Park Systems Corp.(KOSDAQ 140860) _ A company that leads the world in atomic force microscopes
Park Systems Corp.
A company that leads the world in atomic force microscopes is moving from selling research instruments to selling semiconductor metrology tools. In half a year the order backlog grew 77% while operating profit fell to a quarter of its prior level.
Currency. Korean won figures are shown with a US dollar equivalent in parentheses, converted at a flat ₩1,400 = US$1 for reference only. The rate is not a market quote for any specific date, and figures originally reported in dollars are left as reported.
Scope of this note. It covers the strength of the business itself. Market-price items — share price, market capitalisation, target prices, price-earnings ratios — are excluded from the body because they can cloud the judgement being made here. Contract terms that are required to size dilution, such as a bond's exercise price, are kept.
［Disclosure］ expand
- This is one individual's analysis based on publicly filed disclosures. It does not recommend buying or selling any security and is not investment advice. The author is not a licensed financial investment business or investment adviser and does not provide individual consultations.
- All data and figures are drawn from sources believed reliable, such as regulatory filings, but their accuracy and completeness cannot be guaranteed. Any estimates reflect the author's own assumptions and may differ from actual outcomes; all figures are as of the reference date stated in the text.
- The analysis deliberately sets aside the market's short-term assessment and looks first at the company's core competitiveness and the direction of its fundamentals.
- Investment decisions and their consequences belong to the investor. Loss of principal is possible. This material may not be used as evidence in any determination of legal liability.
- Disclosures and results published after the reference date are not reflected. As of the reference date the author holds no position in the security discussed. The author may, however, buy or sell it after publication without prior notice.
- "The critical analysis in this note is a subjective opinion offered in the public interest — for investors' right to know and for the sharing of information. It is not intended to damage any company's reputation or to interfere with its business."
The bottom line
Being the world's largest supplier of AFMs (Atomic Force Microscopes) and the first-half 2026 results tell two different stories. First-half consolidated revenue was ₩92.4bn (US$66m), down 10.5% year on year, and operating profit fell 74.9%, from ₩25.1bn (US$18m) to ₩6.31bn (US$4.5m).22 Over the same period, however, the order backlog grew 77%, from ₩63.6bn (US$45m) to ₩112.5bn (US$80m).22
How one reads that gap is the whole question here. The evidence so far points to "not shipped yet" rather than "not selling," and the collapse in profit came not from weak demand but from the company adding people, buildings and development spend ahead of 2027. There is, however, no guarantee yet that this front-loaded spending converts into revenue — and the added costs do not reverse.
Both revenue and profit fell in the first half. Revenue −10.5%, operating profit −74.9%. But the decline is concentrated in Q1 — Q1 revenue −22.7% and operating profit −84.0%; Q2 revenue +1.4% and operating profit −64.8%.22
The cause is spending ahead, not failing to sell. Gross margin actually rose, from 67.2% to 68.6%. What cut profit was SG&A, up 28.9% from ₩44.28bn (US$32m) to ₩57.07bn (US$41m). Three lines — salaries (+₩6.55bn), ordinary R&D expense (+₩2.84bn) and depreciation (+₩1.40bn) — account for 84% of the increase, and none of the three reverses.22
Demand has not cooled. New orders of ₩141.3bn (US$101m) in the first half; backlog of ₩112.5bn (US$80m) at end-June, up 77% from year-end. 76% of the backlog is industrial. Plant utilisation was 99.0% for research tools and 97.7% for industrial tools — effectively full.22
Growth depends on changing the arena, and that arena belongs to others. The AFM market was only about ₩760bn (US$543m) in 2025 and Park already holds roughly 27% of it.21 In the next arena — advanced packaging metrology — the leader KLA's revenue in that single business is about 13 times Park's entire industrial line.24 New products (FX-Wafer, TSH310PLP) are mostly scheduled to contribute to profit in 2027.3
There are three claims on the shareholder's slice. Adding the 5.34% from the bond with warrants (BW) and 1.25% from unexercised stock options gives potential dilution of about 6.6%.22·31 First-half operating cash flow was −₩6.04bn (US$4.3m): the company booked ₩7.72bn (US$5.5m) of net profit yet cash went out.22 The dividend sits at the bottom of policy (payout ratio 11.6%) and treasury shares go to employee compensation rather than cancellation.2 One widely repeated claim — that the articles of incorporation were amended to raise the CB and BW ceilings to ₩500bn (US$357m) each — did not happen. The amendment was tabled and then voted down at the March 2026 general meeting, with 42.3% against or abstaining (see 05-a ③).34
Five numbers
01 — Business structure
Not a microscope company, a metrology-equipment company
Park Systems was founded in 1997 by Sang-il Park. He had earlier founded and sold the US AFM maker PSI (Park Scientific Instruments) before returning to Korea to start this company; PSI changed hands several times and today forms the AFM division of Bruker, the number-two player — meaning the world's first- and second-largest AFM businesses trace back to the same person.10
Products split in two: research instruments for universities and institutes, and industrial tools that go into semiconductor lines. Industrial tools divide again into wafer metrology, EUV (Extreme Ultraviolet) mask tools and others, with unit prices differing by a single-digit multiple — so as the industrial mix rises, margin rises with it.18
That mix shift has been real. In 2025, consolidated revenue was 71% industrial, 22% research and 7% service, and industrial was 78% of the backlog.2 In the first half of 2026, however, the mix moved backwards — 65% industrial, 28% research and 7% other on a consolidated basis.22 The company presents the changing character of its earnings as the industrial share rises as the core of its business case.21
The portfolio has been widened through M&A: Accurion of Germany in 2022 (ISE, Imaging Spectroscopic Ellipsometry), Lyncée Tec of Switzerland in March 2025 (DHM, Digital Holographic Microscope) and Rocky Mountain Nanotechnology of the US in April 2026, adding optical metrology and a components axis (see 03-e).22 Consolidated subsidiaries went from eight to nine. The company's stated medium-term picture is a move from "AFM equipment maker" to "integrated metrology solutions provider."2
Customers include major manufacturers — Samsung Electronics, SK hynix, TSMC, Intel, Micron — alongside universities and research institutes, and about 92% of first-half 2026 consolidated revenue came from outside Korea (₩7.1bn domestic versus ₩85.3bn overseas).22 The head office moved from Suwon to Gwacheon in March 2026 and the new building opened in April. The company says more than 26% of its people work in R&D.11·12
From here the question is how that product becomes money. In short, this business ① sells hard-to-substitute tools ② at high prices ③ in small numbers, which is why the ceiling on revenue is not the size of the market but the number of units it can build.
a. What it sells and at what price — the price ladder
Setting the reference figures quoted in the press beside the actuals disclosed through DART (Korea's electronic disclosure system):
| Category | Press reference18 | Disclosed avg. 1H 202622 | 202521 | 202422 |
|---|---|---|---|---|
| Research tool (domestic) | ~₩150m (US$107k) | ₩214m (US$153k) | ₩252m | ₩279m |
| Research tool (export) | — | ₩157m (US$112k) | ₩217m | ₩168m |
| Industrial tool (domestic) | ~₩1.7bn (wafer metrology) | ₩2.13bn (US$1.5m) | ₩1.66bn | ₩1.96bn |
| Industrial tool (export) | — | ₩2.23bn (US$1.6m) | ₩2.11bn | ₩1.82bn |
| EUV mask tool | ~₩3.0bn (US$2.1m) | not separately disclosed | — | — |
How to read this (caution). The disclosed figures are not list prices but the average of what actually shipped that year. The company itself notes that unit prices differ by model, so the average moves with the model mix.21 Research-tool domestic prices fell three years running while industrial prices moved up and down — this does not mean discounting; it reflects which models happened to ship. These are also K-IFRS separate-entity figures, i.e. the price at which headquarters transfers goods to overseas subsidiaries and dealers, so the end customer pays more. The ₩33.1bn gap between 2025 separate revenue of ₩172.5bn and consolidated revenue of ₩205.6bn is the margin captured in that stretch.21
b. Why customers buy it — the places only AFM reaches
AFM sells not because it performs better in general, but because there are things nothing else can measure. The company's own filing sets out the comparison:21
| What is measured | Optical | Electron (SEM) | Atomic force (AFM) |
|---|---|---|---|
| Maximum magnification | Thousands× | Hundreds of thousands× | Tens of millions× (0.01nm) |
| 3-D height and angle values | Difficult | 2-D images only | Quantitative |
| Measurement environment | Ambient | Vacuum only | Ambient, vacuum, liquid |
| Sample preparation | Simple | Elaborate and complex | Almost none |
| Sample damage | None | Beam contaminates or deforms | Almost none |
| Electrical / mechanical / thermal properties | No | No | Yes |
The signature application is hybrid bonding (joining chips directly without adhesive). Bonding-surface roughness must be held to Rq (root mean square roughness) of 0.1–0.2nm, and at present the only practical non-destructive way to measure that is the AFM family.23 How large that opportunity becomes, and who is already standing there, is dealt with in 03.
c. How the money arrives — units × price, and a thin recurring layer
| Item | 2024 | 2025 | 1H 2026 |
|---|---|---|---|
| Research units produced | 219 | 265 | 97 |
| Industrial units produced | 58 | 75 | 43 |
| Consumables & service, share of revenue (separate) | 1.97% | 1.98% | 2.38% |
On a consolidated basis service is 7%. The separate-entity figure captures only headquarters shipments, so it reads lower.22
Revenue is recognised in whole units at delivery. That means quarterly results swing directly on how many tools went out and when. Consumables (cantilevers and the like) and service could become a recurring cushion, but the share is thin enough that today they do not smooth anything.
Customer concentration looks very different depending on the basis — a point that is often quoted incorrectly.22
| 1H 2026 | Top-customer share | What the number shows |
|---|---|---|
| Separate basis (HQ shipments) | Top three = 38% (16 + 12 + 10) | Includes subsidiaries and dealers that headquarters ships to |
| Consolidated basis (end customers) | Customer A 8% · Customer B 3% | Only actual buyers outside the group |
On a consolidated basis, end-customer concentration actually fell from a year earlier (13% and 11%). The sentence "the top three customers are 38%" is true only on the separate basis.
02 — Market position
Sources agree on who is first. They do not agree on how big the market is.
QY Research reported that Park's 2022 AFM revenue (US$84.72m) passed Bruker's (US$78.44m), and kept Park in first place in later years.6·9 MarketsandMarkets classified the company in its 2024 evaluation matrix as the only "Star Player" ahead on both revenue and product coverage.20 Share figures nevertheless differ by source, and estimates of market size vary by more than threefold depending on definition.
a. Market share — by source
| Source · base year | Park | Bruker | Oxford |
|---|---|---|---|
| QY Research · 20226 | 20.3% | 18.8% | — |
| QY Research · 20239 | 20.6% | 18.5% | 10.5% |
| Korea Economic Daily · 20258 | 20% | 19% | 9% |
| Company value-up filing · 20242 | 21.7% | 15.8% | 14.3% |
Every source agrees on the ranking (Park first, Bruker second). Absolute figures diverge within about ±2 percentage points, and the gap between second and third reads narrow depending on the source.
b. How big the market is — the bowl this business sits in
Start with the one cited figure whose definition can be verified. The DART annual report reproduces the MarketsandMarkets table directly.21
| (US$ M) | 2025 | 2026 | 2028 | 2030 | CAGR |
|---|---|---|---|---|---|
| Industrial | 241.8 | 259.2 | 300.9 | 347.7 | 7.5% |
| Research | 299.9 | 318.9 | 364.4 | 414.5 | 6.7% |
| Total | 541.8 | 578.1 | 665.2 | 762.2 | 7.1% |
Park's 2025 consolidated revenue of ₩205.6bn is roughly US$147m, or about 27% of that market. Bruker and Oxford shares are from the company value-up filing (2024 basis).2·21
In other words, within the AFM bowl the company already holds a large share. Future growth depends less on taking more share than on making the market itself bigger (advanced packaging, hybrid bonding) or moving into the market next door (integrated metrology).
The company does not fully believe this table either. Immediately after reproducing it, the annual report adds that "the company estimates the AFM market will grow faster than this."21 No supporting basis is given — read it as the company's view only.
| Source | Market-size estimate |
|---|---|
| MarketsandMarkets (in the annual report)21 | US$541.8m in 2025 → US$762.2m in 2030 (CAGR 7.1%) ← baseline |
| Research and Markets19 | US$707m in 2026 |
| MarketsandMarkets (cited by Sisajournal-e)3 | 3-D metrology market of US$2.4bn in 2025, of which AFM is US$220m |
Interpretation (inference) — the gap between US$220m and US$700m looks like a difference of definition rather than a dispute about growth. Given that Park alone did roughly US$147m in 2025, a US$220m market is arithmetically incompatible with a 21.7% share. Definitions need checking before any of these numbers is used as evidence.
c. What the alternatives are — the real contest is against different physics
Bruker and Oxford in the share table are a ranking contest among AFMs. What customers actually weigh, however, are tools that solve the same problem with different physics, and those are the effective competitors.
| Alternative | Good at | Not good at | Unit price26 |
|---|---|---|---|
| CD-SEM Critical Dimension SEM | Fast, full-lot inspection, the standard for width | Needs vacuum, weak on height, beam damage | ~US$4M–8M+ |
| OCD Optical Critical Dimension | Very fast, non-destructive, whole wafer | Only for repeating patterns, model-dependent | ~US$2M–5M |
| WLI White Light Interferometry | Wide areas quickly, height | Lacks nanometre resolution | Below AFM |
| Stylus profilometer | Cheap, simple | Low resolution, sample damage | Low |
| TEM cross-section | Highest resolution | Requires cutting the sample, very slow | High |
| X-ray (XRR · XRD) | Film thickness, crystallinity | Cannot see individual feature shapes | Medium |
| AFM (Park industrial) | Quantitative 3-D height, non-destructive, material properties | Slow — cannot inspect every unit | ~US$1.2M–1.6M |
These unit prices are industry estimates whose original sources could not be verified. It is safer to take only the direction of the multiples and not the absolute values.
What to read here (inference). Park's industrial tool costs far less than a CD-SEM. For a fab this is not "replacing a workhorse tool" but "adding one more machine alongside," which lowers the barrier to entry. For the same reason, revenue per unit is small and there is a ceiling on how many a fab will take. The company's move to bolt optics onto the platform (NX-Hybrid WLI) and to buy Accurion and Lyncée Tec is, generously, an expansion into integrated metrology — and, read coldly, a detour that concedes AFM's own speed limit.
03 — Growth drivers
The arena it stands in, and the one it is trying to move into
The growth case reduces to one sentence. In the arena it stands in, it already holds a lot and the arena is small; so growth has to come from changing arenas.
a. Today's arena — a bowl already full
Three lines tell the same story.22 Demand is already secured, there is not much more share to take inside the market, and the space to build in is full. So the next leg of growth has to be one of three: ① more places to use the tool, ② the market next door, or ③ more capacity. The company is doing all three at once — and in 2026 all three show up only as cost (inference).
b. The ceiling on revenue — the factory is the limit
| Separate basis | 2024 | 2025 | 1H 2026 |
|---|---|---|---|
| Research capacity | 240 units | 272 units | 98 (196 annualised) |
| Research utilisation | 91.3% | 97.4% | 99.0% |
| Industrial capacity | 64 units | 78 units | 44 (88 annualised) |
| Industrial utilisation | 90.6% | 96.2% | 97.7% |
| Production headcount | 33 | 43 | 38 |
How to read this (caution · inference). Utilisation uses the company's own capacity figure as the denominator, and in the first half of 2026 that denominator shrank — research capacity annualises to 196 units against 272 in 2025, and production headcount fell from 43 to 38.22 The March head-office move may explain it, but the filing offers no explanation. "99% utilisation" is safer read as "full within the capacity currently counted" than as "we could sell everything even with more capacity."
In response the Gwacheon building opened in April 2026 — 15 floors above ground and five below, 27,000㎡ of floor area, with a basement industrial cleanroom designed to hold up to 35 industrial metrology systems.27 Press reports say the company intends to lift capacity toward a ₩400bn (US$286m) revenue scale and is pursuing a second production site on roughly 13,200㎡ in the Yongin semiconductor cluster, targeting occupancy in 2027.27
(Inference) Placing a ₩112.5bn backlog beside ₩400bn of planned capacity suggests the bottleneck over the next two to three years is delivery capability rather than demand. But it cuts both ways — capacity is built first, so fixed costs arrive first, and if revenue does not keep to plan the margin is squeezed first. That worry has already shown up in the numbers: SG&A depreciation rose 61% in the first half (see 04-b).
c. The arena it is moving into — six expected markets
| Expected market | Why AFM is needed | When it opens | Confidence |
|---|---|---|---|
| Hybrid bonding | Dielectric roughness at the bonding interface must be held to Rq 0.1–0.2nm, and the only practical non-destructive measurement today is the AFM family23 | Now, alongside HBM4 and 3-D stacking | High |
| Advanced packaging metrology | RDL step height, micro-bump height variation, TSV | In progress | High |
| GAA leading-edge logic | Sidewalls and heights of 3-D transistors are hard to measure accurately by SEM21 | 2026– | Medium |
| EUV photomask defect repair | Find → mill → verify in one tool (NX-Mask) | In progress | Medium |
| Glass substrate & panel-level packaging | Flatness over large areas (NX-TSH310PLP) | 2027– | Low (early) |
| Bio and medical diagnostics | Measuring living cells in liquid (SICM) | Long term | Low |
The size of the arena changes. The advanced-packaging inspection and metrology equipment market alone is estimated at about US$890.9m for 2025,28 and total semiconductor process control above it runs into the trillions of won. That is a different order of size from the AFM market (₩760bn). CEO Sang-il Park has said he intends to grow the company, through M&A and expansion, into a global player comparable to Nasdaq-listed KLA over the long run.27
Caution (inference). The market sizes above are research-house estimates, and the "threefold spread by definition" problem flagged in 02 applies here too. "The arena is big" and "our share of it grows" are different sentences. This is the passage in this note that most needs care.
d. What it carries in — four new products and their timing
AFM's structural weakness is throughput; it is slower than optical or electron microscopes. Most of the new-product strategy is a way around that weakness.
AFM combined with white-light interferometry: wide areas fast by WLI, fine structures precisely by AFM. Because it adds technology to an existing platform, the price uplift drops through to profitability. Developed for advanced packaging but flowing back into front-end — major non-memory makers beginning adoption.5
The chip is fixed and the measuring head moves; suited to large areas. A 310mm version for PLP was developed at the request of a leading global foundry and packaging firm; the first prototype has been delivered and supply for production qualification is planned for 2027.3
Low-noise, high-precision next-generation platform. The company targets up to 20× the throughput of existing AFMs and 2.5× that of AFP tools.3 These are company targets, not verified measurements. Hybrid bonding and leading-edge processes make speed the condition of adoption, so this product's success is effectively the entry ticket to that arena (inference).
EUV mask defect repair. Demand rises as integrated device makers push part of their DUV mask volume out to specialists.3
The timing converges on 2027. Two of the four products (TSH310PLP, FX-Wafer) only reach production or beta in 2027.3 So the P&L contribution of the new-product cycle is a 2027 event, and 2026 is the year the money is spent first (inference). The company has said revenue above ₩250bn (US$179m) in 2026 is achievable,27 but that is a target, not a result.
e. What it buys — the direction of acquisitions has changed
| Date | Target | Character | Stated purpose and size22 |
|---|---|---|---|
| 2022.10 | Accurion GmbH (Germany) | ISE — adjacent tech | Product diversification |
| 2025.03 | Lyncée Tec SA (Switzerland) | DHM — adjacent tech | Strengthening optical metrology |
| 2026.04 | Rocky Mountain Nanotechnology (US) | Components — vertical integration | "Bringing components in-house" · total consideration ₩1.46bn (US$1.0m); goodwill ₩1.36bn |
The first two widened the company sideways; the third moves it downward. It is small but points a different way — a signal that the company wants to own speed and cost itself (inference).
The filing does not say what component, but the company's own press release does.36 RMN, founded in 2003 in Utah, manufactures AFM probes made from ultra-high-purity platinum and platinum-iridium. Tip radii go down to about 10 nanometres, and the probes are used in modes that pass current or measure potential (C-AFM, KPFM and others). This is Park's first entry into probe manufacturing, and the company said it will build a dedicated probe production facility in Korea.
- Probes are consumables bought continuously while a tool is in use — a way to thicken the thin recurring revenue seen in 01-c (2.4% separate basis)
- Electrical measurement and nanoprobing connect directly to the next-generation products in development
- The target makes speciality probes for electrical measurement, a narrow market. It does not replace ordinary silicon cantilevers
- A new plant in Korea is one more fixed cost — another step added to the cost staircase in 04-b (inference)
There is a funding thread too. When the BW was issued, ₩40bn (US$29m) was earmarked for "acquisition of securities in other entities (details undetermined),"31 but the half-year report narrowed this to "capacity expansion and M&A funding to secure technology."22 The target, timing and amount remain undisclosed.
f. Who is in that arena — here the competitor set changes completely
Tier 1 — those aiming at the fab floor with AFM
| Company | What it brings |
|---|---|
| Bruker (US) | Thirty years of in-line AFM lineage in the InSight 300 and InSight AFP automation systems, and already supplying hybrid-bonding surface metrology.23 The most threatening rival, and evidence that this is not a place only Park can serve |
| Oxford Instruments (UK) | Launched the Jupiter Discovery AFM for large samples in May 2025, leading with faster throughput29 |
| Hitachi High-Tech (Japan) | CD-SEM home turf plus fab customer relationships |
| Horiba · Nanosurf · NT-MDT · Semilab | Research-focused, niche |
Tier 2 — those solving the same problem without AFM
| Company | Scale and moves |
|---|---|
| KLA | Leader in advanced-packaging metrology and inspection with roughly 24.6% share.25 Expects about US$1.1bn (₩1.5tn) of advanced-packaging process-control revenue in 2026 alone24 |
| Onto Innovation | Guiding to more than 50% growth in advanced packaging in 2026. Dragonfly G5 qualified at a leading 2.5-D logic customer24 |
| Camtek | US$25m Hawk order for AI applications from a tier-1 IDM in February 202630 |
| Applied Materials · Lasertec · Nova · ZEISS · Nearfield Instruments | Each aiming at the same place with different physics (e-beam, infrared, X-ray, optical) |
This market is concentrated, with the top five taking about 61.4% of revenue,25 and Park Systems is not yet on that list.
About 13 times. "The opportunity is large" and "we capture that opportunity" are entirely different sentences (inference). The left-hand figure is a separate-basis shipment value, so on an end-customer basis it is larger — but the order of magnitude does not change.
04 — Results
Six years of growth stopped in the first half
| Consolidated | 2023 | 2024 | 2025 |
|---|---|---|---|
| Revenue (₩bn) | 144.8 (US$103m) | 175.1 (US$125m) | 205.6 (US$147m) |
| Operating profit (₩bn) | 27.6 (US$20m) | 38.5 (US$28m) | 42.2 (US$30m) |
| Operating margin | 19.1% | 22.0% | 20.5% |
The 2024–2025 figures match the summary financial data in the half-year report.22 Operating margin is calculated from the two lines. Full-year 2026 forecasts are not carried in this note — the rule here is not to cite brokerage estimates; the only forward figure kept for reference is the company's own target (2026 revenue above ₩250bn).27
a. The 2026 quarterly path 〔all items as disclosed〕
| Consolidated (₩bn) | 1Q25 | 1Q26 | 2Q25 | 2Q26 | 1H25 | 1H26 |
|---|---|---|---|---|---|---|
| Revenue | 50.91 | 39.36 (−22.7%) | 52.31 | 53.04 (+1.4%) | 103.22 | 92.40 (−10.5%) |
| Operating profit | 13.15 | 2.10 (−84.0%) | 11.97 | 4.22 (−64.8%) | 25.12 | 6.31 (−74.9%) |
| Operating margin | 25.8% | 5.3% | 22.9% | 8.0% | 24.3% | 6.8% |
Source: half-year report22
Two things stand out. First, the revenue decline is concentrated in Q1 — Q2 revenue actually rose year on year. Second, Q2 operating profit still fell 65%. Profit did not recover even in a quarter when revenue did, which points to cost rather than revenue as the cause.
b. Where the margin broke — SG&A, not cost of goods
| Consolidated (₩bn) | 1H 2025 | 1H 2026 | Change |
|---|---|---|---|
| Revenue | 103.22 | 92.40 | −10.5% |
| Cost of sales | 33.82 | 29.02 | −14.2% |
| Gross profit | 69.40 | 63.38 | −8.7% |
| Gross margin | 67.2% | 68.6% | +1.4%p |
| SG&A | 44.28 | 57.07 | +28.9% |
| SG&A ratio | 42.9% | 61.8% | +18.9%p |
| Operating profit | 25.12 | 6.31 | −74.9% |
Gross margin actually rose. That matters — even though the industrial share fell (71% → 65%) as noted in 01, that mix change did not cut margin at the cost-of-goods stage. Profit disappeared in SG&A.
So what grew inside SG&A?
| SG&A line (consolidated, 1H, ₩bn) | 2025 | 2026 | Change | Reverses? |
|---|---|---|---|---|
| Salaries | 16.59 | 23.14 | +6.55 (+39.5%) | ✕ headcount added |
| Ordinary R&D expense | 9.05 | 11.89 | +2.84 (+31.4%) | ✕ strategic spend |
| Depreciation | 2.28 | 3.69 | +1.40 (+61.3%) | ✕ it grows further |
| Fees paid | 1.87 | 3.21 | +1.34 | △ |
| Sales commissions | 3.46 | 1.89 | −1.57 (−45.4%) | — |
| All other | 11.03 | 13.25 | +2.22 | △ |
| Total | 44.28 | 57.07 | +12.79 (+28.9%) |
Two things emerge. First, those three lines are 84% of the increase, and none of them reverses. Depreciation in particular appears to include only about three months from the new building, which opened in April, so the second half will carry more of it (inference).
Second, sales commissions — frequently named as the cause of the profit drop — actually fell 45% on a first-half cumulative basis. Looking only at the three months of Q2 they rise from ₩0.98bn to ₩1.56bn, but that is 6% of the ₩9.32bn increase in Q2 SG&A. The disclosure gives thin support for explaining this decline as "one-off costs" (inference).
c. The added cost sits at the subsidiaries, not headquarters
| 1H 2026 | Consolidated | Separate (HQ) |
|---|---|---|
| Revenue | ₩92.40bn (US$66m) | ₩76.34bn (US$55m) |
| Operating profit | ₩6.31bn (US$4.5m) | ₩9.57bn (US$6.8m) |
| Net profit | ₩7.72bn (US$5.5m) | ₩11.62bn (US$8.3m) |
Taken alone, headquarters earned more than the consolidated group. The roughly ₩3.3bn difference reads as losses at consolidated subsidiaries (inference).22 The reversal of direction is what matters — in 2025 it was ₩40.4bn separate against ₩42.2bn consolidated, and in 2024 ₩33.0bn against ₩38.5bn, with subsidiaries adding to profit.21·22
| First half (₩bn) | Separate 2025 → 2026 | Consolidated 2025 → 2026 | Subsidiary share |
|---|---|---|---|
| Total SG&A | 33.14 → 38.48 (+5.34) | 44.28 → 57.07 (+12.79) | +7.45 (58%) |
| Salaries | 7.97 → 9.44 (+1.47) | 16.59 → 23.14 (+6.55) | +5.08 (78%) |
78% of the increase in salaries arose outside headquarters. Headquarters headcount rose by 27, from 392 at end-2025 to 419 at end-June 2026, while average pay per person edged slightly down.22 In other words, headquarters labour cost rose gently and the surge sits at overseas subsidiaries — the most likely explanation for the reversal between consolidated and separate profit (inference). However, the consolidated group reports a single operating segment, so which entity it is cannot be identified.22
d. Where the orders are stacked
| (₩bn) | Backlog end-2025 | New orders 1H | Delivered 1H | Backlog end-June 2026 |
|---|---|---|---|---|
| Industrial | 49.7 | 96.0 | 59.8 | 85.9 |
| Research | 11.5 | 36.6 | 25.7 | 22.5 |
| Other | 2.4 | 8.7 | 6.9 | 4.2 |
| Total | 63.6 | 141.3 | 92.4 | 112.5 |
This table answers the mix worry above. 68% of new orders and 76% of the remaining backlog are industrial. The fall in the industrial revenue share during the first half therefore looks like a delivery that has not happened yet rather than an absence of product to sell (inference). The company does add that at the exchange rate on the filing date (14 August) the backlog is about ₩102.3bn — meaning currency alone moves the figure by around ₩10bn.22
e. What rides on the second half
First-half consolidated revenue was ₩92.40bn.22 To meet the company's stated 2026 target of more than ₩250bn in revenue, the remaining two quarters must produce about ₩157.6bn (US$113m).27 Second-half 2025 revenue was about ₩102.4bn, so that means roughly 54% more than last year's second half (calculated).
Lead times (research 2–3 months, industrial 4–6 months) make recognition of the backlog within the year possible, but if industrial deliveries slip by a quarter the arithmetic does not hold. And as seen in (b), SG&A grows further in the second half. The target above is the company's own figure, not a confirmed result.
05 — Risks
Five things seen from the other side
a. Dilution — two strands, and a third that shareholders blocked
① Bond with warrants (BW) — confirmed terms
The ₩100bn (US$71m) BW resolved on 21 May 2026 was fully paid in on 8 June.31
| Item | Confirmed terms31 |
|---|---|
| Character | Private, non-detachable BW — a hybrid security treated as equity for accounting |
| Coupon | 0.0% running / 0.0% at maturity · maturity 8 June 2056 (30 years) |
| Exercise price | ₩267,747 (US$191) |
| New shares if exercised | 373,486 = 5.34% of shares outstanding |
| Exercise window | 8 June 2027 – 8 May 2056 |
| Subscribers | Kiwoom Securities ₩50bn · NV Mezzanine Growth Equity PEF ₩50bn (managed by Dominus Investment) |
| Use of proceeds | Facilities (capacity expansion) ₩40bn · acquisition of securities in other entities ₩40bn · R&D ₩20bn |
- Zero coupon and classified as equity — consolidated equity rose from ₩225.1bn to ₩340.3bn, cash from ₩75.1bn to ₩104.5bn22
- Dilution only after June 2027, so 2026 EPS is unaffected
- Capital with no interest burden funds capacity and R&D — the money to relieve the "factory is the limit" bottleneck seen in 03-b
- If exercised, 5.34% dilution remains
- From one year after issuance, the company or its designee may call back up to 20% of the bonds
- The specific use of the ₩40bn is still undisclosed (see 03-e)
One addition. "Raised without debt" is true of the BW itself only. Over the same period long-term borrowings (including bonds) rose from ₩42.28bn to ₩50.89bn, and interest paid jumped from ₩0.25bn to ₩1.70bn.22 The company is not debt-free.
② Stock options
As of the half-year report, 87,490 unexercised options remain22 — about 1.25% of the 7,002,469 shares outstanding. During the first half 5,992 options were exercised and 10,000 newly granted, and further grants were approved at the March 2026 general meeting.22
BW 373,486 shares + 87,490 unexercised options. The BW is exercisable only from June 2027.22·31 Bar widths are illustrative, not to scale.
③ "Charter ceilings raised to ₩500bn each" — the agenda item was real; the outcome was rejection
Press reports said the CB and BW issuance ceilings had been raised from ₩100bn each to ₩500bn each.15 Checking the original documents, the content of the agenda item was as reported — but the item was voted down at the general meeting. What was wrong was not the content but the outcome. Agenda item 6, "Partial amendment of the articles of incorporation," bundled the following.35
| Article to be changed | Before → after |
|---|---|
| Art. 16 Convertible bonds (CB) | Issuance ceiling ₩100bn → ₩500bn |
| Art. 17 Bonds with warrants (BW) | Issuance ceiling ₩100bn → ₩500bn |
| Art. 8-2 Classes of shares | Ceiling 500,000 shares → one quarter of total authorised shares, and shares with excluded or restricted voting rights newly added |
| Art. 12 Treasury shares | New clause allowing a holding and disposal plan with shareholder approval |
| Arts. 31, 36-2, 44 | Reflecting the amended Commercial Act (outside director → independent director) |
| Art. 45 Audit committee | Separately elected members 1 → 2 |
A special resolution needs two-thirds of the votes cast. 57.7% did not clear that bar. This also explains why the 21 May BW filing recorded "remaining issuance capacity under the articles" as zero — the existing ₩100bn ceiling was fully consumed by this issue (inference).31
- From a shareholder-rights view one worry is gone — the ceilings were not raised
- The stock-option item passed the same day with over 90% support — what shareholders rejected reads less like dilution itself than open-ended financing capacity (inference)
- It also revealed what the company intended: a fivefold bond ceiling and the basis for issuing non-voting shares were bundled into one item
- Any further mezzanine issuance needs another shareholder vote, and there is no guarantee a once-rejected item passes on a second attempt
Who voted against is not disclosed.
b. A cost base that steps upward
As shown in 04-b, gross margin actually improved, so product mix is not the cause of the margin decline. The real risk is on the cost side. SG&A rose 28.9% in the first half, and 84% of the increase came from salaries, ordinary R&D expense and depreciation.22 What the three have in common is that once they step up they do not step back down.
Put together: the order table in 04-d gives grounds for revenue returning in the second half. Costs, however, do not return. So margin recovers only if revenue grows enough to clear the cost step — which is exactly the second-half requirement in 04-e (about 54% above last year's second half).
c. Profit and cash moved apart
Investing includes ₩45.43bn of PP&E purchases; financing includes ₩100.0bn of hybrid securities. Operating cash flow a year earlier was +₩13.89bn.22
Profit was booked but cash went out. Cash generated from operations nearly vanished, from ₩17.15bn to ₩0.48bn, and after interest paid (₩1.70bn) and income taxes (₩5.27bn) the operating line turned negative overall.
- Inventories rose from ₩54.8bn to ₩61.2bn while revenue fell 10.5%
- ₩45.4bn went into the new building, with Yongin still ahead
- Another year of this structure would consume the capital raised
- The backlog grew 77% in the same period, so this may be inventory awaiting delivery
- The gap was covered by the ₩100bn BW, and cash rose to ₩104.5bn
- If second-half deliveries go to plan, working capital unwinds into cash
There is one way to check: whether third-quarter operating cash flow turns positive. If it does not, the "delayed delivery" reading is hard to sustain (inference).
d. Customer and export concentration, and currency
Industrial revenue tracks leading-edge semiconductor capex, and overseas revenue was about 92% of the total in the first half of 2026.22 Because revenue is recognised tool by tool, quarterly results swing with the order and delivery calendars of a handful of customers — the −22.7% in Q1 is the example.22
The size of the currency exposure has been measured for the first time. The effect on pre-tax profit of a 10% move in each currency (consolidated, as of 30 June 2026):22
| Currency | Effect on pre-tax profit of a 10% move |
|---|---|
| US dollar (USD) | ±₩4.89bn (US$3.5m) |
| Euro (EUR) | ±₩3.63bn (US$2.6m) |
| Swiss franc (CHF) | ±₩1.77bn (US$1.3m) |
| Others (JPY, TWD, SGD) | ±₩1.60bn (US$1.1m) |
| Total | about ±₩11.9bn (US$8.5m) |
That is close to twice the ₩6.31bn of first-half consolidated operating profit. It cuts both ways, and this half it helped — financial income of ₩7.36bn (including ₩4.49bn of FX translation gains) less financial costs of ₩4.18bn leaves +₩3.18bn, which is why net profit of ₩7.72bn exceeded operating profit of ₩6.31bn.22 Put the other way, next half it could subtract by the same order. The sensitivity analysis covers foreign-currency assets and liabilities at the reporting date and may differ from realised results.
e. The direction of competition
Park's profit comes from the industrial premium (its industrial AFM share is cited at 80%).8 As seen in 03-f, Bruker and Oxford are aiming at exactly that industrial market with faster scanning and automation, and Bruker in particular already supplies hybrid-bonding metrology.23 If research-focused rivals succeed in crossing into industrial, margin gets cut before share does (inference). There is a threat from the opposite direction too — if optical metrology catches up to the required precision, the case for AFM itself narrows. On top of that, AFM is slow enough that even in hybrid bonding it is used for sampling rather than full inspection, which puts a ceiling on how many tools a fab takes in the first place.
06 — Shareholder returns
Reinvestment over dividends, and the word TSR
The financial goals in the value-up plan released in May 2026 are to hold an operating margin above 20%, keep ROE above the cost of equity, and upgrade the shareholder-return policy.2 The 6.8% first-half operating margin is far from that goal (see 04-a).
Dividends are set by the board within a range of 10–50% of separate-basis FCF,16 and over the last five years have averaged an 11.6% payout ratio and 12.7% total return ratio — near the bottom of the range. Dividend per share went ₩250 in 2021 → ₩400 in 2022–2023 → ₩500 in 2024–2025.2
The company states that it prefers improving TSR (total shareholder return) through investment and profit growth over raising dividends near term.2 Its stated policy is to use treasury shares for RSU compensation and strategic purposes rather than cancellation, and it did so twice — 2,000 shares on 31 March and 732 shares on 11 June, 2,732 in total, delivered to employees, reducing treasury holdings from 45,327 to 42,595 shares (0.61%).17·32·38
Assessment (inference) — prioritising reinvestment is reasonable for this kind of business. Taken together, though, the shareholder's slice is not confirmed through dividends. Treasury shares go out as compensation rather than being cancelled, options continue to be granted, and the BW adds to the count from June 2027. While the payout ratio stays at the bottom of the range (11.6%), all three strands work in the direction of more shares.
Ownership snapshot (as of 30 June 2026)22
| Item | Value |
|---|---|
| CEO Sang-il Park (personal) | 32.26% (2,258,871 shares) |
| CEO and related parties combined | 32.39% (2,267,760 shares) |
| NORGES BANK | 5.88% (411,989 shares) |
| National Pension Service | 2.89% (202,495 shares) |
| Employee stock ownership association | 0.20% (14,292 shares) |
| Treasury shares | 0.61% (42,595 shares) |
| Minority shareholders | 52.3% (14,772 holders) |
| Shares outstanding | 7,002,469 |
The second-largest holder changed after that date. Capital Research and Management Company, part of the US Capital Group, appeared with 5.01% (351,149 shares) as of 4 August 2026 and had raised that to 6.07% (424,760 shares) by 31 August — a further 1.06 percentage points in about a month, filed for simple investment purposes, not to participate in management.33
It reads both ways. Buying into a half-year in which results fell sharply can be taken as a long-horizon purchase (inference), but a simple-investment filing carries no restriction on selling back. Large-holding reports are also filed after the fact, so the disclosed stake is no guarantee of the current one.
Capital Research and Management Company — 5.01% new on 4 Aug → 6.07% on 31 Aug33
Samsung Asset Management — 5.00% (350,152 shares) in Feb → 2.89% (201,905) in May, selling 148,247 shares in three months33
"Institutions are accumulating" and "institutions are leaving" are both one-sided sentences. The two filings cover different periods (Samsung in the first half, Capital after the half-year results), and both are for simple investment.
07 — What could not be verified
Questions public sources do not answer
Method and limits
This note was written by cross-checking press reports, company IR material and filings, and research-house citations published through 3 September 2026. Brokerage estimates, investment opinions and target prices are not cited. That is why the space for a 2026 full-year forecast is left empty, with only the company's own target kept there for reference.
Scope — only the strength of the business itself (product, market, competition, results, capital structure). Market-price items such as share price, market capitalisation, target prices and price-earnings ratios are excluded from the body. Contract terms needed to size dilution, such as a bond's exercise price, are kept.
Use of primary sources — the original DART filings (annual report 2026.03.19, half-year report 2026.08.14) were read directly for consolidated and separate statements of comprehensive income (three-month and cumulative, with prior-year comparatives), the SG&A note, cash flow statements, borrowings, unit prices, capacity and utilisation, revenue composition, order status, customer concentration, currency sensitivity, stock options, employee data, shareholder composition, general meeting resolutions, subsidiary status, business combinations and the market-size table. In addition, all 38 filings received by DART between 1 January and 3 September 2026 were checked against the list, and the following originals were read directly: the general meeting result (2026.03.26), the proxy solicitation reference document (2026.03.10), two treasury-share disposal reports, four large-holding reports, the material-event report (2026.05.21) and two investor-meeting filings.
Still not reviewed — the original audit report and the IR materials posted on KIND.
To separate fact from interpretation, the author's judgements are marked "(inference)." This is not a recommendation for any investment decision.
Sources
- 1Shinhan Securities report summary — Etoday, 2026.08.25 — excluded from citation in the body under the rule against citing brokerage research
- 2Coverage of the company value-up plan — Bloter, 2026.05.27 (share, revenue mix, dividends, treasury shares, M&A)
- 3Sisajournal-e, 2026.08.20 (corrected 08.21) (TSH310PLP, FX-Wafer, NX-Mask, M&M market size)
- 4iM Securities report summary — Korea Economic Daily, 2026.06.09 — excluded from citation in the body
- 5iM Securities report summary — Edaily, 2026.06.09 — excluded from citation in the body
- 6QY Research citation — PRNewswire, 2023.03.07 (2022 AFM revenue US$84.72m vs Bruker US$78.44m)
- 7Samsung Securities company report, "All-Time Steady Seller," 2025.10.20 — excluded from citation in the body
- 8Korea Economic Daily industry report, 2025.05.19 (80% industrial share, competitive landscape)
- 9Financial News, 2025.08.06 (QY Research 2023 share of 20.6%)
- 10ChosunBiz (via Daum News), 2025.05.12 (founding background)
- 11Park Systems Q1 2026 report (backlog ₩97.8bn, corporate history)
- 12Company profile and new-building announcement — parksystems.com/about-us, PRNewswire 2026.04.07 (26% CAGR, 26% R&D headcount)
- 13FnGuide company overview (2024 results, accessed 2026.09.03) — price-related items excluded
- 14WISEreport company status (accessed 2026.09.03) — price items excluded
- 15Coverage of the Korea Corporate Governance Forum watchlist — Daily Brief (CB/BW ceiling increase) — the agenda content matches the original, but the item was rejected; see 05-a ③
- 16Park Systems dividend policy (10–50% of FCF)
- 17Treasury share disposal coverage — Data Investment, 2026.05.13
- 18Korea Economic Daily, 2024.05.13 (tool prices of ₩150m / ₩1.7bn / ₩3.0bn)
- 19AFM market report — GlobeNewswire, 2026.01.30 (market of US$707m)
- 20MarketsandMarkets number-one citation — PRNewswire, 2025.08.30
- 21Park Systems annual report (2026.03.19), DART — 2025 unit prices, production, utilisation, revenue and order status; the MarketsandMarkets market table (2025–2030); AFM/SEM technical comparison; the company's own view on market growth; notice of the proposed charter amendment; 2025 employee data
- 22Park Systems half-year report (2026.08.14), DART — the principal primary source. Consolidated and separate statements of comprehensive income, SG&A note, cash flow statement, borrowings and interest paid, summary financial data, revenue composition, average unit prices, capacity, utilisation and production headcount, order status by product, revenue by region, major customers, currency sensitivity, unexercised stock options, treasury share disposal, employee data, holders above 5%, general meeting resolutions, charter amendment history, subsidiary status, business combination (RMN), planned use of issuance proceeds
- 23Bruker application note AN-5001, hybrid bonding surface metrology — Rq 0.1–0.2nm requirement, InSight 300/AFP
- 24KLA advanced-packaging process control outlook — Yahoo Finance (Zacks), 2026.07.30 (about US$1.1bn in 2026; Onto 50%+ growth outlook)
- 25Advanced-packaging metrology and inspection market share — dataintelo, 2026.03 (research-house material; directional only)
- 26CD-SEM and OCD tool prices — exponentialindustry.com (2026.07), semiconductorinsight.com (2025.07) (industry estimates; original source unverified)
- 27Gwacheon building and capacity plans — ZDNet Korea 2026.04.09 (building size, 35-system capacity) / TheBell 2026.02.19 (₩400bn capacity, Yongin site) / Money Today 2026.06.02 (KLA-scale ambition, 2026 revenue target)
- 28Advanced-packaging inspection and metrology market size — Global Growth Insights, 2026.06 (about US$890.9m in 2025) (research-house material)
- 29Oxford Instruments Jupiter Discovery AFM launch — included in MarketsandMarkets report, 2025.05
- 30Camtek Hawk US$25m order (2026.02) — cited in a factmr report (secondary citation; original unverified)
- 31Material-event report (resolution to issue bonds with warrants, 2026.05.21) and issuance result (2026.06.08), DART — exercise price, share count, exercise window, hybrid-security treatment, subscribers, use of proceeds, call option, remaining issuance capacity under the articles
- 32Treasury share disposal result (2026.06.15), DART — 732 common shares delivered to employees on 2026.06.11; 42,595 treasury shares (0.61%) after disposal
- 33Large-holding reports (short form), DART — Capital Research and Management Company 2026.08.07 (new, 5.01%) and 2026.09.03 (change, 6.07%); Samsung Asset Management 2026.02.23 (new, 5.00%) and 2026.05.29 (change, 2.89% — a reduction of 148,247 shares). Both filed for simple investment purposes
- 34General meeting result (2026.03.26), DART — outcome and vote split by agenda item (item 6 rejected: 57.7% in favour of votes cast, 42.3% against or abstaining, 40.4% of total shares outstanding); items 7 and 8 (stock options) and item 9 (treasury share plan) approved
- 35Proxy solicitation reference document (2026.03.10), DART — the before/after table for agenda item 6
- 36Press release on completing the Rocky Mountain Nanotechnology acquisition — PRNewswire, 2026.04.30 — platinum and platinum-iridium solid-metal AFM probes, tip radii to about 10nm, plan for a dedicated probe facility in Korea
- 37Investor-meeting filings ×2 (2026.08.18, 2026.09.01), DART — Shinhan Securities NDR on 19–20 August and Shinhan Premier Corporate Day on 2 September; materials posted on KIND
- 38Treasury share disposal result (2026.03.31), DART — 2,000 common shares delivered to employees on 2026.03.31
댓글
댓글 쓰기