Gaonchips (KOSDAQ: 399720) _ The order backlog is trending up, but more top-line growth is needed to cover fixed costs.
Gaonchips (KOSDAQ: 399720) — Business Review
Prepared 25 August 2026 · Revised 27 August 2026
Units All figures are in KRW billion (bn) unless otherwise noted. KRW 1bn = 10억 원, which lets you cross-check every number against the Korean original.
※ The disclosure notice sits at the very bottom of this document. Please read it first.
What Has Not Been Confirmed — Stated Up Front
Three items were not confirmed as of the preparation date. The reason, the effect on this review, and when each can be resolved are set out below.
① What is unconfirmed
- The timing of the operating-profit turn. This review covers only 2026–2028, so it cannot settle whether the turn occurs inside that window.
- A cross-check against third-party industry forecasts.
- Competitors' revenue mix and business model (candidate ⓓ in Section 8-2).
② Why it is unconfirmed
- On the author's estimate, costs that do not move with revenue run to roughly KRW 43.6bn a year, and the revenue needed to cover them is about KRW 104.4bn. The author's 2028 revenue estimate is KRW 98.9bn — 5.6% short. Within a three-year window, both outcomes remain open.
- The intermediate working papers carried a summary rather than the original text.
- Of the five peer companies, only one discloses revenue broken out by type.
③ Effect on the conclusion — direction and size
| # | Direction | Size |
|---|---|---|
| 1 | [VIEW] Works only in the direction of holding the author's estimate at the lower end | Once operating profit clears zero, profit-based measures such as operating margin and interest coverage become calculable. A single quarter above zero does not by itself confirm that the earnings structure is durable; separate items remain to be checked |
| 2 | [VIEW] No effect on the financial estimate. Third-party forecasts are not inputs to the calculation; they appear only as narrative context | The premises of the estimate are unchanged |
| 3 | [VIEW] Prevents Section 8-2 from narrowing the cause of the share-of-peer-revenue decline to a single explanation | Whether the cause sits outside or inside the company determines how recovery should be judged |
④ When it can be confirmed
| # | What resolves it | Date |
|---|---|---|
| 1 | Whether Q3 2026 operating profit is above zero, and whether the three-year estimate turns positive | 16 November 2026 (Q3 report filing deadline) |
| 2 | Obtaining forecast comparison material in the next review | After 16 November 2026 |
| 3 | Whether peers disclose revenue by type | March 2027 annual reports |
0. Thirty-Second Summary
① How the company makes money. [FACT] Gaonchips takes on design, fabrication, packaging, and test as a single bundle on behalf of fabless chip companies, then delivers the finished chip under its own name and is paid for the contract. [BASIS] The 40.6% figure below annualises the five Korean peers' H1 2026 combined revenue by doubling it and compares that with full-year 2025; differences in the five companies' business mix are not adjusted for. [FACT] On that basis, combined peer revenue rose 40.6%. [VIEW] The competitive set looks closer to the four domestic peers and to Samsung Electronics, which grants the certification, than to large overseas players.
② Direction of results. [VIEW] In the author's base estimate, revenue grows on the back of new orders but does not cover the costs that do not move with revenue, so operating losses continue through 2028.
③ Where the judgement sits. [FACT] Confirmed order backlog rose from KRW 61.1bn at end-2025 to KRW 76.4bn at end-June 2026, up 25.0%. Q2 2026 operating profit was KRW 0.33bn, above zero. The revenue needed to cover non-revenue-linked costs is about KRW 104.4bn on the author's estimate.
④ Summary. [FACT] H1 2026 revenue KRW 41.9bn (+72.1% year on year), operating loss KRW 3.6bn, net cash KRW 30.2bn at end-June 2026. [VIEW] The gap between the author's 2028 revenue estimate of KRW 98.9bn and the KRW 104.4bn threshold is 5.6%. That 5.6% assumes the low end of the disclosed outsourcing-cost range (57.6%); at the high end (65.9%) the threshold rises to KRW 127.9bn and the gap widens to 29.3%.
1. What the Company Is
[FACT] Gaonchips is a certified design partner of Samsung Foundry (contract semiconductor manufacturing). It performs design, fabrication, packaging, and test as one bundle on behalf of companies that only design chips. Founded 6 August 2012; listed on KOSDAQ 20 May 2022. The chief executive is Jung Kyu-dong, who is also the largest shareholder.
| Item | Detail |
|---|---|
| People | [FACT] 230 of 267 employees (86.1%) are in R&D or technical roles |
| Facilities | [FACT] No in-house production facilities. Fabrication, packaging, and test are fully outsourced |
| Subsidiaries | [FACT] Japan (2022), United States (2024), China (February 2026) — all wholly owned |
| Largest shareholder | [FACT] Jung Kyu-dong 28.18%. Combined with related parties, 44.71% as of end-July 2026, down 2.16pp from 46.87% |
| Dividends | [FACT] None in the last three fiscal years. No treasury share purchases or cancellations |
[COMPANY-STATED] The articles of incorporation list nine business purposes, including manufacture, design, and layout of semiconductor integrated circuits, system-semiconductor design services, and design-IP development.
2. The Core Capability — What the Company Does On Others' Behalf
[FACT] Not every company that wants a chip owns a fab. Fabless companies contract manufacture out to Samsung Electronics, TSMC, or similar. At advanced nodes, a design must be redrawn to the specific fab's rules from an early stage, which creates a need for a specialist that bridges the design house and the fab. That role is called a Design Solution Partner (DSP), and this is what Gaonchips does.
[VIEW] The author's read is that customers use the company because producing a chip on a node such as Samsung's 2nm requires design know-how specific to that node, and building it in-house costs a fabless company significant time, headcount, and EDA software spend.
First, certification. [FACT] The company has been a certified Samsung Foundry design partner since 2019 and holds official design-partner status with Arm of the United Kingdom, which supplies the underlying design framework. It has a track record in functional-safety compliance for automotive chips. [COMPANY-STATED] The company says it has received Arm's "Partner of the Year" award several times; the author could not confirm this in a regulatory filing (Unsourced Item 1).
Second, people. [FACT] 86.1% of headcount is technical. The patent count is small.
[VIEW] The author treats both as real barriers to entry. However, the two certifications are granted by Samsung Electronics and Arm, so the granting authority sits outside the company. If Samsung adds partners, the scarcity of that status falls with it. Sections 7 and 8 return to this.
3. Contracts in Progress and Their Timing
[FACT] Work arrives as bundles of multi-year design projects. The H1 2026 semi-annual report's "Order Status" section lists five disclosed major contracts.
[BASIS] "Total (semi-annual report)" below is stated on the semi-annual report basis, which reflects re-translation at updated exchange rates. "Original disclosed amount" is the confirmed contract value in the single sales/supply contract disclosure filed at signing. The difference between the two is mostly the exchange rate applied on different dates.
| # | Contract date | Disclosure date · rcpNo | Content · counterparty | Total (semi-annual report) | Original disclosed amount | Scheduled end | Remaining backlog |
|---|---|---|---|---|---|---|---|
| ① | 2024-02-08 | 2024-02-13 · 20240213900265 | AI ASIC (Application-Specific Integrated Circuit) design development. Counterparty: GAONCHIPS JAPAN Co., Ltd — a wholly owned subsidiary; supply region Japan | KRW 59.7bn | KRW 55.66bn (US$42.00m) | 2026-08-31 | KRW 3.3bn |
| ② | 2025-09-15 | 2025-09-16 · 20250916900085 | Custom ASIC design development. Counterparty withheld (trade-secret request) | KRW 9.8bn | KRW 9.75bn (US$7.02m) | 2027-06-30 | KRW 9.8bn |
| ③ | 2025-10-29 | 2025-10-29 · 20251029900487 | Custom semiconductor design development. Counterparty withheld | KRW 33.9bn | KRW 31.79bn | 2028-07-31 | KRW 21.7bn |
| ④ | 2025-12-24 | 2025-12-26 · 20251226900414 | Custom semiconductor design development. Counterparty withheld | KRW 17.9bn | KRW 17.88bn | 2026-10-01 | KRW 13.3bn |
| ⑤ | 2026-06-10 | 2026-06-11 · 20260611900165 | ASIC design development. Counterparty stated as "a global AI semiconductor company". Supply method: in-house production applicable / outsourced production NOT applicable | KRW 29.9bn | KRW 29.61bn (US$19.50m) | 2028-06-30 | KRW 28.3bn |
| Total | KRW 151.2bn | KRW 144.69bn | KRW 76.4bn |
On Contract ① — the structure, stated plainly. [FACT] The disclosure text reads: "This contract is one under which the company receives the order for an original contract concluded by its subsidiary GAONCHIPS JAPAN Co., Ltd with a local Japanese counterparty." In other words, the Japanese subsidiary wins the original contract from the end customer and the parent takes the work from it. [UNVERIFIED] The author could not confirm whether this contract appears in the order-status table on a consolidated or a separate-entity basis. [VIEW] It is set out here because reading this contract as "a US$42m customer the parent won directly" would misdescribe the structure.
Extension of Contract ①. [FACT] An amended disclosure filed 31 December 2025 (rcpNo 20251231900107) moved the end date from 31 December 2025 to 31 August 2026. The stated reason was "change of contract period".
On Contract ⑤. [FACT] The disclosure records the supply method as "in-house production applicable / outsourced production not applicable", unlike the other four, all of which are marked outsourced-production applicable. [VIEW] If no outsourced cost attaches, the margin retained on a given amount of revenue could be higher than on the other contracts. The actual cost composition is not disclosed and was not confirmed.
[FACT] Three further contracts, each exceeding 5% of prior-year revenue, are in progress but not separately disclosed (semi-annual report, contract-level notes).
[FACT] Three of the five contracts are denominated in US dollars. Contract ① moved from KRW 55.66bn at original disclosure to KRW 59.7bn on the semi-annual report basis, up 7.3% — the result of a different exchange-rate reference date, not additional work.
[FACT] Confirmed order backlog rose from KRW 61.1bn at end-2025 to KRW 76.4bn at end-June 2026, up 25.0%. Within that, Contract ①'s remaining KRW 3.3bn (4.3% of the total) is scheduled to end on 31 August 2026 — six days after the preparation date.
[FACT] No new single sales/supply contract has been disclosed since 11 June 2026. That is 75 days to the preparation date of 25 August.
4. Three Places Where the Judgement Sits
4-1. Orders are in fact arriving
[FACT] H1 2026 revenue KRW 41.9bn, up 72.1% year on year. Confirmed backlog KRW 76.4bn, up 25.0% in six months. Q2 2026 operating profit KRW 0.33bn, above zero. H1 2026 gross margin moved from −7.6% (FY2025) to +5.1%.
[VIEW] The author reads the current operating loss as arising less from order decline than from revenue not yet reaching the scale that covers non-revenue-linked costs.
[FACT] In 2025 revenue fell 29.0% year on year and the operating loss was KRW 16.7bn. In H1 2026 revenue rose 72.1% and the operating loss narrowed to KRW 3.6bn.
[VIEW] Whether quarterly operating profit stays above zero for consecutive quarters is the next checkpoint.
4-2. Operating profit is still below zero
[FACT] Costs fall into two blocks. One is the amount paid to outside parties (outsourced processing costs), which rises with revenue and runs at 57.6–65.9% of revenue. The other is labour-centred cost that does not move with revenue, which the author estimates at roughly KRW 43.6bn a year.
[VIEW] The revenue that covers that KRW 43.6bn works out at about KRW 104.4bn.
[FACT] The author's base estimate for 2028 revenue is KRW 98.9bn, 5.6% below the KRW 104.4bn threshold.
[VIEW] The near-term question is less the absolute size of profit than whether operating profit clears zero. While operating losses continue, profit-based measures such as operating margin and interest coverage cannot be calculated, and the short-term buffer is gauged by net cash of KRW 30.2bn.
[FACT] Q2 2026 net income was KRW 0.73bn while operating profit was KRW 0.33bn. The difference came from foreign-exchange and derivative valuation, not from operations.
[VIEW] Reaching the threshold requires either 5.6% more revenue or about a 2.3pp improvement in the cost ratio (covering KRW 43.6bn on KRW 98.9bn of revenue requires a 44.1% contribution margin against the 41.8% assumed).
4-3. Share of the five peers' combined revenue has fallen for three consecutive years
[BASIS] The comparison below simply adds the consolidated revenue reported by five companies — SemiFive, CoAsia SEMI, ADTechnology, Gaonchips, and Alphachips — in their semi-annual and annual reports. The five differ in fiscal calendar, consolidation scope, and revenue-recognition model (how much finished-chip delivery each books), and those differences are not adjusted for. H1 2026 figures are annualised by doubling; because this company's 2025 revenue split KRW 24.3bn in H1 against KRW 44.1bn in H2, annualising may understate it.
[FACT] On that basis Gaonchips' share was 22.4% in 2024, 13.8% in 2025, and 12.0% in H1 2026. On the same basis, the five companies' combined revenue rose 40.6% while Gaonchips rose 22.3%.
[BASIS] The 22.4% for 2024 is measured in the company's largest revenue year (KRW 96.5bn).
[VIEW] The premise that growth in the end market flows straight through to this company's results is not yet confirmed in the financial figures. Growth of the overall market and expansion of an individual company's share need to be looked at separately.
[FACT] Revenue from customers outside the top five fell 64.2%, from KRW 17.3bn in 2024 to KRW 6.2bn in 2025.
[BASIS] "Other" is the residual after the top five customers, so when the composition of the top five changes from year to year, the same customer can move into or out of "other". That reclassification is not adjusted for.
[VIEW] Whether the share recovers is what determines if industry growth is absorbed.
Where AI cuts both ways
| Axis | Direction that helps | Direction that pressures | Signal to watch |
|---|---|---|---|
| Demand and pricing | [FACT] A KRW 29.61bn contract signed with an overseas AI company in June 2026 (rcpNo 20260611900165) | [COMPANY-STATED / UNVERIFIED] Synopsys began a collaboration with Samsung Foundry to embed AI across the design flow (May 2026; outlet and publication date unconfirmed — Unsourced Item 2). [VIEW] As design automation advances, the price paid for design time can fall | Direction of revenue per employee — March 2027 annual report |
| Cost and productivity | [FACT] Revenue per employee KRW 257m (2025) to about KRW 314m (H1 2026 annualised). [BASIS] Annualised by doubling H1; changes in headcount composition are not adjusted for | [UNVERIFIED] The author could not confirm evidence that adoption costs or workforce restructuring land ahead of the benefit | (same) |
| Substitution and moat | [VIEW] Large, complex chip design and advanced packaging integration are performed by relatively few firms, which can also work to raise the barrier | [VIEW] As design automation advances, the value added by the partner role may change, and if the foundry handles more directly, the order environment shifts | Whether margin by revenue type is disclosed for two years — March 2027 annual report |
[VIEW] If design automation advances further, the asset in this business may become verified, production-proven design data rather than headcount. In that case the revenue-per-employee measure now in use would not capture the company's asset. This is not a confirmed fact and is treated only as a signal.
5. The Business Model — How Money Comes In and Goes Out
5-1. What is sold is not "design" but "a finished chip"
[FACT] The company completes the design, uses it to place wafer production with Samsung, has packaging and test performed, and delivers the finished chip to the customer under its own name. The industry calls this a turnkey model.
| Category | H1 2026 |
|---|---|
| Product revenue (prototype development + volume chip supply) | [FACT] 92.2% |
| Service revenue (design work only) | [FACT] 7.8% |
[VIEW] Selling design alone produces smaller revenue, most of which is value added. Selling a finished chip brings the wafer cost paid to Samsung, the packaging and test cost, and design-IP royalties in as revenue and back out as cost. Revenue is recorded large and gross margin is calculated low.
5-2. Money coming in — three streams
| Stream | What it is | H1 2026 share | Nature |
|---|---|---|---|
| Prototype development | Advanced-node design and prototype fabrication | [FACT] 67.1% (KRW 28.1bn) | Project-based |
| Product (volume) | Volume wafer and chip supply | [FACT] 25.1% (KRW 10.5bn) | Recurring |
| Service | Partial design work | [FACT] 7.8% (KRW 3.3bn) | Project-based |
[FACT] About two-thirds of revenue is project-based, so that revenue ends when the project ends.
[FACT] Payment arrives as advance, interim, and final instalments, while revenue is recognised on percentage of completion. As a result, "received but not yet revenue" and "performed but not yet received" both sit large on the balance sheet.
5-3. Money going out — two blocks
| Cost block | What it is | Size | Nature |
|---|---|---|---|
| Outsourced processing | Wafer cost paid to Samsung; packaging and test | [FACT] 57.6–65.9% of revenue | Moves with revenue |
| Everything else | Labour-centred | [VIEW] About KRW 43.6bn a year (author's estimate) | Does not move with revenue |
[FACT] The only item classified as raw material is design IP (Intellectual Property), of which KRW 10.0bn was purchased in 2025. Suppliers include Arm and Synopsys.
[VIEW] Under this structure the profit-and-loss outcome converges on one variable: revenue scale. Above roughly KRW 104.4bn (range KRW 102.8–127.9bn) operating profit clears zero; below it, operating losses continue. That is about 1.5 times FY2025 revenue of KRW 68.5bn.
5-4. How selling prices are set
[VIEW] The selling price is set by the customer, and a substantial part of the cost is set by Samsung and Arm.
[COMPANY-STATED] "Unit prices are not disclosed because price variation is wide owing to differences in specifications required by each customer" (FY2025 annual report, Section II-4).
[BASIS] The wafer price below is a weighted average of purchased items, while the cost ratio is a company-wide figure. Changes in revenue mix (prototype / volume / service) and in process node are not adjusted for.
| 2023 | 2024 | 2025 | |
|---|---|---|---|
| Average wafer purchase price (US$) | [FACT] 4,910 | 4,857 | 4,411 |
| Cost of sales ratio | [FACT] 80.0% | 85.2% | 107.6% |
[FACT] Over three years the wafer price fell 10.2% while the cost of sales ratio rose 27.6pp.
[VIEW] The rise in the cost ratio is not explained by the wafer price decline alone. However, as noted in the basis above, the figures are unadjusted for mix and node, so the cause cannot be attributed to selling-price decline alone.
5-5. Operating margin fell in the year revenue was largest
[BASIS] In a model where finished chips are delivered under the company's own name, outsourced cost enters both revenue and cost, so margins are calculated lower. The figures below are unadjusted for that effect and for changes in revenue mix.
| 2023 | 2024 | 2025 | |
|---|---|---|---|
| Revenue | [FACT] KRW 63.6bn | KRW 96.5bn | KRW 68.5bn |
| Operating margin | [FACT] 6.8% | 3.7% | −24.4% |
[VIEW] In 2024, when revenue rose 51.7%, operating margin was 3.1pp lower than the prior year. With much of the cost base being outsourcing that scales with revenue, the structure limits how far operating margin improves as revenue grows.
6. Market Environment — Why the End Market Is Growing
6-1. Three drivers
First, big tech developing its own chips. [FACT] Cloud providers and device makers are moving to design AI-specific chips fitted to their own services rather than buying general-purpose parts. [VIEW] They are not positioned to execute advanced-node design themselves, so more in-house chip programmes means more demand for partners. This is already under way.
Second, Samsung Foundry's advanced nodes returning to activity. [VIEW] This company's revenue is closely tied to the number of projects running on Samsung's advanced nodes. [COMPANY-STATED / UNVERIFIED] Samsung Electronics said at its Q2 2026 results briefing that it had won 2nm projects from major cloud and AI customers and begun product design, and that it intends to more than double the number of 2nm design wins in 2026 versus the prior year. Utilisation at 8nm and below is described as effectively at maximum, and the Taylor, Texas fab is planned to begin operating in 2026 (briefing date and document name unconfirmed — Unsourced Item 3).
Third, automotive and robotics adopting high-performance chips. [FACT / UNVERIFIED] Automotive revenue was KRW 14.6bn (2023), KRW 21.3bn (2024), KRW 30.6bn (2025), and KRW 23.1bn (H1 2026); the source citation is unconfirmed (Unsourced Item 4). [VIEW] If those figures hold, automotive would be 44.7% of company revenue in 2025 and 55.1% in H1 2026. Because the source is unconfirmed, these shares are left as reference only.
6-2. Total market size is not stated as a number
[UNVERIFIED] There is no official size or share statistic for the design-service market, so this review excludes any market-size estimate built on a cited figure.
[FACT / UNVERIFIED] A near proxy is system-semiconductor exports — US$4.26bn in January 2026, up 22.3% year on year (Ministry of Trade, Industry and Energy ICT trade statistics, January 2026; release date and document link unconfirmed — Unsourced Item 7). [VIEW] That figure combines foundry, fabless, and back-end processing, so it is defined differently from "design services"; it is used for direction only and is not an input to any calculation.
[BASIS] The figures below are the author's own extraction and simple summation of consolidated revenue reported by the five companies in their H1 2026 semi-annual reports and FY2025 annual reports on DART. They are not an official market statistic. Individual receipt numbers were not obtained (Unsourced Item 5).
[FACT] Combined revenue of the five was KRW 497.0bn in 2025 and KRW 349.4bn in H1 2026, or 70.3% of the prior full year.
6-3. The author's reading
[VIEW] End-market demand growth is confirmed. What has moved is the question. It used to be "is the market growing?"; it is now "why is this company's share of it falling while it grows?"
[VIEW] The ceiling on this market is the number of advanced-node customers Samsung secures multiplied by the portion this company is given. Results therefore track the foundry partner's advanced-node order performance, and if the partner struggles to secure large customers, both favourable drivers for this company are delayed with it.
[FACT] Three of the five disclosed contracts are dollar-denominated. [COMPANY-STATED] The company's stated sensitivity is that a 10% move in the KRW/USD rate shifts pre-tax profit by KRW 2.8bn (note location not cited — Unsourced Item 6). [VIEW] A stronger won compresses won-translated revenue and foreign-currency assets at the same time.
7. Competitive Structure
[VIEW] The structure is a narrow oligopoly. Samsung Foundry design partners require formal certification, which limits the number of participants. Entry requirements are five: (i) foundry certification, (ii) an advanced-node production track record, (iii) design-IP partnerships, (iv) automotive functional-safety experience, and (v) senior design engineers.
[BASIS] The figures below were compiled by the author from each company's H1 2026 consolidated financial statements. The five differ in consolidation scope and revenue-recognition model, and this is not adjusted for. Individual receipt numbers were not obtained (Unsourced Item 5).
| Company | Revenue | Operating profit | Net income |
|---|---|---|---|
| SemiFive | KRW 94.7bn | −KRW 10.9bn | −KRW 10.9bn |
| CoAsia SEMI | KRW 90.2bn | [UNVERIFIED] | [UNVERIFIED] |
| ADTechnology | KRW 82.0bn | −KRW 1.5bn (consolidated) | −KRW 6.1bn |
| Gaonchips | KRW 41.9bn | −KRW 3.6bn (Q2: +KRW 0.33bn) | −KRW 0.8bn |
| Alphachips | KRW 40.7bn | −KRW 1.6bn | −KRW 0.9bn |
[UNVERIFIED] CoAsia SEMI did not file a separate semi-annual report, so its operating profit and net income could not be confirmed.
[FACT] ADTechnology's operating loss was KRW 1.5bn against a net loss of KRW 6.1bn.
[FACT] All four companies that could be confirmed were in operating loss.
[FACT] On H1 2026 revenue, Gaonchips ranks fourth of the five (KRW 41.9bn).
[VIEW] This looks like a condition common to the sector — revenue short of the non-revenue-linked cost base — rather than a situation confined to one company.
Structural variables within the sector. [VIEW] Changes in the foundry partner's vendor policy and the addition of new partners could shift the competitive environment; this is an external variable affecting how solid the barrier to entry is.
Substitution paths. [VIEW] The author regards three items as more structurally significant than head-to-head competition: (i) customers building their own design teams, (ii) the foundry absorbing the partner role, and (iii) a major customer shifting production wholesale to a competing foundry. If (iii) occurs, uncertainty widens across the entire order platform on which this company is based.
8. Barriers to Entry and Drivers of the Share Change
8-1. Defensibility, split six ways
| Source of defensibility | Confirmed? | Evidence |
|---|---|---|
| Technology and patents | [FACT] Confirmed | 2nm advanced-node design execution, large-and-complex chip design methodology, advanced packaging integration. Patents are few and the core capability sits with people |
| Economies of scale | [FACT] Not confirmed | In 2024, when revenue rose 51.7%, operating margin was 3.1pp lower than the prior year (6.8% to 3.7%) |
| Network effects | [UNVERIFIED] | No material available |
| Switching costs | [VIEW] Partly confirmed | Transferring a project in progress to another firm carries switching cost. That effect does not persist once a contract ends — the nature of order-based work |
| Certification and entry barriers | [FACT] Confirmed | Samsung certified partner since 2019, Arm official partner, automotive functional-safety experience. The granting authority sits outside the company |
| Brand and reputation | [COMPANY-STATED] | Arm "Partner of the Year" awards (not confirmed in filings — Unsourced Item 1) |
[VIEW] Barriers are confirmed to a degree, but four items remain alongside them: (i) pricing authority sits with the customer; (ii) the link between revenue growth and margin improvement is not confirmed in the actuals; (iii) the granting authority for two certifications sits outside the company; and (iv) the FY2025 operating loss of KRW 16.7bn is the outcome this structure produced.
8-2. Why revenue growth lagged the peer group
[BASIS] The 40.6% and 22.3% below use the same annualised, simple-sum basis as Section 4-3.
[FACT] While combined peer revenue rose 40.6%, this company rose 22.3%.
[VIEW] The author could not narrow the reason to one explanation. There are four candidates, and only ⓐ is confirmed in filings.
| Candidate cause | Confirmed facts | What remains unconfirmed |
|---|---|---|
| ⓐ Termination of a large contract in December 2024 | [FACT] Single sales/supply contract termination disclosed 24 December 2024 (rcpNo 20241224900546). Terminated amount KRW 12.02bn, equal to 27.75% of the then-latest annual revenue of KRW 43.3bn. Counterparty an overseas customer; contract period 2023-12-20 to 2025-10-15. The stated reason in the filing is "a request from the counterparty to discontinue development". Of an original contract value of KRW 24.53bn (US$18.80m), KRW 12.51bn (US$9.59m, 50.98%) had been delivered, and the residual was terminated | Whether follow-on work replaces the terminated volume. [VIEW] How much of the KRW 28.0bn revenue decline in 2025 this termination accounts for is not confirmed in filings |
| ⓑ Shrinking revenue outside the top customers | [FACT] Revenue outside the top five customers fell 64.2% in one year (KRW 17.3bn to KRW 6.2bn). [BASIS] The definition of "other" shifts year to year | Whether new customers were not won, or smaller projects were not taken |
| ⓒ Concentration in one large project | [FACT] 75.4% of cumulative revenue from the five disclosed contracts came from Contract ① alone, which is scheduled to end 31 August 2026 | Whether follow-on work exists |
| ⓓ Competitors taking the lead in specific areas | [UNVERIFIED] | Competitors' process and technology mix could not be confirmed |
[VIEW] With ⓐ now confirmed in a filing, it becomes possible that a substantial part of the 2025 revenue decline came from a customer's decision to stop development — a factor outside the company — rather than from a change in its competitiveness. Because the correspondence between the terminated amount and the revenue decline is not confirmed in filings, this is not asserted as settled.
8-3. Selling chips under its own name — what is gained and what is paid
[VIEW] Three things gained.
- The customer deals with one party. There is no need to separately find and coordinate a design house, a fab, and a packaging vendor. The less the customer is itself a semiconductor company, the larger this convenience.
- Revenue and backlog are recorded large. Fabrication and packaging values enter revenue, making it easier to meet the delivery-track-record requirements large customers impose.
- Relationships last longer. Taking responsibility through volume production brings recurring revenue while the chip sells. [FACT] Volume supply is the only stream at this company whose margin has stayed above zero for three straight years.
[VIEW] Three things paid.
- Revenue grows while gross margin is calculated low. Because outside costs pass through, roughly 60 of every 100 won of revenue goes to Samsung and design-IP suppliers.
- The pace of covering non-revenue-linked costs is slow. This is the direct reason the threshold rose to KRW 104.4bn.
- No pricing authority develops. What the company adds in this model is design value added, and the customer sets its price.
[UNVERIFIED] Whether this model is the cause of the growth gap against peers could not be confirmed. A comparison requires knowing whether peers use the same model and whether they sell a higher proportion of design-only work; that material was not obtained. This is why ⓓ above is empty.
[FACT] That said, Contract ⑤ (KRW 29.61bn) is marked "outsourced production not applicable", so the model does not apply uniformly across all contracts.
8-4. How the principal risk would materialise
[VIEW] The main path to a contraction in the business base is the foundry partner struggling to secure large customers at advanced nodes (2nm and similar), reducing volume. In that case order volume can fall even with certification, track record, and headcount intact.
[FACT] FY2025 revenue −29.0% and an operating loss of KRW 16.7bn are the actuals from such a period.
[VIEW] The places to check are Samsung Electronics' quarterly results briefings and changes to its partner roster.
9. Financials — What Has Happened and What Is Estimated
9-1. The last three years and H1 (actuals)
| (KRW bn) | 2023 | 2024 | 2025 | H1 2026 |
|---|---|---|---|---|
| Revenue | 63.6 | 96.5 | 68.5 | 41.9 |
| Operating profit | 4.4 | 3.5 | −16.7 | −3.6 |
| Net income attributable to owners | 6.3 | 7.5 | −14.4 | −0.8 |
| Operating cash flow | 5.7 | 10.8 | −0.8 | 4.3 |
| Capex and intangibles | 10.9 | 10.8 | 8.3 | 5.1 |
| Free cash flow | −5.2 | about 0 | −9.2 | −0.8 |
| Cash and short-term financial instruments | [UNVERIFIED] | [UNVERIFIED] | [UNVERIFIED] | 59.2 |
| Borrowings and convertible bonds | [UNVERIFIED] | [UNVERIFIED] | [UNVERIFIED] | 29.0 |
| Total equity | [UNVERIFIED] | [UNVERIFIED] | [UNVERIFIED] | 71.8 |
| Debt-to-equity ratio | [UNVERIFIED] | 62.5% | 191.3% | [UNVERIFIED] |
[FACT] FY2022 revenue was KRW 43.3bn.
[UNVERIFIED] Year-by-year actuals for cash, borrowings, and equity were obtained only for end-June 2026. The 62.5% and 191.3% debt-to-equity figures and the KRW 71.8bn equity figure carry different reference dates, and their consistency was not confirmed.
9-2. The next three years — the author's base estimate
[VIEW] The table below is an estimate, not a settled figure. It assumes the current backlog is consumed on schedule and that new contracts arrive at the pace of the last three years. If the assumptions change, the figures change.
| (KRW bn) | 2025 (actual) | 2026 | 2027 | 2028 |
|---|---|---|---|---|
| Revenue | 68.5 | 87.1 | 99.2 | 98.9 |
| Year on year | −29.0% | +27.1% | +14.0% | −0.4% |
| Gross profit | −5.2 | 5.2 | 9.5 | 9.1 |
| Operating profit | −16.7 | −6.2 | −2.1 | −2.8 |
| Pre-tax profit | −19.9 | −6.9 | −3.3 | −4.0 |
| Net income attributable to owners | −14.4 | −6.9 | −3.3 | −4.0 |
| Operating cash flow | −0.8 | about 2.3 | about 4.7 | about 4.0 |
| Capex and intangibles | 8.3 | about 10.3 | about 10.3 | about 10.3 |
| Free cash flow | −9.2 | about −8.0 | about −5.6 | about −6.3 |
[BASIS] "Gross profit" above deducts the whole of cost of sales, whereas the threshold calculation in Section 4-2 uses contribution margin after outsourced processing costs only. The definitions differ, so the two ratios will not reconcile if compared directly.
[VIEW] Three things to read alongside this table.
- Revenue grows, but whether it reaches the KRW 104.4bn threshold cannot be settled; the gap to the 2028 estimate is 5.6% on the low-end premise and 29.3% on the high-end premise.
- The operating loss narrows and then widens again in the final year, because 2028 revenue is 0.4% below 2027 on the curve at which the current backlog is consumed.
- Free cash flow stays below zero even after operating cash flow clears zero, because roughly KRW 10bn a year goes to design equipment and software. [FACT] Free cash flow was −KRW 5.2bn in 2023, about zero in 2024, and −KRW 9.2bn in 2025, and is estimated below zero for 2026–2028.
9-3. The nature of the loss
[FACT] The FY2025 loss of KRW 14.4bn arose in operations, not from valuation or other book items.
- [FACT] R&D expense runs at 2.74–5.23% of revenue and is fully expensed (nil capitalised). At most KRW 3.6bn in 2025, it is not of a size that explains a KRW 14.4bn loss.
- [FACT] Capital expenditure is design equipment and software, not plant. Construction in progress is under 1% of five years of investment.
- [VIEW] The substance of the loss is outsourcing cost that scales with revenue plus non-revenue-linked cost not yet covered.
9-4. Cash and burn rate
[FACT] At end-June 2026: cash and short-term financial instruments KRW 59.2bn, borrowings and convertible bonds KRW 29.0bn, net cash KRW 30.2bn.
[VIEW] At the FY2025 free cash flow rate of −KRW 9.2bn, KRW 59.2bn represents about 25.8 quarters, or a little over six years. The base for this calculation is the KRW 59.2bn before deducting borrowings.
[VIEW] Because design IP and outsourcing costs are incurred early in a project, short-term cash-flow pressure can appear in a period when orders are rising.
9-5. Sources of potential share-count increase
| Item | Disclosed detail |
|---|---|
| Shares outstanding | [FACT] 11,858,844 (30 June 2026); treasury shares nil |
| Convertible bonds | [FACT] KRW 34.5bn issued April 2025 (0% coupon, maturity April 2030; rcpNo 20250401003248); KRW 10.35bn converted April 2026, KRW 24.15bn outstanding |
| Stock options | [FACT] 47,000 unexercised |
| Diluted share count | [FACT] 12,462,167 (+5.1% versus basic) |
[VIEW] The roughly 556,323 shares that would arise on conversion of the outstanding convertible bonds is an estimate. [FACT] The bonds carry a conversion-price adjustment clause, so the actual number issued could exceed the estimate. The dilution ratio above assumes the lower bound.
9-6. Business review summary
[VIEW] This review focuses on business continuity and the direction of operating cash flow.
- [FACT] Net cash at end-June 2026 was KRW 30.2bn, about 25.8 quarters at the FY2025 burn rate.
- [VIEW] The direction is "revenue is growing; operating profit clearing zero is not yet confirmed."
- [VIEW] The author's standard method counts only disclosed contracts, which sets future work to zero; in practice new contracts arrived in each of the last three years. This review therefore assumes that pace continues. The figure produced by the standard method has also been retained internally.
10. Scenarios and Variables
| Case | If this happens | Then the company |
|---|---|---|
| Upside | [VIEW] One or two of the advanced-node customers Samsung secures select this company as design partner, and two more contracts of around KRW 30bn arrive | Revenue passes the KRW 104.4bn threshold and operating profit clears zero |
| Base | [VIEW] The current backlog is consumed on schedule and new contracts arrive at the current pace | Revenue grows but stays below the threshold and operating losses continue |
| Downside | [VIEW] The contract gap persists and the outsourcing ratio moves to the unfavourable end (65.9%) | The threshold rises to KRW 127.9bn, the operating loss widens, and cash depletion accelerates |
10-1. End-market growth versus this company's share
[VIEW] The author's base estimate assumes the company's share stays at its current level even as the end market grows.
| 2024 | 2025 | H1 2026 | 2028 (implied by the author's estimate) | |
|---|---|---|---|---|
| Gaonchips' share of five-peer combined revenue | 22.4% | 13.8% | 12.0% | about 11–12% |
[VIEW] Material on both sides of that premise is set out below.
First, why share recovery is not assumed. Why the share fell has not been narrowed to one cause (Section 8-2), so the evidence for recovery and the evidence for continued decline are equally thin. In that situation the author does not put the better outcome into the assumption. That is not the same as saying decline is settled.
Second, material that could support recovery.
- [FACT] A KRW 29.61bn contract was signed with an overseas AI company in June 2026, and the filing records "no performance of same-type contracts with the company in the last three years" (rcpNo 20260611900165). This runs opposite to candidate ⓑ.
- [FACT] The stated reason for the December 2024 termination was "a request from the counterparty to discontinue development" — a factor outside the company (rcpNo 20241224900546).
- [FACT] One large project's end date was extended from 31 December 2025 to 31 August 2026 (rcpNo 20251231900107). Delayed work is deferred, not lost.
- [FACT] H1 2026 revenue rose 72.1%. The fall in share to 12.0% reflects faster growth at the comparison companies, not a decline at this company.
Third, material on the other side.
- [FACT] Contract ①, which produced 75.4% of cumulative revenue, is scheduled to end 31 August 2026.
- [FACT] No new supply-contract disclosure since 11 June 2026 (75 days to the preparation date).
- [FACT] The share has fallen for three consecutive years.
[VIEW] Which way it goes is decided in two places: the Q3 2026 report in November (comparison against the five peers) and new contract disclosures from September 2026.
10-2. Why share recovery is not built into the estimate
[VIEW] The upside rests on two things that have not happened — that a customer Samsung secures selects this company, and that this occurs twice. Neither is a confirmed fact.
[FACT] The downside material is the continuation of what has already occurred — no new contract disclosure since 11 June, three consecutive years of falling share against peers, and a stronger won.
[VIEW] That said, the distance between the current estimate and the breakeven case is 5.6% of revenue on the low-end premise.
10-3. Where the estimate could be wrong
| Material supporting the author's estimate | Nature | Material supporting the other side | Nature |
|---|---|---|---|
| Non-revenue-linked costs KRW 43.6bn; threshold revenue KRW 104.4bn | [VIEW] Author's estimate | Q2 2026 operating profit +KRW 0.33bn | [FACT] |
| Share of peers 22.4% to 12.0% | [FACT] (annualised, simple-sum basis) | Gross margin −7.6% to +5.1% | [FACT] |
| No new contract disclosure since 11 June | [FACT] | Volume-supply margin above zero for three straight years and improving | [FACT] |
| Free cash flow below zero in two actual years and three estimated years | [FACT / VIEW] | Advance-payment-type items up KRW 8.5bn | [FACT] |
| Three-year profit estimate below the threshold | [VIEW] | Samsung's 2nm customer wins and Taylor fab schedule | [COMPANY-STATED] |
[VIEW] The variable most likely to produce error is reading the margin decline in prototype revenue as structural. Contrary evidence remains: [FACT] one large project was extended by eight months (rcpNo 20251231900107), and [COMPANY-STATED] the company has said it will bill the resulting additional cost to the customer. [FACT] In Q2 2026 the prototype margin moved above zero.
[VIEW] If the margin decline was an accounting timing effect rather than structural, the author's cost estimate is overstated and the turn arrives earlier. The point of resolution is the March 2027 annual report, when margin by revenue type would be disclosed for two years.
11. What to Watch
| # | Variable | Author's base assumption | Which way it turns | Where it is confirmed |
|---|---|---|---|---|
| 1 | Reaching the threshold | Revenue stays below KRW 104.4bn through 2028 | Above it, profit-based measures become calculable and the author's estimate is recalculated. Below it, net cash of KRW 30.2bn continues to act as the buffer | Q3 2026 report, November |
| 2 | Sustained quarterly operating profit | The Q2 2026 +KRW 0.33bn may be a one-off | Two consecutive quarters is a signal of approach; a return to loss hardens the structural reading | Q3 2026 report, November |
| 3 | Inflow of new contracts | Continues at KRW 55.7bn a year. [UNVERIFIED] Basis not presented (Unsourced Item 9). [FACT] The three-year average on the original-disclosure basis is KRW 50.6bn | If the gap persists, the revenue premise fails and cash begins to fall | Ad hoc · DART |
| 4 | Contract ending 31 August | KRW 3.3bn of backlog remains | Termination cuts backlog by 4.3%; extension holds it | Disclosure, early September 2026 |
| 5 | Margin decline: structural or timing | Not yet resolved | Structural lowers the estimate; timing means margin recovers | March 2027 annual report |
| 6 | Scarcity of advanced-node track record | Falls if Samsung adds partners | Adding partners is the unfavourable direction for this company | Ad hoc · Samsung Foundry announcements |
[VIEW] The heaviest item is number 1. The other five are closer to signals that indicate, before or after, when that one is crossed.
Two measures to check each quarter
| Measure | Where to look | Improving / deteriorating |
|---|---|---|
| Confirmed order backlog | Order-status table in quarterly, semi-annual, and annual reports (base KRW 76.4bn) | +10% / −10% versus prior quarter |
| New supply-contract disclosures | DART (rolling 12 months) | Holds at KRW 55.7bn (three-year average KRW 50.6bn on the original-disclosure basis) / two consecutive quarters at zero |
12. Strengths, Weaknesses, Opportunities, Threats
| Strengths — already held | Weaknesses — internal burdens |
|---|---|
| [FACT] Samsung certified design partner status, Arm official partner status, advanced-node design track record | [FACT] Operating profit is still below zero, and revenue is 5.6% short of the threshold on the low-end premise |
| [FACT] Confirmed backlog KRW 76.4bn, up 25.0% in six months | [VIEW] Delivering finished chips under its own name means outside cost inflates both revenue and cost, so gross margin is calculated low |
| [VIEW] Handling design through fabrication, packaging, and test in one place gives customers that are not semiconductor companies a stronger reason to use it | [FACT] In 2024, when revenue rose 51.7%, operating margin was 3.1pp lower than the prior year |
| [FACT] 86.1% of 267 employees are technical | [FACT] Contract ①, which produced 75.4% of cumulative revenue, is scheduled to end 31 August 2026 |
| [FACT] Net cash KRW 30.2bn; cash and short-term instruments KRW 59.2bn, about 25.8 quarters at the FY2025 burn rate | [FACT] Free cash flow is below zero across the actual and estimated periods, and KRW 24.15bn of convertible bonds remains convertible |
| Opportunities — external | Threats — external |
|---|---|
| [FACT] Big tech's in-house AI chip programmes are increasing demand for design partners | [VIEW] The foundry partner's policy on adding new partners could change the scarcity of certified design-partner status |
| [COMPANY-STATED] Samsung has secured advanced-node large customers and begun design, and has said it will double 2nm wins | [VIEW] A major customer moving production wholesale to a competing foundry would weaken the order base |
| [FACT] Automotive revenue rose for three consecutive years (KRW 14.6bn to 21.3bn to 30.6bn) | [COMPANY-STATED] A 10% move in KRW/USD shifts pre-tax profit by KRW 2.8bn |
| [VIEW] As design automation advances, verified production-proven design data could become a new asset — not a confirmed fact, treated as a signal | [VIEW] Design automation could reduce the price paid for design time itself |
| [FACT] Contract ⑤ (KRW 29.61bn) carries no outsourced production | [FACT] All four peers that could be confirmed are in operating loss |
[VIEW] In summary, this is a company that must cross a 5.6% revenue gap — on the low-end premise — while certification and cash hold.
13. Limitations — Please Read Alongside
First, the author has not closed this review as settled. The three unconfirmed items are set out in "What Has Not Been Confirmed" at the top.
Second, one of the checks the author normally applies did not operate. It tests whether profit arrives as actual cash, and with the company in operating loss the test itself does not resolve. For reference, five years of free cash flow measured against net income was below zero, driven by equipment and software investment.
Third, this review carries the following limitations.
- [UNVERIFIED] The company does not disclose volumes and unit prices by product. The revenue estimate is therefore built from the rate at which contract backlog is consumed, not from volume times price.
- [UNVERIFIED] Total market size could not be confirmed, so whether the estimated revenue runs into a market ceiling was not tested.
- [UNVERIFIED] The share count on conversion of the outstanding convertible bonds is an estimate and could be higher in practice.
- [UNVERIFIED] The author intended to compare margins against peers but only one company discloses comparable figures, so the comparison was not made.
- [UNVERIFIED] CoAsia SEMI's operating profit and net income, and peers' business models, could not be confirmed.
- [UNVERIFIED] Lease liabilities and the revenue contribution of the subsidiary established in 2026 could not be confirmed. The former means net cash of KRW 30.2bn may be recorded slightly higher than the true figure.
- [FACT] The data cut-off is 25 August 2026; disclosures and market developments after that date are not reflected.
Fourth, this review adjusted two elements of the author's standard method. New contracts were not set to zero but assumed to continue at the recent pace, and projected financial statements — normally omitted for a company in operating loss — were included in Section 9-2. The standard-method figure for the former has been retained internally; the latter does not affect the concluding figures.
Fifth, this document addresses only the business base and the direction of cash flow. Net cash of KRW 30.2bn is cited solely as a measure of the short-term buffer available to absorb non-revenue-linked costs.
Appendix — Key Figures on One Page
| Item | Value | Reference date |
|---|---|---|
| Shares outstanding | 11,858,844 (treasury nil) | 2026-06-30 |
| Diluted share count | 12,462,167 (+5.1%) | 2026-06-30 |
| Net cash | KRW 30.2bn (cash and short-term instruments 59.2 − borrowings and CBs 29.0) | 2026-06-30 |
| Runway at current burn | About 25.8 quarters | FY2025 burn rate |
| Confirmed order backlog | KRW 76.4bn (+25.0% versus KRW 61.1bn at end-2025) | 2026-06-30 |
| Disclosed major contracts | 5 · KRW 151.2bn on the semi-annual report basis (KRW 144.69bn on the original-disclosure basis; three USD-denominated) | 2026-06-30 |
| FY2025 revenue / operating profit | KRW 68.5bn / −KRW 16.7bn | FY2025 |
| H1 2026 revenue / operating profit | KRW 41.9bn (+72.1%) / −KRW 3.6bn | H1 2026 |
| Q2 2026 operating profit | +KRW 0.33bn | Q2 2026 |
| Estimated revenue (2026 / 2027 / 2028) | KRW 87.1bn / 99.2bn / 98.9bn | Author's base estimate |
| Estimated net income to owners (2026 / 2027 / 2028) | −KRW 6.9bn / −3.3bn / −4.0bn | Author's base estimate |
| Non-revenue-linked cost / threshold revenue | KRW 43.6bn a year / about KRW 104.4bn (range 102.8–127.9) | Author's estimate |
| Share of five-peer combined revenue | 22.4% to 13.8% to 12.0% | 2024 / 2025 / H1 2026 (annualised, simple sum) |
| Largest shareholder and related parties | 44.71% (previously 46.87%) | 2026-07-31 |
| Dividends | None (three consecutive fiscal years) | — |
| In-house production facilities | None (fully outsourced) | — |
| Certifications | Samsung certified design partner (2019–) · Arm official design partner | — |
Unsourced Items — What Could Not Be Sourced
No source has been invented. The places where a source could not be obtained are listed as they are.
| # | Location | What is missing |
|---|---|---|
| 1 | Sections 2, 8-1 | Arm "Partner of the Year" award — outlet and publication date unconfirmed |
| 2 | Section 4 table | Synopsys–Samsung Foundry AI collaboration (May 2026) — outlet and publication date unconfirmed |
| 3 | Section 6-1 | Samsung Electronics Q2 2026 results briefing remarks on 2nm — briefing date and document name unconfirmed |
| 4 | Section 6-1 | Automotive revenue KRW 14.6 / 21.3 / 30.6 / 23.1bn — source citation unconfirmed |
| 5 | Sections 6-2, 7 | Five-peer financial figures — receipt numbers for the five semi-annual reports not obtained |
| 6 | Sections 6-3, 12 | Pre-tax profit sensitivity of KRW 2.8bn to a 10% FX move — note location not cited |
| 7 | Section 6-2 | MOTIE ICT trade statistics, January 2026 — release date and document link not cited |
| 8 | Sections 4-2, 5-3 | Internal composition of the KRW 43.6bn non-revenue-linked cost — not confirmable from filings; author's estimate |
| 9 | Section 11 | The KRW 55.7bn annual new-contract pace — basis not presented. [FACT] On the original-disclosure basis, single sales/supply contracts totalled KRW 62.86bn in 2024 (2 contracts), KRW 59.42bn in 2025 (3 contracts), and KRW 29.61bn in January–August 2026 (1 contract) |
Disclosure Notice
This is an individual business review by the author based on publicly available regulatory filings. It does not recommend the purchase or sale of any security and does not constitute investment advice. The author is not a licensed financial investment business operator or investment adviser and does not provide individual consultations.
The data and figures are drawn from electronic disclosures and comparable sources, but their accuracy and completeness are not guaranteed. Estimates reflect the author's assumptions and may differ from actual outcomes; all figures are as of the reference dates stated in the text.
This document contains no judgement of any kind on corporate value, share price, or fair value.
The preparation reference date is 25 August 2026. Disclosures and results after that date are not reflected.
Investment decisions and their outcomes rest with the investor. Loss of principal is possible. This material may not be used as evidence in determining legal liability.
· The author does not hold shares in this company.
· After publication of this review, the author may buy or sell without prior notice.
The author does not provide individual stock consultations and does not respond to security-specific enquiries received through comments or email.
Language. This is an English rendering of a document originally written in Korean. Where the two differ, the Korean original governs. Figures have been converted to KRW billion (KRW 1bn = 10억 원) and rounded; the Korean original carries the figures as filed.
Jurisdiction. The subject company is listed on KOSDAQ in the Republic of Korea. This document is written for a general readership, is not directed at any individual, and is not tailored to any recipient's circumstances. It is not an offer or solicitation in any jurisdiction.
댓글
댓글 쓰기