a competing cleaning tool on an advanced node is costly and risky for the customer, so an installed, qualified Zeus tool tends to keep pulling repeat and follow-on orders for that process. The barrier cuts both ways: it protects Zeus's installed base but makes displacing an incumbent at a new customer slow.

Contract structure. Predominantly project-based equipment orders tied to the customer's capex schedule, not recurring subscription revenue. The Avaco robot deal (~KRW 47.2 billion total, with receivables largely collected by March 2026) is illustrative: a large lumpy contract recognised over a delivery period. This lumpiness is why quarterly results are volatile and why Zeus's revenue tracks the memory capex cycle so tightly.

5. Competitive Landscape

Zeus is a mid-tier specialist competing against much larger incumbents in each of its arenas, winning on localisation and niche tools rather than scale.

Semiconductor wet cleaning. The real competitors are SEMES (Samsung's captive equipment subsidiary - both a competitor and, through Samsung, adjacent to Zeus's biggest customer), SCREEN Holdings / DNS and Tokyo Electron of Japan, and Lam Research of the US. The structural reality: SEMES has the inside track at Samsung; the Japanese incumbents own decades of qualified installed base at advanced nodes. Zeus wins the localisation-mandated share, the cost-sensitive batch work, and the HBM niche tools (ring-frame cleaning, TBDB, debonding) where the giants have left gaps. It loses head-to-head at the leading-edge logic clean where an incumbent is already locked in.

HBM bonding/debonding and advanced packaging. Here the named competitors are SUSS MicroTec and EV Group (EVG) in temporary bonding/debonding, with Tokyo Electron also present. Hanmi Semiconductor dominates an adjacent HBM step (thermo-compression bonders) but is not a direct cleaning/debonding rival. Zeus's differentiation is the PulseForge photonic-debonding approach and the ring-frame-cleaning specialty.

Robots. Against Hyundai Robotics, Doosan Robotics, Rainbow Robotics, Yujin Robot (Korea) and Fanuc, Yaskawa, Mitsubishi (global), Zeus is a small entrant whose only real edge is customer access through its equipment relationships. This is the segment where the moat is thinnest.

Barriers to entry. In wet cleaning, genuinely high - process recipe know-how, customer qualification, and installed-base lock-in. A new entrant cannot simply build a tank and sell it; it must be qualified on a live advanced node, which customers grant only to credible, well-capitalised suppliers. In robots, the barriers are far lower and the field is crowded.

Where Zeus is strong: localised wafer cleaning for Korean memory, HBM niche tools, and customer proximity. Where it is exposed: it is sub-scale versus every major incumbent, dependent on a handful of customers, and its robot diversification competes in a commoditising field.

6. Industry

What drives demand. Zeus's fortunes are levered to wafer-fab-equipment (WFE) capex by memory makers, and within that, increasingly to HBM and advanced-packaging build-out driven by AI accelerator demand. Every incremental HBM line needs thinning, bonding/debonding, ring-frame cleaning and wafer cleaning - the exact steps Zeus tools. Secondary drivers are OLED display capex (notably Chinese LCD-to-OLED conversion) and the broader factory-automation/robot adoption curve.

Size and trajectory. The global WFE market is a multi-tens-of-billions-of-dollars annual market; the semiconductor cleaning-equipment slice alone runs into the low-tens of billions of dollars and grows with node complexity (more process steps mean more cleans per wafer). HBM is the fastest-growing pocket, expanding far faster than the broader memory market on the back of AI demand. (Industry sizing per general WFE/SEMI market commentary; exact figures vary by source.)

Where Zeus sits in the supply chain. Upstream-of-production, mid-tier: it is a process-equipment supplier to the chipmakers and panel makers, one tier below the fab operators and one tier above the components/chemicals vendors that feed its tools.

Import substitution. This is the policy backbone of the bull case. Korea historically imported the bulk of its advanced cleaning and thermal tools from Japan and the US. After Japan's 2019 export-control episode, Korea accelerated a national 소부장 localisation drive, and Zeus is a named beneficiary - explicitly framed in trade press as helping lead the "탈일본" (de-Japanification) of cleaning equipment and robots. The localisation share is rising but the leading edge remains foreign-dominated.

Regulation and qualification. No consumer regulator gates this market; the binding "regulation" is the customer's own qualification regime plus export-control sensitivity around advanced semiconductor equipment (relevant to the US IDM and China-exposure dimensions).

Cyclicality. Severe. Equipment orders track the memory capex cycle, which is one of the most volatile in all of industrials - feast in upcycles, famine when memory makers freeze capex. Zeus's 2024-to-2025 swing from solid profit to an operating loss, on a high-teens revenue decline, is the cycle in one frame. The HBM segment is structurally less cyclical than commodity DRAM/NAND tooling, which is precisely why management is steering toward it.

Tailwinds: AI-driven HBM build-out, Korea localisation policy, OLED conversion, US partnership optionality. Headwinds: memory-capex volatility, China display/semi slowdown (which hit the J.E.T. subsidiary), and rising R&D intensity needed just to stay in the HBM race.

7. Growth Triggers

Source caveat: Zeus holds no public concalls. The following are forward-looking items drawn from the four most recent DART periodic disclosures (Q1 FY2026 / FY2025 / Q3 FY2025 / 1H FY2025) and on-record management statements in press releases over the same window. Each is dated to the disclosure or announcement period rather than to a transcript.

• HBM tool order ramp - TBDB and ring-frame cleaning (ATOM). Management positions the newly developed HBM tool set to convert into orders as memory makers build HBM capacity into 2026. (Repeated theme across 1H FY2025 → FY2025 → Q1 FY2026 disclosures.)

• US debonding equipment launch with PulseForge, targeting US IDMs. A photonic-debonding system co-developed with PulseForge, slated for delivery to a major US integrated-device manufacturer. (Announced Feb 19, 2025; reiterated through FY2025 disclosure.)

  CEO Lee Jong-woo: the partnership "combines industry-leading technological capabilities in photonic debonding and semiconductor manufacturing," with first equipment scheduled for delivery to major US IDM customers.

• US single-wafer market entry via YES distribution and local manufacturing. The Yield Engineering Systems partnership gives Zeus's single-wafer (SWPP) platform a US sales/service channel, with a stated long-term plan to manufacture the systems in the US. (Originating July 2023; an ongoing channel in current disclosures.)

• Robot business scale-up - overseas markets and larger projects. Management has guided to pushing the ZERO-brand robot business into export markets and bigger projects beyond the anchor Avaco display-transfer contract. (Repeated across FY2025 and Q1 FY2026 commentary.)

• Capacity build at the Yongin semiconductor cluster. A land investment (2024-2027) to expand production capacity ahead of the next Korean fab-construction wave. (FY2025 disclosure.)

• Co-CEO appointment to deepen Samsung relationships. The March 2025 addition of ex-Samsung EVP Hwang Ha-seop as co-CEO is framed by management as a move to strengthen the Samsung account and source new opportunities. (FY2025 / Q1 FY2026 governance disclosures.)

8. Key Risks

• Memory-capex cyclicality - the dominant risk. Zeus's revenue is project-based and concentrated in a few memory customers, so a capex pause swings the whole company. This is not theoretical: FY2025 swung to an operating loss on a high-teens revenue decline as orders softened, while the company could not cut its skilled-engineer payroll or its (rising) HBM R&D spend. Mechanism: fixed cost base + lumpy orders = operating-leverage whiplash. Calibration: high-probability, recurring, moderate-to-severe drag.

• HBM thesis execution risk. The entire bull narrative rests on TBDB, photonic debonding and ring-frame cleaning converting into repeat production orders. These are newly developed tools that still must win qualification at scale. If memory makers standardise on SUSS/EVG/TEL for bonding-debonding, Zeus's new tools could under-monetise after the R&D was spent. Calibration: medium-probability, high-impact on the growth case.

• Single-customer concentration (~23.7%). One account at roughly a quarter of revenue means a single customer's roadmap or capex timing can dominate a year. Calibration: structural; high-probability moderate volatility.

• Japanese subsidiary J.E.T. drag. J.E.T. posted an operating loss (reported around ¥594 million) on customer delivery delays and a China-market slowdown, weighing on consolidated results even as it supplies core single-wafer technology. Mechanism: a strategically important subsidiary that is currently a financial drag and exposed to China. Calibration: medium-probability moderate drag; watch for the guided 2Q recovery.

• Near-term liquidity tightness. Disclosures show short-term debt due within a year exceeding liquid assets (a reported ~KRW 19 billion gap), bridged in 2025 partly by liquidating inventory to generate operating cash. Healthy as the overall balance sheet is (low debt ratio, large distributable reserves), inventory drawdown is not a repeatable cash source and could constrain future supply if orders snap back. Calibration: low-probability acute risk, monitor.

• Tax / overseas-account dispute. An ongoing litigation over an unreported-overseas-account penalty (~KRW 2 billion). Calibration: low-probability, contained financial impact, but a governance optics item.

• Robot diversification competing in a crowded field. The cyclicality hedge sits in a commoditising market against larger robot specialists; if robots fail to scale, the diversification thesis weakens and the company stays a pure memory-capex play. Calibration: medium-probability, moderate.

9. Walk the Talk

Reporting periods used (no concall transcripts exist for this issuer): 1H/Q2 FY2025 (반기보고서, ~Aug 2025), Q3 FY2025 (분기보고서, 2025-11-14), FY2025 (사업보고서, ~Mar 2026), Q1 FY2026 (분기보고서, 2026-05-15). The most recent is within ~90 days of today. Because there are no verbatim management calls, this assessment cross-references management's stated intentions in press releases and periodic disclosures against subsequently disclosed outcomes - and it is calibrated accordingly (directional, not transcript-precise).

The clearest promise-and-outcome pair is HBM. Through 1H 2025, management leaned hard into the HBM-tooling story - new TBDB and ring-frame-cleaning tools developed, the PulseForge debonding launch announced in February 2025, and press framing the half-year as benefiting from an "HBM effect." The follow-through was mixed: the tools were delivered and the partnerships announced as promised, but the order ramp did not arrive fast enough to offset the broader capex downturn, and FY2025 closed in an operating loss with the loss widening again in Q1 FY2026. So management delivered on product and partnership milestones (debonding system launched on schedule, shown at Semicon Korea 2025, supply intent to a US IDM stated) but the financial payoff lagged the optimism - a pattern of being early and product-credible rather than promotional.

On shareholder returns, management said it would steady the stock and return capital, and it has visibly done so: a treasury-stock trust agreement (a ~KRW 5 billion contract in April 2024, with larger trust figures reported subsequently), treasury-share cancellation, a bonus share issue, and a multi-year unbroken dividend record. Here the talk matched the action.

On the Avaco robot contract, management guided that the ~KRW 47.2 billion deal would be delivered and collected; by the FY2025/Q1 FY2026 disclosures most of the proceeds had been received and the residual ~KRW 7.7 billion receivable was collected by March 2026. Promise kept, cash in the door.

On diversification away from cyclicality, management has repeated the robot-scale-up ambition across multiple periods. This is the least-proven commitment: robots remain a modest share, and the 2025 downturn showed the company is still, fundamentally, a memory-capex name. The intent is consistent and repeated; the outcome is unproven.

Management framing (paraphrased from disclosures and press): the HBM tool set and US debonding launch would drive the next leg; the co-CEO hire from Samsung would deepen the customer relationship.

Assessment. This is product-credible, financially over-optimistic-on-timing management - not promotional, not erratic. They build what they say they will build (debonding tool, ring-frame clean, robot contracts delivered, buybacks executed, dividends maintained), and they are straight about the cyclical pain. Where they consistently miss is timing the order ramp - the HBM payoff has been "coming" across several periods without yet showing up in the P&L. Read them as honest engineers who under-deliver on the when, not on the what. Track the actual HBM order announcements as the credibility test.

10. Shareholder Friendliness Index

Dividends. Zeus has paid a dividend for well over a decade (press cites 12-plus consecutive years), making it unusually consistent for a cyclical KOSDAQ equipment maker. The exact per-share dividend for each of the last three fiscal years could not be verified to a primary filing within the search budget; what is verifiable is the unbroken multi-year dividend record and a large distributable-reserve base (commercial-law distributable profit cited around KRW 161.8 billion), so the dividend is comfortably covered even through the FY2025 loss year. (Stated rather than estimated: I am not inventing a DPS series I could not confirm.)

Buybacks and dilution. Capital return via buybacks has been active and real: a treasury-stock trust agreement of ~KRW 5 billion announced in April 2024 (term to April 2025) explicitly for price stability, with larger treasury-trust figures (around KRW 8 billion) reported subsequently, plus an actual treasury-share cancellation (shares retired, not just held). Offsetting the retirement, the company ran a bonus/free share issue (reported as a 1-for-2 free allotment), which increases the share count without changing economic ownership, and introduced an RSU program in 2025 for talent retention (a modest future dilution source). Net of the cancellation and the bonus issue, the share count has been managed actively rather than allowed to drift up on option dilution.

Verdict: Returns Capital - a rare KOSDAQ cyclical that pairs a decade-plus dividend streak with genuine treasury cancellation, even while absorbing a loss-making year, with the only offset being a bonus issue and a small RSU plan.

11. Insider Activities

Source attempt and limitation. The authoritative venue for Korean insider data is DART (dart.fss.or.kr), specifically the 임원·주요주주 특정증권등 소유상황보고서 (Officer & Major Shareholder Securities Holdings Reports) and 5%-rule filings. DART's filing-viewer and the Korean insider aggregators (fintel.io, goinsider.kr) returned 403/blocked responses to automated retrieval within this search budget, so I could not extract a granular, transaction-by-transaction list of individual open-market insider buys and sells for the trailing 12 months. I am disclosing that explicitly rather than fabricating transactions. What is verifiable from filings and press is set out below.

Ownership and company-level insider actions (verified):

Buys - reading the signal. No specific open-market purchase by the CEO, CFO or an individual director was located in retrievable filings for the trailing 12 months. The strongest company-level conviction signal is the treasury buyback combined with actual share cancellation - management spending corporate cash to retire stock through a loss-making year is a constructive signal, though it is a company action, not a personal-conviction open-market insider buy. The founding family's large, stable ~39% block means insiders are heavily aligned by holding, but I found no evidence of fresh personal accumulation.

Sells - working out the why. No material individual insider open-market disposals were located in retrievable filings within the search budget. The bonus share issue mechanically changes share counts in insider accounts but is not a sale. I will not infer sells that the data does not show.

Net assessment. On the verifiable record, the picture is neutral-to-constructive at the company level and aligned at the family level: a founder family controlling ~39%, active treasury buybacks with genuine cancellation, and a sustained dividend - but no confirmed individual open-market insider buying or selling in the trailing 12 months that I could retrieve. A reader who needs the transaction-level granularity should pull the 임원·주요주주 특정증권등 소유상황보고서 directly from DART for ticker 079370; that primary source was not machine-accessible in this session.

12. Scenarios

Bull case. AI demand keeps HBM the fastest-growing pocket of memory, and the 2026 fab build-out finally pulls Zeus's niche tools into volume. The ring-frame-cleaning (ATOM) and TBDB tools win repeat qualification at Samsung and SK Hynix, and the PulseForge photonic-debonding system lands real production orders at a US IDM, opening a second geography that is far less correlated to Korean capex. The Samsung-veteran co-CEO converts relationship into share gains in core wet cleaning. The robot business, leaning on its display and factory access, graduates from the Avaco reference into export and larger projects, finally giving the company a credible non-memory leg. The Japanese subsidiary J.E.T. recovers as Chinese demand stabilises. In this world, Zeus stops being a pure cyclical and becomes an HBM-packaging-and-cleaning specialist with a diversifying revenue base - and the decade of dividends and buybacks looks like the prelude to a re-rating.

Base case. Memory capex normalises rather than booms. Zeus's HBM tools convert to orders steadily but not explosively, the US debonding channel contributes modestly, and the robot business grows but stays a minority of revenue. Earnings recover off the 2025 trough as the cycle turns, but the company remains fundamentally levered to a handful of Korean memory customers and the timing of their capex. Management keeps doing what it has done: building credible niche tools, maintaining the dividend, retiring stock, and being early-but-honest on the HBM ramp. A solid, founder-controlled, cyclical equipment maker delivering roughly what it guides, with the HBM option still maturing.

Bear case. The HBM order ramp keeps slipping. Memory makers standardise their bonding/debonding on SUSS, EVG or Tokyo Electron, leaving Zeus's new tools under-monetised after heavy R&D spend that could not be cut during the downturn. A prolonged memory-capex freeze meets a fixed skilled-engineer payroll, and the operating loss persists rather than reverses. The single large customer pushes orders out, J.E.T.'s China exposure deepens its losses, and the near-term liquidity gap forces uncomfortable choices once the inventory-liquidation cash lever is exhausted. The robot diversification fails to scale against better-funded specialists. In this world Zeus stays a sub-scale, single-region, single-cycle equipment vendor whose HBM story was real in the lab but never paid off in the order book.

Sources: Stock Analysis - company profile · Wikipedia (KO) - 제우스 · YES / Global Zeus partnership · Businesspost - Lee Jong-woo profile · Bosoop - FY2025 report analysis · Bosoop - company analysis · TheLec - debonding to US IDM · Newsis - PulseForge photonic debonding · ZDNet Korea - KRW 5bn buyback trust · TheBell - J.E.T. subsidiary loss · KIND - Q1 FY2026 분기보고서 (2026-05-15) · DART

A few transparency notes on the deliverable: (1) Zeus does not host public concalls, so Sections 7 and 9 are built from the four most recent DART periodic disclosures plus on-record press statements, disclosed as such. (2) Section 11 individual-transaction granularity could not be retrieved - DART and the Korean insider aggregators blocked automated access - so I reported only verifiable ownership and company-level actions and said so plainly rather than fabricating. (3) Section 13 (Further Reading) is omitted because SemiAnalysis, Stratechery and MBI Deep Dives have no coverage of this company. (4) Per your rules I kept revenue/margin figures out of the report body (segment mix % only); the absolute figures appear solely in the chart-data block, as that block explicitly calls for real research data.

Would you like me to save this as a .md file in the repo, or adjust any section's depth?