ASICLAND (445090) — Independent Filing-Based Company Note

Korea's only channel into TSMC, the first half-year operating profit in three years, and a gross margin that swung from −3.0% to 24.4% quarter by quarter

Date of writing: 27 August 2026

Disclosure cut-off: 25 August 2026

Unit: KRW 0.1 billion (i.e. 100 million won) unless otherwise stated. For example, 1,137.3 means KRW 113.73 billion. Figures are on a consolidated basis unless otherwise noted.

[Disclosure]

This report is one individual's analysis based on publicly available regulatory filings. It does not recommend the purchase or sale of any security.

I am not a licensed financial investment business operator or investment adviser, and I do not provide individual consultations.

Investment decisions and their consequences belong to the investor, and loss of principal may occur.

Reference date: 2026-08-25. Disclosures and results after this date are not reflected.

0. Thirty-Second Summary — Company Overview

① Business model, market and competition. ASICLAND provides design-implementation services and turn-key manufacturing to fabless companies (chip designers that own no factory) by routing their chips through the fabs of TSMC (Taiwan Semiconductor Manufacturing Company, the Taiwanese contract chip manufacturer). It earns money from project service fees and from the margin on wafers it delivers. The market is expanding as demand for custom AI (Artificial Intelligence) chips grows, and the foundry market itself is forecast at USD 210 billion for 2026. Its real rival is not a Korean company but GUC of Taiwan, which holds the same TSMC partner status.

② What the company may become. Today it is a company that books short-term revenue by working down a backlog of design orders it has already won. If the chips it has finished designing move repeatedly into mass production, it becomes a company with stable production revenue that does not depend entirely on winning new orders. Whether that transition has actually happened will be visible in the Q3 2026 report: the test is whether quarterly product (mass-production) revenue exceeds 20% of sales for two consecutive quarters.

③ The key structural observation. ASICLAND is the only Korean company holding official TSMC partner status — but that status was not built by the company; it was granted by TSMC. And in Q2 2026 product (mass-production) revenue rose from KRW 0.31 billion in the same quarter a year earlier to KRW 16.56 billion, producing the first observation of the transition the market has been waiting for.

④ What is confirmed / not yet confirmed. Confirmed: first-half 2026 revenue of 1,137.3 (versus 304.6 a year earlier) and operating profit of 49.9 — and among the four listed Korean peers, ASICLAND was the only one with a first-half operating profit. Also confirmed: the remaining work on hand (development contract backlog) fell from 1,614.1 to 776.2, and in a half-year with positive operating profit, operating cash flow was −141.4. Not yet confirmed: that mass-production supply contracts underpinning production revenue have been fixed by disclosure, and at what level the gross margin — which ultimately determines profitability — settles.

1. What Kind of Company Is This — When It Was Founded and What It Has Done

ASICLAND takes semiconductors designed by other companies, optimises that design for the manufacturing process of a TSMC fab, and — when the customer asks — takes the whole job on a turn-key basis and delivers finished parts.

It was founded in 2016 and listed on KOSDAQ in November 2023. The company is based in Suwon, Gyeonggi Province, and the chief executive officer is Lee Jong-min. Headcount was 260 as of 30 June 2026, down by four from 264 at 31 December 2025. Average tenure is three years and eight months.

It owns no production facilities. Wafer production is entrusted entirely to TSMC, and packaging and test entirely to specialist back-end firms. There are six consolidated subsidiaries; of these, the operating ones are the Taiwanese local subsidiary and a building-management company. The other four — an investment advisory subsidiary and investment partnerships held by the company — do not form separate revenue segments.

The largest shareholder is CEO Lee Jong-min with 2,573,631 shares, a 23.56% stake. There has been no change in the largest shareholder's own holding over the past three years. On the basis of the disclosure dated 25 August 2026, the largest shareholder and related parties held 3,061,541 shares (28.02%), of which 466,738 shares (4.27%) are pledged as collateral for a share-backed loan; the pledge period was extended on 18 August 2026.

During the first half of 2026, the combined stake of the largest shareholder and related parties fell from 40.32% to 30.49%, a decline of 9.83 percentage points. What the disclosures confirm is this: about 6.9 percentage points of that reflects one registered director resigning on 6 April 2026 and therefore dropping out of the "related party" perimeter — that person separately reported holding 755,000 shares on the date of resignation. Another 0.9 percentage points or so dropped out because another executive's title changed. The portion that actually decreased through selling is about 2.0 percentage points.

Note, however, that the two figures above use different reference dates. The 28.02% is as of the 25 August 2026 disclosure and the 30.49% is as of 30 June, and the definition of the related-party perimeter may have changed in between. The author could not determine the nature of the difference.

The filings state that there are no sanctions, administrative dispositions or licence revocations, and no material litigation in progress. However, after the close of the reporting period the company became a co-defendant in one lawsuit (claim amount KRW 3.3 billion), and the filing states that the outcome cannot be predicted.

(Sources: Semi-annual report, 2026-08-14, 20260814000520 / Annual report, 2026-03-19, 20260319001064 / Large shareholding report, 2026-04-13, 20260413000557 / Large shareholding report, 2026-08-25, 20260825000367)

2. Core Technology — What It Makes and What It Does on the Customer's Behalf

2-1. How it works and what sets it apart

The company converts the circuit a design house has drawn into a physical layout that a fab can actually print, and it opens the door to TSMC processes that the customer could not open alone.

Making a chip has three broad stages: ① drawing the circuit that defines what the chip does; ② converting that circuit into a layout that obeys the process rules of a real fab; and ③ printing, dicing and packaging wafers according to that layout. Design houses usually do only ①. ASICLAND handles ②, and if the customer wishes, takes ③ as well and delivers the finished chip. The industry calls this all-in arrangement a turn-key order.

The reason it is hard to do alone lies on the fab side. TSMC does not accept orders from just any company; it takes design volume through firms it has granted official partner status. That status is the VCA (Value Chain Alliance), which the company obtained in 2019, making it the only holder in Korea and one of only seven worldwide. It also holds ARM design partner status.

There is, however, a sentence the company itself has written about the nature of that status. It states that a partner's sales territory is anchored on the region where the partner is based but may subsequently expand worldwide [company statement (disclosure)]. In other words this is closer to a convention than to a contractually locked regional exclusivity — and the same logic applies in reverse to Taiwanese partners.

On the technical record, the company states that it has executed 5, 7, 12, 16 and 28 nanometre process work and completed Korea's first 7-nanometre AI chip [company statement (disclosure)]. The author could not find third-party verification of these items.

2-2. Pipeline and roadmap

Five government-funded R&D projects are currently in development, with government grants totalling KRW 31.092 billion. These five are listed exactly as recorded in the disclosure as "in development", and one of them is recorded as having a research period that ended in December 2025.

Project	Government grant	Research period
2,000 TFLOPS-class server AI deep-learning processor and module development	188.30	~ Dec 2026
Chiplet heterogeneous-integration ultra-high-performance AI semiconductor	77.57	~ Dec 2027
Thermal modelling for AI semiconductor heterogeneous-integration packages	25.20	~ Dec 2027
Vector database accelerator for large language models	13.00	~ Dec 2026
Open-call project	5.85	Ended Dec 2025

Two things can be read here. First, the direction of development sits with server AI chips and with chiplets — the technique of splitting a chip into several pieces and joining them. Second, 201.30, or 64.7% of the grants, ends in December 2026. Those grants have been accounted for as an offset to research and development expense, so when the projects end, the same amount flows through as an increase in reported SG&A (Selling, General and Administrative expenses). How much is allocated to each project each year is not disclosed, so the author could not measure the annual size of the increase.

3. Business Model — How It Makes Money

3-1. What it sells

The filings show a single operating segment, ASIC (Application Specific Integrated Circuit), and revenue splits three ways.

Revenue line	Revenue model	Share of H1 2026	How revenue is recognised
Prototype development	Chip design services	57.9%	Project-by-project; once a project ends the company must win another order
Supply of services (engineer dispatch)	Engineers are placed with customers and their labour cost becomes revenue	23.5%	Recurring; only 9.07 is bought-in external manpower, so most of it is the company's own payroll
Products (mass production)	Delivery of wafers and finished parts	18.6%	Recurring within the contract term; for the company to grow, this line has to grow

By quarter, the three lines move at very different speeds.

Quarter	Products (mass production)	Prototype development	Supply of services	Total
Q1 2025	4.8	127.5	25.4	157.7
Q2 2025	3.1	116.0	27.8	146.9
Q3 2025	5.1	129.8	26.9	161.8
Q4 2025	42.4	194.6	25.2	262.2
Q1 2026	46.3	449.8	43.7	539.9
Q2 2026	165.6	208.8	223.0	597.4

3-2. To whom — customer structure and concentration

Customers are Korean and overseas design houses and system-semiconductor companies. Three customers each accounted for more than 10% of first-half 2026 revenue, at 260.2, 264.5 and 215.0 respectively — 739.7 in total, or 65.0% of all revenue. In the same half a year earlier, two customers together came to 61.9, or 20.3%.

None of the three customers is named. The company does not disclose this item, and the author could not identify who they are.

The counterparties that can be confirmed from disclosed contracts are as follows.

Counterparty	Amount	Term	Status
SK hynix (controller design for enterprise storage — eSSD, enterprise Solid State Drive)	409.6	~ 2027-12-31	Increased from 318.5 on 2026-08-19 as volume rose
Mobilint	165.4	~ 2027-12-31	Increased and extended by one year on 2026-07-30
Company A (mass-production supply)	257.1	~ 2026-12-31	Increased from 254.3 on 2026-08-18; counterparty name withheld
PRIMEMAS	94.7	~ 2027-06-30	In force
FADU	299.8	Ended 2026-08-20	End date pulled forward by 11 months from 2027-07-11
Company F	141.7	Ended 2026-07-31	Counterparty name switched to withheld

(Sources: single sales / supply contract disclosures — 20260819900179, 20260730900296, 20260818900207, 20260820900104, 20260731901028, PRIMEMAS Inc. 2026-01-12, 20260112900138)

3-3. How it gets paid — billing and revenue recognition

Design service revenue is recognised in proportion to how far the work has progressed, with progress measured against costs and hours incurred. Under this method, if the estimate of total cost changes, the timing at which revenue is booked moves with it.

The auditor has designated "measurement and period allocation of development revenue recognised over time" as a Key Audit Matter (KAM) for three consecutive years. The same item was designated even after the auditor changed from Deloitte Anjin to EY Han Young. A Key Audit Matter is a mechanism for flagging high-risk areas, not a statement that something has been found wrong; the audit opinion has been unqualified for three years running, and the filings state there is no material uncertainty about the company continuing as a going concern. Audit hours rose from 860 in the eighth fiscal year to 1,915 in the tenth, and audit fees from KRW 129 million to KRW 185 million. The intensity of verification itself has increased.

There is one place inside the same annual report where the description of revenue recognition is inconsistent. The note to the order-backlog table states that progress is not measured by percentage of completion, while the notes to the financial statements state that progress is measured using the input method. The author could not determine which is a clerical error in the form.

3-4. Cost structure

Purchases in the first half of 2026 came to 831.1, made up as follows.

Purchase item	Amount	Share
IP (Intellectual Property, i.e. design assets) licence fees	338.95	40.8%
Wafer purchases for development	313.46	37.7%
Wafer purchases for products	169.12	20.3%
Outsourced processing and external manpower	9.07	1.1%

Design-asset licence fees and wafer costs together come to 78.5%. This is cost the company does not price itself: it takes the prices set by the IP owners and by TSMC and passes them through. The share of cost that ASICLAND can cut on its own is comparatively limited.

3-5. Pricing power and economies of scale

Unit prices are not disclosed for any period. The company states that it does not record selling prices because processes and specifications differ by customer and commercial protection is needed [company statement (disclosure)]. The same applies to sales volumes. In other words, there is no way to confirm from the filings whether the company has pricing power. The author tried to measure the power to set prices and could not find a single disclosed fact that would show it.

Nor is it yet confirmed that larger scale leaves more behind. In 2024, when revenue was 940.6, the gross margin was in fact 1.1%. Bigger revenue did not translate into a better margin.

4. The Market — Is the Field This Company Stands On Getting Bigger?

4-1. Size of the field and where the company stands

[Fact] The global semiconductor market for 2026 was forecast at USD 909.8 billion, growing 17.8% (Korea International Trade Association, November 2025 forecast), and the foundry market at USD 210 billion, growing 20% (Export-Import Bank of Korea, forecast of 30 December 2025). TSMC's revenue rose from TWD 2,161.7 billion in 2023 to TWD 3,809.1 billion in 2025, and its foundry share in Q3 2025 was put at about 71.0% (TrendForce, announced 12 December 2025).

[Fact] The box that could not be filled. The field in which this company actually competes — the market for design service houses measured on its own — has no published statistics. No third-party tally exists, so the author leaves this box empty. Neither the company nor its competitors disclose market share.

[Interpretation] Market-size figures are therefore used only as reference coordinates. The real coordinates are contracts and revenue. Combined first-half 2026 revenue of the four listed Korean peers was 4,914.8, of which ASICLAND accounted for 1,137.3, or 23.1%. That is second behind CoAsia (51.7%).

[Fact] What the chips are used for. Split by application, the concentration shifts from year to year.

Application	2024	2025	H1 2026
Memory	129.3	234.0	682.7 (60.0%)
Artificial intelligence	331.2 (35.2%)	194.2	130.2 (11.5%)
Consumer electronics	not separately disclosed	not separately disclosed	231.7 (20.4%)
Automotive	170.9	82.0	10.8
Display	27.6	75.1	24.0
IoT (Internet of Things) and 5G	72.6	68.3	16.5
Industrial	not separately disclosed	not separately disclosed	7.0
Other	209.0 (22.2%)	74.8	34.4

In 2024 artificial intelligence was the largest application; in the first half of 2026 memory is 60.0%. Because the company works project by project, the mix swings sharply depending on which application one large customer's project happens to serve.

4-2. Forces that expand the field

Growth engine	What it is and a sense of size	Where it stands now	Evidence
① AI and memory controllers moving into mass production	Memory-application revenue of 129.3 (2024) → 234.0 (2025) → 682.7 (H1 2026); large	In progress	Disclosed actuals
② TSMC expanding leading-edge capacity	TSMC's 2025 capital expenditure of TWD 1,272.6 billion, an all-time high; medium	In progress	TSMC data cited by the company
③ Automotive and overseas customer expansion	Automotive design supply contract with a US integrated device manufacturer; size unknown	Not started	[Press] amount undisclosed — Korea Economic Daily, 2025-06-11 (https://www.hankyung.com/article/202506118526i)

There is one trace of these forces actually reaching the company. Overseas revenue in the first half of 2026 was booked at 312.8 (27.5% of the total), against zero in the same half a year earlier.

But to take that work the company needs people. Having no fab, it carries no burden of adding equipment; instead, headcount is the first place it runs into a wall. Annualising first-half 2026 revenue gives KRW 870 million per employee, and the final year of the author's base scenario (2028 revenue of 2,060) would require KRW 790 million per employee at today's headcount. That level has already been exceeded, so it is not impossible — but it should be read alongside the fact that headcount is falling, from 264 to 260.

4-3. Ceiling and headwinds

[Fact] First, there is no statistic measuring the maximum this company could capture, so the ceiling cannot be drawn as a number. Second, contracts are denominated in dollars. The conversion rate written into contracts rose from KRW 1,379.30 in November 2024 to KRW 1,544.20 in June 2026. So far a weakening won has worked in favour of revenue; if that direction reverses, it works the other way. Third, Korean design houses have another road available to them — the Samsung Foundry channel.

[Interpretation] The field getting bigger and this company's slice getting bigger are two different questions. Because every customer name is withheld, it is hard to trace through the filings the path by which a market tailwind reaches this company.

5. Competition — Who It Competes With, and On What

5-1. Competitors by layer

The market structure is oligopolistic. The company explains that each foundry designates its partner design houses, which makes entry difficult [company statement (disclosure)]. Three things block entry: foundry partner status, design engineers with 15 or more years of experience, and a track record of work actually delivered.

Competitor	Which fab camp	Scale and results	Position relative to ASICLAND
GUC (Taiwan)	TSMC partner	The author could not confirm published data on either share or results	Ahead on scale and track record; has a sales office in Korea [company statement (disclosure)]
Alchip (Taiwan)	TSMC partner	As above	Reported to be strong in AI ASICs [company statement (disclosure)]
ADTechnology	Samsung Foundry camp	2025 revenue 1,645.1, operating profit +28.5; H1 2026 revenue 819.8, operating loss −15.3	Grew revenue 54.4% and turned profitable in 2025, then posted an operating loss in H1 2026
Gaonchips	Samsung Foundry camp	2025 revenue 685.1, operating loss −167.2; H1 2026 revenue 418.8, operating loss −35.8	Similar in size
CoAsia	Samsung Foundry camp	2025 revenue 3,751.3, operating loss −21.8; H1 2026 revenue 2,538.9, operating loss −50.7	Largest by revenue

(Sources: 2025 results — 20260209901090, 20260209900347, 20260213902009 / H1 2026 results — each company's semi-annual report, 2026-08-14; ADTechnology 20260814002567, Gaonchips 20260814003050, CoAsia 20260813001792)

The two years point in opposite directions, so which company is ahead is not settled. In 2025, when ASICLAND's revenue fell 22.6%, ADTechnology's grew 54.4%; in the first half of 2026, ASICLAND was the only one of the four with an operating profit.

5-2. Substitute technologies and substitute products

[Interpretation] Three things are worth watching. First, customers doing the implementation design themselves: small design houses lack the staff, but large customers can run it in-house. Second, switching the fab to Samsung Foundry — in which case the customer uses a Samsung-camp partner design house. Third, design automation tools getting better: as tools improve, the price of having someone else do physical design can be squeezed. The author rates the size of this threat as medium.

5-3. The line of defence and what remains unresolved

[Fact] The company holds one qualification as its line of defence, and that qualification is not owned by the company. TSMC granted it, and the company has itself written that a partner's sales territory may expand worldwide. [Interpretation] So the defence rests on a permission that a third party can redefine.

Two admissions. First — the author could not find a single case of a Taiwanese partner contracting directly with a Korean customer. That is to say, there is also no evidence that this line of defence has already been breached. Second — no company discloses market share, so the author cannot state in numbers how much of the market ASICLAND holds.

The competitor most worth watching is GUC. It holds the same qualification, and the larger Korean customers become, the more reason they have to deal directly with a Taiwanese firm.

6. Moat — How Long Can This Position Be Held?

Source of moat	Applies?	Evidence
Regulation, certification, entry barrier	☑☑	Official TSMC partner status — the only one in Korea, one of seven worldwide; ARM design partner. This is the body of the company's defence.
Switching cost	☑	It is hard to move the design of a chip to another firm midway. But work is project-based, so the next chip is competed for again.
Technology and patents	☑	Track record at 5, 7, 12, 16 and 28 nanometres — a company statement for which the author could not confirm third-party verification.
Brand and reputation	☑	Minister of Trade, Industry and Energy commendation at Semiconductor Day in October 2025; ARM Design Partner of the Year 2023 — both [press], not cross-checked against original filings.
Economies of scale	☐	Gross margin of 1.1% in 2024 on revenue of 940.6 — no scale effect confirmed.
Network effects	☐	Not applicable.

[Interpretation] Overall assessment. Of the six moat sources above, four are marked and two are not. The qualification and the technical record are both real, but no disclosed evidence has yet appeared that the qualification has converted into price or into profit. In 2024 and 2025 — the seventh and eighth years of holding the status — operating profit was −169.7 and −275.8; unit prices are undisclosed for every period; and 23.5% of first-half 2026 revenue is engineer dispatch, where payroll simply becomes revenue.

[Interpretation] The first gate of the analysis — neither factor may be zero. The author multiplies two things together: the defence line of the core business, and change in the external environment. If either is zero, the result is zero however large the other is. [Interpretation] Neither box is zero here: the defence line rests on one qualification that is real but not yet converted into disclosed price or profit evidence, and the external change (rising demand for custom AI chips) is present but has not yet been traced to this company through named customers.

[Interpretation] What would weaken the defence line. If TSMC designates a second partner in Korea, or if it becomes established practice for large Korean customers to contract directly with Taiwanese partners.

7. Financials — The Company's Present Condition in Numbers

7-1. Summary financial statements

Item	2023	2024	2025	H1 2026
Revenue	741.5	940.6	728.5	1,137.3
Gross profit	144.7	10.5	−4.1	182.8
Gross margin	19.5%	1.1%	−0.6%	16.1%
SG&A	106.0	180.2	271.6	132.9
Operating profit	+38.6	−169.7	−275.8	+49.9
Shareholders' share (net profit attributable to owners)	+36.4	−140.1	−265.6	+63.5
Operating cash flow	−105.9	+24.0	−63.0	−141.4
Purchases of PP&E and intangibles	336.2	155.5	369.6	40.2
Free cash flow (FCF)	−442.1	−131.5	−432.6	−181.6

First-half 2026 revenue was 273.3% higher than the same half a year earlier (304.6). By quarter, the picture is as follows.

Quarter	Revenue	Gross margin	SG&A	Operating P/L
Q3 2025	161.8	10.4%	61.5	−44.8
Q4 2025	262.2	−3.0%	91.6	−99.4
Q1 2026	539.9	6.8%	66.9	−29.9
Q2 2026	597.4	24.4%	66.0	+79.8

What stands out in this table is the gross margin. Across five quarters it moved from −3.0% in Q4 2025 to 24.4% in Q2 2026. SG&A stayed at roughly 6.6 billion won, almost unchanged, and yet quarterly operating profit flipped from −29.9 to +79.8. In other words, what turned this company's quarterly result around was not cost cutting but a single line: gross margin.

One more point. In fourth quarters, costs have also spiked. SG&A was 1.54 times the Q1–Q3 average in 2024 and 1.53 times in 2025, and in Q4 2025 the gross margin also fell to −3.0%. Both years pointed the same way, so the author reads fourth-quarter numbers with this pattern in mind.

(Sources: annual report, 2026-03-19, 20260319001064; semi-annual report, 2026-08-14, 20260814000520. Quarterly figures are derived by subtraction from the cumulative figures in the same filings.)

7-2. History of losses and the nature of those losses

2024 and 2025 were two consecutive years of operating losses. The 2025 operating loss widened by 10.61 billion won versus the prior year, of which the increase in SG&A was 9.14 billion won, and within that the increase in R&D expense was 5.61 billion won (61%). Over the same period the company bought land for its own headquarters for 29.14 billion won (3,230.80 square metres in Geumto-dong, Sujeong-gu, Seongnam, Gyeonggi; registered 27 December 2025), set up and capitalised a Taiwanese subsidiary, and raised 45.0 billion won through a convertible bond in March 2026.

To be fair, the facts pointing the other way. In 2024 and 2025 gross profit was 10.5 and −4.1 respectively — effectively zero. When gross profit is zero, the whole of SG&A drops straight into the loss. On that fact alone, the cause of the losses is not spending for the future but cost eating all of revenue. The author could not separate these two readings by any mechanical criterion, and which way one reads it changes the judgement on this company substantially.

R&D is worth recording alongside. R&D expense net of government grants was 0.76% of revenue in 2024, 8.75% in 2025 and 2.52% in the first half of 2026. It rose sharply in 2025, then fell as a share once revenue grew in the first half of 2026.

(Sources: annual report, 2026-03-19, 20260319001064; semi-annual report, 2026-08-14, 20260814000520. Land purchase registered 2025-12-27; convertible bond, material fact report 2026-03-13, 20260313000791.)

7-3. Cash runway, annual cash burn and its composition

As of 30 June 2026, cash and cash equivalents were 437.0 and short-term financial instruments 50.0, for a combined 487.0.

In the first half of 2026, 141.4 flowed out of operations and 40.2 was spent on property and intangible assets — 181.6 for the half, or 90.8 per quarter. If that pace continues, cash on hand covers roughly 5.4 quarters.

Why cash went out. First-half 2026 operating profit was 49.9, yet operating cash flow was −141.4. The largest reason is that money received from customers in advance (contract liabilities) fell by 590.8, from 950.1 to 359.3. That money came in as cash last year; this year it was only booked as revenue. Moreover, the −141.4 figure comes after trade payables were increased by 100.7, from 15.2 to 115.9 — that is, after paying suppliers later. The gap was filled by 448.1 of inflows from financing activities (the 45.0 billion won convertible bond).

Two offsetting points should be read alongside. One is that both operating profit and the shareholders' share had already turned positive in the first half of 2026; the other is that the 45.0 billion won financing was completed in March 2026.

On borrowings: short-term debt 55.0, current portion of long-term debt 70.0, long-term debt 340.0 and the carrying value of the convertible bond 265.6, totalling 730.6. Deducting cash equivalents and short-term financial instruments gives net debt of 243.6. Lease liabilities of 11.7 are not included here.

The capital structure is worth recording too. The debt-to-equity ratio rose from 98.24% at end-2024 to 299.71% at end-2025. At 30 June 2026, current assets were 1,202.5 and current liabilities 793.9, a current ratio of 151%. Total equity was 903.7, which includes an accumulated deficit of −297.6 and 235.3 separated into equity from the convertible bond.

Cumulative free cash flow over the three years 2023–2025 was −1,006.2, and over the same three years there was not a single dividend or share buyback or cancellation.

(Sources: annual report, 2026-03-19, 20260319001064; semi-annual report, 2026-08-14, 20260814000520. Quarterly figures are derived by subtraction from the cumulative figures in the same filings.)

7-4. Financing history and potential dilution

On 23 March 2026, a convertible bond (CB — debt that can be converted into shares) of 45.0 billion won was paid in. The coupon is 0.0%, maturity is 23 March 2031, and conversion may be requested from 23 March 2027. The stated use of the 45.0 billion won was 7.0 billion won for working capital and 38.0 billion won to acquire equity stakes in other companies — that is, 84% of the proceeds was allocated to buying equity outside the company.

The size of the dilution — the reduction in existing shareholders' share as the share count rises — is as follows.

Item	Shares
Shares outstanding (2026-06-30)	10,925,658
If the convertible bond converts	1,427,800
Unexercised stock options	227,960
Total	12,581,418

The dilution rate is 15.2%. There are no preferred shares, bonds with warrants or exchangeable bonds.

(Sources: material fact report, 2026-03-13, 20260313000791 / semi-annual report note 24)

7-5. Revenue forecast (the author's own estimates)

Year	Base scenario	Bull scenario	Bear scenario
2024 (actual)	940.6	—	—
2025 (actual)	728.5	—	—
2026	1,750	—	—
2027	1,821	—	—
2028	2,060	2,576	878

These figures are the author's base-scenario estimates and may differ from actual outcomes. If the scenario changes, these values change with it.

The 2024 and 2025 rows are disclosed actuals; the rows below are estimates constructed by the author. The bull and bear scenarios are drawn only for the final year. How the three paths diverge is discussed in section 9.

One addition. The author's base scenario puts 2028 revenue at 2,060 (the author's own estimate). If the gross margin is held at the first-half 2026 actual of 16.1% (a cost ratio of 83.9%), the break-even point — where money earned from selling equals money spent — is revenue of 2,149. In other words, in the base scenario the sign of profit does not change at any point over the three years.

8. Verification Axes — How Far Has the Evidence Come?

8-1. The qualification that opens the door to the fab

Confirmed fact — the company obtained official TSMC partner status in 2019 and is the only holder in Korea, one of seven worldwide. It also holds ARM design partner status.

What it means for judgement — without this qualification, Korean design houses would struggle to use TSMC leading-edge processes. But the qualification belongs to TSMC, and the company itself has written that its scope may expand worldwide.

Illustration — the company states that after obtaining the status it became able to order wafers directly, turning from a firm that dispatched engineers into one that takes the whole job from design through back-end processing [company statement (disclosure)].

What comes next — if cases emerge of Taiwanese partners contracting directly with Korean customers, the space this qualification provides can narrow.

8-2. The first observation of the shift into mass production

Confirmed fact — product (mass-production) revenue went from 3.1 in Q2 2025 to 165.6 in Q2 2026, and its share of that quarter's revenue, at 27.7%, is the highest in the observable period.

What it means for judgement — design services must be re-won once a project ends, whereas mass production repeats within the contract term. This is the first number for the transition the market has waited a long time to see.

Illustration — the Company A mass-production supply contract of 257.1 runs to 31 December 2026 and was increased from 254.3 on 18 August 2026. Press reports describe it as a storage-controller mass-production deal signed in January 2026, structured so that a year's volume is fixed in advance [press, Korea Economic Daily, 2026-01-08, https://www.hankyung.com/article/202601089202i]. Having volume set in advance is what makes it different from design services.

What comes next — if quarterly product revenue exceeds 20% of sales for two consecutive quarters, this transition stops being a single-quarter observation.

Also worth recording — of the 211.9 of product revenue in the first half of 2026, only 63.8 can be traced to disclosed contracts. The author could not trace through the filings where the remaining roughly 148 came from. Revenue at the Taiwanese subsidiary in the same period was 218.7 (zero a year earlier), so a substantial part may have come from there, but matching the items one by one was not possible.

8-3. The first volume to cross the border

Confirmed fact — overseas revenue of 312.8 (27.5% of the total) was booked in the first half of 2026, against zero in the same half a year earlier. The number of customers accounting for more than 10% of revenue also rose from two to three.

What it means for judgement — it is a trace that the qualification has led to actual volume, and not only from domestic customers.

Illustration — customers above 10% of revenue rose from two to three, and the newly arrived one is the starting point of overseas revenue. The Taiwanese subsidiary's first-half 2026 revenue of 218.7 (the same revenue noted in 8-2) appeared from zero a year earlier, and its half-year net loss over the same period was 14.2.

What comes next — if overseas revenue continues into the next half, the company stops being one that leans on a single domestic channel.

8-4. Increases and terminations arriving in the same period

Confirmed fact — in July and August 2026 there were three disclosures of contract increases (totalling +105.0), and in the same two months three disclosures of termination or shortening. The end date of the FADU contract of 299.8 was pulled forward by 11 months, from 11 July 2027 to 20 August 2026, and the Company F contract of 141.7 ended on 31 July 2026.

What it means for judgement — contracts coming in and contracts going out are moving at the same time, so the direction cannot be read from any single contract announcement.

Illustration — one SK hynix contract was cut from 398.4 to 197.4 on 26 June 2026 and its end date brought forward (stated reason for correction: contract amount and end date corrected). The order-backlog table in the semi-annual report still carries the uncorrected 398.4, so two different figures appear within the same report. Of the disclosed order backlog of 618.2, 388.3 comes from this discrepancy.

What comes next — if increases and terminations keep arriving in pairs, it will remain hard to gauge results from any one disclosure.

(Sources: 20260626901013, 20260819900179, 20260820900104)

8-5. Is the remaining work being refilled?

Confirmed fact — the development contract backlog was 1,614.1 at end-June 2025, 1,312.7 at end-December 2025 and 776.2 at end-June 2026. It has fallen to less than half in a year.

What it means for judgement — this backlog is work already contracted but not yet booked as revenue. Design service revenue comes out of it, so if this box is not refilled, revenue eventually converges on the level of newly arriving orders.

How it shrank — the speed at which backlog converts into revenue suddenly accelerated. On a half-year basis it was 20.9% in H1 2025 and 23.4% in H2 2025, then 70.5% in H1 2026. Hence revenue grew and backlog fell.

Placing new orders received alongside amounts booked as revenue makes the relationship clearer.

Period	New orders received	Booked as revenue	Backlog
2024	1,135.6	884.1	Increased
2025	711.9	674.4	Broadly flat
H1 2026	555.1	925.4	Fell by 536.5

Design service revenue ultimately converges on the level of new orders received. Drawing down the backlog quickly makes that year's revenue larger, but after that only what newly arrives remains.

Illustration — new projects received in the first half of 2026 came to 555.1, the largest of the four observable half-years. Yet in the same half, 166.3 was cut from existing projects, so the backlog was not refilled. The company discloses only the newly added portion of the increase and does not record the details of the reductions.

What comes next — if the adjustment to existing projects turns positive in Q4 2026, it means new orders really are refilling the backlog.

9. Three Futures — Why the Upside and Downside Are Not Symmetrical

All three are drawn for 2028.

[Interpretation] Bull scenario — revenue 2,576. A world in which new projects keep arriving at the Q2 2026 pace (406.5 per quarter), the product (mass-production) share rises to 37%, and the cost ratio falls to 80.0%. On the author's estimates, operating profit in this world is 13.5 billion won and the shareholders' share 13.7 billion won — the sign flips. If this happens, a company that manufactured revenue by consuming its backlog becomes one that builds revenue on repeating production volume.

[Interpretation] Base scenario — revenue 2,060. A world in which product (mass-production) revenue climbs in steps to 70.0 billion won in 2027 and 95.0 billion won in 2028, while the gross margin stays tied to the first-half 2026 actual of 16.1%. If this happens, revenue grows to nearly three times the level of three years earlier, but falls 4% short of the break-even point (2,149), so the sign of profit does not change.

[Interpretation] Bear scenario — revenue 878. A world in which new order intake stays at the lowest observed level, the Company A mass-production contract is not renewed after it ends in December 2026, and the cost ratio returns to 88.0% (89.6% in Q3 2025 and 103.0% in Q4 2025 are actuals). If this happens, the half-year profit of 2026 proves to be a one-off and the company reverts to the size of a design services house.

[Interpretation] Why the upside and downside differ in width. Going up requires three premises that are not yet confirmed: that new order intake continues at the Q2 2026 pace, that the rising product share actually translates into margin, and that SG&A growth is contained. The road down is an extension of things that have already happened: contract reductions occurred twice (676.5 in total), and quarters with a cost ratio of 103.0% have actually existed.

How the author's judgement could be wrong

The author sets the pillar he leaned on and the opposing pillar side by side.

The author held the gross margin at the first-half 2026 actual and assumed it does not move for three years. Yet taking Q2 2026 alone, the cost ratio was 75.6% — lower even than the lowest of the past five years of actual cost ratios that the author referenced (80.5% in 2023). The product share is rising, and the author did not put the resulting margin improvement into the numbers.

If product contracts are renewed repeatedly and the cost ratio settles in the low 70s, things the author judged not yet to be facts and therefore excluded become facts, and the picture the author has drawn would have to be redrawn entirely. In that case another world appears outside the upper edge of the three paths above.

That said, the basis for such a judgement is, for now, a single quarter of observation. Whether it continues is not yet known.

(Inputs: disclosed actuals from the annual report, 2026-03-19, 20260319001064 and the semi-annual report, 2026-08-14, 20260814000520. The 2026–2028 revenue, margin and profit figures in all three scenarios are the author's own estimates and are not disclosed figures.)

10. What to Watch — Which Signals, and When to Check Them

#	Variable	Author's base assumption	If wrong / if right, where it goes	★ Signal and when it can be checked
1	Product (mass-production) revenue	70.0 billion won in 2027, 95.0 billion won in 2028	If wrong, the revenue path reverts to 916 in 2027 and 878 in 2028	★ Quarterly product share above 20% for two consecutive quarters — Q3 2026 report
2	Gross margin	16.1% for all three years	At 24.4% profit stands; at 12.0% (cost ratio 88.0%) the loss widens — the sign diverges	★ Quarterly gross margin level — Q3 2026 report
3	Profitable quarters	Loss in all three estimated years	If profit is confirmed, calculations that start from earnings become possible for the first time	★ Positive shareholders' share for two consecutive quarters — Q3 2026 report, then FY2026 annual report
4	Adjustment to existing projects	The pattern of reductions continues	If it turns positive, the judgement on growth changes	★ Sign of the Q4 2026 adjustment — FY2026 annual report, note 27
5	Company A mass-production contract	No renewal after it ends on 31 December 2026	If renewed, the basis for production revenue is upgraded to an actual contract	★ Presence or absence of a renewal disclosure — ad hoc disclosure
6	Government R&D projects	SG&A increases by 2.5 billion won in 2027	If new projects are won, the increase is smaller	★ Project selection disclosure — FY2026 annual report
7	Five-year cash history	Listed for less than five years, so only three years exist	Once five years accumulate, the relationship between profit and cash can be measured properly for the first time	★ FY2027 annual report
8	Regional convention around partner status	Sole domestic channel is maintained	If broken, the assessment of the defence line is downgraded	★ One case of a Taiwanese partner contracting directly with a Korean customer — ad hoc
9	Convertible bond	Already included in the dilution count	Once conversion requests begin, the free float increases	★ Start of conversion requests — 23 March 2027

A possibility that is not yet a fact — watched only as a signal. If TSMC adds fabs in the United States and Japan and regional capacity disperses, the very notion of a "regional sales territory" attached to partner status could blur. In that case the position of a Korean partner shifts from being a domestic channel to competing on price in a global market. This is not the author's invention: it follows from a sentence the company wrote itself (that the sales territory may expand worldwide), and the cheapest and fastest signal to check is number 8 above.

Three signals registered separately.

Signal	Where to see it	Upside / downside	Frequency
Quarterly product (mass-production) revenue share	Note 27 of the quarterly and semi-annual reports on the electronic disclosure system	Above 20% for two consecutive quarters / below 10%	Quarterly
Quarterly gross margin	Consolidated statement of comprehensive income on the electronic disclosure system	20% or above / below 10%	Quarterly
Adjustment to the existing portion of the development contract backlog	Note 27 on the electronic disclosure system	Turns positive / a reduction of more than 20.0 billion won	Half-yearly

If one had to choose a single issue for this company — it is whether mass-production contracts are renewed repeatedly. That is item 1 in the table above; the other eight follow behind it.

11. Strengths, Weaknesses, Opportunities and Threats — One Page

	Box	Content
Inside	Strengths (already verified)	Sole holder of official TSMC partner status in Korea; Q2 2026 operating profit of 79.8, the only first-half operating profit among the four peers; no fab of its own, hence no burden of capacity expansion
Inside	Weaknesses (not yet verified, and cash)	Top three customers are 65.0% of revenue and all are unnamed; unit prices are undisclosed for every period, so pricing power cannot be confirmed; operating cash flow of −141.4 in a half-year with an operating profit; cumulative three-year free cash flow of −1,006.2
Outside	Opportunities	Rising demand for custom AI chips; TSMC 2025 capital expenditure of TWD 1,272.6 billion, an all-time high; the contract conversion rate rose from KRW 1,379.30 to KRW 1,544.20
Outside	Threats	GUC and Alchip of Taiwan hold the same qualification; migration to the Samsung Foundry camp; 201.30 of the 310.92 in government R&D grants (64.7%) ends in December 2026

In one sentence — this company's strength is its position, its weakness is what is left over from that position, and both the opportunities and the threats lie outside, where the company does not decide.

(Sources: annual report, 2026-03-19, 20260319001064; semi-annual report, 2026-08-14, 20260814000520; peer results as cited in section 5-1; contract disclosures as cited in section 3-2.)

12. Conclusion — The Market's Received View and What the Author Confirmed

[Interpretation] The market sees this company through one qualification: the only official TSMC partner in Korea. It carried large expectations into its listing on that basis and has been named early whenever the foundry ecosystem is discussed in the AI era. But results lagged those expectations for a long stretch, and the shift from development services into mass-production revenue has been identified as the trigger for results to turn.

[Fact] What the author confirmed is that this trigger has actually been pulled for the first time. The product (mass-production) share of Q2 2026 revenue, at 27.7%, was the highest in the observable period, and in the same half ASICLAND was the only one of the four listed Korean peers with an operating profit. But what the author also confirmed is that what determines what is left over at this company is not whether it shifts into mass production but a single line, the gross margin — and that this line moved from −3.0% to 24.4% over the last five quarters. Why it moves that way is not explained by the filings, because unit prices are not disclosed.

Fact assessment

What	Where it stands now
Direction of the business	Revenue is growing (H1 2026 at 1,137.3 against 304.6 a year earlier). The number of customers has also risen from two to three. But the work remaining has fallen from 1,614.1 to 776.2. All three are disclosed actuals.
Position on verification	One quarter of the shift into mass production has been observed, and it has not yet been fixed by a disclosed mass-production supply contract. The next step is whether the quarterly product share exceeds 20% for two consecutive quarters; the checkpoint is the Q3 2026 report (November).

What would change this picture

Direction	Checkable signal	Timing
Up	Quarterly product (mass-production) share above 20% and quarterly gross margin of 20% or more, together, for two consecutive quarters	Q3 2026 report (November)
Down	No renewal disclosure after the Company A mass-production contract (257.1) ends on 31 December 2026, or the adjustment to existing projects turns negative again in Q4 2026	Ad hoc disclosure; FY2026 annual report

A closing sentence — this company holds the corridor into the fab, and work has begun coming through that corridor; but how much is left over from that work is something the company does not yet decide.

(Sources: annual report, 2026-03-19, 20260319001064; semi-annual report, 2026-08-14, 20260814000520; contract disclosures as cited in section 3-2. The closing sentence is the author's interpretation.)

13. Limitations — Please Read These Alongside

The author's analysis went through a separate review gate, and seven items were left unresolved there. All of them are recorded here.

1. The author held the margin fixed at the first-half 2026 actual and assumed no movement for three years. It is recorded alongside that a better quarter did in fact exist (a cost ratio of 75.6% in Q2 2026).

2. The later part of the revenue forecast rests on premises not yet confirmed as facts. For one of those three premises — that new order intake shifts to the maximum observed level from 2027 — there is no fact that can be attached today.

3. In the base scenario the sign of profit does not change at any point over the three estimated years. The author therefore did not perform, in this report, any calculation that starts from earnings.

4. The author could not cross-check with a second yardstick, because the reference table has no row corresponding to the industry this company belongs to.

5. Six items remain unconfirmed, for differing reasons. Three could not be seen because the company does not disclose them — the size of the design services market, the market shares of the company and its competitors, and the content of revenue from other applications. Two could not be seen because the author could not gain access — twelve months of the company's own news page and materials from the supervising ministry, both of which blocked automated retrieval. The remaining one, whether a credit rating has been assigned, could not be confirmed by search. (The status regarding sanctions and export controls was checked afterwards and recorded in the body as not applicable.)

6. The author could not separate, by any mechanical criterion, whether this company's losses should be read as spending during a growth phase or as losses arising from the core business. Which way one reads it changes the author's judgement substantially.

7. The author did not forecast future cash flows, and therefore could not cross-check on a cash basis.

Three further points are added.

● The author adjusted parts of the standard drafting method to fit this company. Step increases in mass-production volume that have not yet been fixed by contract disclosure were included in base-scenario revenue. This does not affect the conclusion.

● Because the company has been listed for less than five years, the verification that measures the relationship between profit and cash over five years cannot yet be performed. It becomes possible for the first time with the FY2027 annual report.

● The author did not manufacture numbers to fill empty boxes. Boxes that are empty have been left empty.

Appendix — Key Numbers on One Page

Item	Value	Basis
Shares outstanding	10,925,658 shares	2026-06-30
Fully diluted shares	12,581,418 shares	Dilution rate 15.2%
Convertible bond	KRW 45.0 billion	Paid in 2026-03-23; conversion from 2027-03-23
Unexercised stock options	227,960 shares	2026-06-30
Development contract backlog	776.2	2026-06-30
Disclosed order backlog	618.2	2026-06-30; of which 388.3 does not reflect the correction of 2026-06-26
Revenue	741.5 / 940.6 / 728.5	2023 / 2024 / 2025
Revenue	1,137.3	H1 2026
Shareholders' share (net profit attributable to owners)	+36.4 / −140.1 / −265.6	2023 / 2024 / 2025
Shareholders' share (net profit attributable to owners)	+63.5	H1 2026
Cash and equivalents + short-term financial instruments	487.0	2026-06-30
Net debt	243.6	2026-06-30
Cumulative three-year free cash flow	−1,006.2	2023–2025
Dividends and buybacks over three years	None	2023–2025
Debt-to-equity ratio	98.24% → 299.71%	End-2024 → end-2025
Contract liabilities (money received in advance)	950.1 → 359.3	End-2025 → 2026-06-30
Government grants on R&D projects in progress	310.92 (201.30 ending December 2026)	2026-06-30
Headcount	260	2026-06-30
Revenue forecast (author's own estimates)	1,750 / 1,821 / 2,060	2026 / 2027 / 2028, base scenario

Sources: annual report (2026-03-19, 20260319001064), semi-annual report (2026-08-14, 20260814000520) and the individual disclosures whose receipt numbers are given in the body.

This report is one individual's analysis based on publicly disclosed material. It does not recommend the purchase or sale of any security and is not investment advice.

The author is not a licensed financial investment business operator or investment adviser and does not provide individual consultations.

This report does not present a fair value or a target price.

All data and figures in the body were prepared on the basis of sources believed to be reliable, such as the electronic disclosure system, but their accuracy and completeness cannot be guaranteed. Estimates in the body follow the author's assumptions and may differ from actual outcomes; all figures are as of the reference dates stated in the body.

Investment decisions and their consequences belong to the investor, and loss of principal may occur.

This report cannot be used as evidence in determining legal liability.

Preparation cut-off: disclosures and results after 25 August 2026 are not reflected.

This report was written for the public purpose of investors' right to know and the sharing of information. There is no intent to damage the reputation of any company or to interfere with its business. Facts and the author's interpretation are separated in the body, and the parts constituting interpretation are the author's own view.

The author analysed the company by first examining the competitiveness of its core business and the direction of its results, rather than starting from the market's short-term assessment.

The author does not hold the security discussed in this report (as of 27 August 2026). After the date of publication, transactions in this security may occur without prior notice.

ASICLAND (445090) _ Korea's only channel into TSMC