SEMIFIVE (KOSDAQ: 490470) _ Demand did arrive, and revenue did grow. What has not arrived is profit.
Date of writing: 25 August 2026 · Author: the author | Source documents: Annual Report (filed 2026-03-23) · Semi-Annual Report (filed 2026-08-14) | Amounts are in KRW bn (billion won) on a consolidated basis unless stated otherwise. Korean filings state amounts in units of KRW 100mn; those figures have been divided by 10 here.
How claims are separated — statements the company made are written as "the company states / the company disclosed"; the author's own judgements are written as "this reads as / the author takes the view that".
Language — this English version is a translation of the Korean original. Where the two differ, the Korean version governs.
Revenue is growing and the cost-of-sales ratio fell from 84.1% in 2025 to 68.4% in the first half of 2026, but the half-year operating result is still a loss of KRW 10.93bn — profitability has not yet been reached.
Before any number is placed, the measuring stick is declared. A comparison whose yardstick is undeclared is the kind that leaves out industry characteristics.
| # | Yardstick | Why this one |
|---|---|---|
| 1 | The company's own history, not an industry average | Peer design-service firms (GUC, Alchip and Faraday in Taiwan; Gaonchips in Korea) disclose cost ratios and market share on different bases, so they could not be compared on identical terms. Every comparison in this document is therefore SEMIFIVE against its own past. |
| 2 | Consecutive occurrence, not a single quarter | Design service revenue is recognised over the progress of the work, while IP (intellectual property, i.e. licensable circuit design blocks) fees are recognised in a lump at contract signing. The mix makes quarterly results swing for reasons unrelated to structure. A single profitable or loss-making quarter is therefore not read on its own. |
| 3 | Absolute amounts, not shares of the mix | The production revenue share falls automatically whenever IP revenue rises. Production is therefore measured in won, not in percent of mix. |
| 4 | Only post-listing data exists | The company listed in December 2025, so half-year splits have been disclosed for one year only. The small sample is a constraint of the data, not a failing of the company. |
| 5 | No legal judgement is made | This document does not assess whether any regulation was breached or what sanction might follow. It stops at reproducing what the company itself wrote in its filings. |
The 2026 revenue target of roughly KRW 269.0bn and the plan to reach profitability, both presented during the listing process, were confirmed only through press reports; the outlet and publication date are not in the author's records. The size of the gap between "company target" and "author's estimate" may therefore change. Check date: 16 November 2026 (Q3 report)
Cash stood at KRW 101.82bn as at 30 June 2026, and on the lower path that cash is drawn down. This does not mean there is no downside risk — it means this document could not put a number on its size. Check dates: 16 November 2026 (cost ratio) · March 2027 (half-year IP fee revenue)
Applying the first-half 2026 revenue mix (design services 57.2% · production 8.5% · IP 34.4%) to half-year revenue of KRW 94.68bn implies IP revenue of about KRW 32.57bn, whereas the figure stated in the text is KRW 30.58bn — a difference of KRW 1.99bn. There may be a separate "other" revenue line, or one of the percentages may reflect rounding or a different cut-off. Until the primary document is checked, this document leaves both figures standing and does not pick one over the other. Check date: segment revenue note in the Q3 report
① What the company sells. It sells custom chips — design through to volume production, bundled — to companies that want an AI (artificial intelligence) chip but lack the engineering headcount to design a whole one. Its real competition is the Taiwanese firms doing the same work, and customers deciding to build in-house design teams instead.
② Where it stands now. First-half 2026 revenue was KRW 94.68bn (78.3% of the KRW 120.95bn recorded for the whole of 2025), with an operating loss of KRW 10.93bn. Design work is growing; a profitable half-year at the operating line has not yet been recorded.
③ Two things worth watching. First, the cost-of-sales ratio moved 68.8% (2023) → 70.3% (2024) → 84.1% (2025) → 68.4% (H1 2026). It has not moved in one direction: 2025 was the outlier, and the first half of 2026 returned to the 2023 level. Second, the shift from design into volume production has arrived in orders (KRW 42.3bn) but not yet, to the same degree, in revenue (KRW 8.00bn).
④ What this document establishes. Revenue grew. By segment, the profitable one is IP (+KRW 9.53bn); the semiconductor segment posted −KRW 20.46bn. The next checkpoint is the Q3 report in November 2026.
SEMIFIVE designs custom system semiconductors — ASICs (application-specific integrated circuits) — on behalf of customers. Given a specification, it designs the chip, places production with a foundry (a contract chip-manufacturing plant), manages back-end processing, and hands over finished silicon. It sells no chips under its own name.
| Item | Detail |
|---|---|
| Founded / listed | Listed on KOSDAQ in December 2025 (KRW 130.6bn raised in the listing process) |
| Head office / CEO | Bundang-gu, Seongnam, Gyeonggi Province · CEO Cho Myung-hyun |
| Business segments | ① Semiconductor design and manufacturing ② IP design and licensing |
| Overseas presence | Subsidiaries in the US, India, Vietnam and Japan; a representative office in China (all five consolidated subsidiaries are unlisted) |
| Largest shareholder | SiFive, Inc. (US) 14.69% (20.35% including related parties) |
| Headcount | 55 in the research centre (14 with master's degrees) |
| Financial profile | Zero borrowings · no dividend · accumulated deficit KRW 615.23bn |
Two structural features stand out. The largest shareholder is SiFive, a US design-IP company, holding 14.69% (20.35% with related parties). And the one segment that posted a positive operating result in the first half of 2026 was IP, which houses the acquired US subsidiary Analog Bits.
Designing a chip from scratch normally occupies hundreds of engineers for years. The company states that it "does not start from scratch each time".
- Subsystem library — common building blocks (power management, signal I/O, memory interfaces) are designed in advance and stored, so only the customer's core logic is added on top, the company states.
- Design automation tools — software performs part of the placement and routing work previously done by hand, the company states.
- Analog Bits design IP — low-power analogue and mixed-signal design IP held by the acquired US subsidiary. The company states it has been supplied into TSMC's 2nm-class process as well as Samsung, Intel and Rapidus processes.
On top of this sit Samsung Foundry Design Solution Partner and GlobalFoundries Platinum Partner status. The share of revenue from leading-edge 2–4nm processes rose from 5% in 2022 to 41.4% on a cumulative basis through Q3 2025, the company disclosed.
| Stage | Elapsed time | What comes in |
|---|---|---|
| ① Design order → development | 1–1.5 years | Design service fees (recognised over the period of work) |
| ② Design complete → production transfer | — | Exclusive mask (the plate used to print chips) production rights secured |
| ③ Volume supply | Product life 4–7 years | Chip supply payments (recognised on each shipment) |
(Source: company filings)
Recurring revenue only appears once stage ③ is reached; a company that stays at stage ① must win fresh work every time a project ends.
| 2023 | 2024 | 2025 | H1 2026 | |
|---|---|---|---|---|
| Annual (half-year) order intake | 87.2 | 123.8 | 168.4 | 118.9 |
| Closing order backlog | — | — | 109.4 | c. 140.0 |
(KRW bn · Source: order status sections of the Annual Report (filed 2026-03-23) and Semi-Annual Report (filed 2026-08-14))
- Orders rose in three consecutive fiscal years, 2023 through 2025. (The 2022 figure was not obtained for this document, so whether the run extends further back is unconfirmed.)
- New production orders in the first half of 2026 were KRW 42.3bn, roughly twice the KRW 21.2bn of production orders taken in the whole of 2025.
- In June 2026 the company announced an order of around KRW 11bn for volume supply of an AI chip for high-performance computing in Japan, with phased shipments from the second half.
| Contract | Amount | Period |
|---|---|---|
| AI NPU (neural processing unit) dedicated chip | KRW 18.23bn | Nov 2025 – Dec 2027 |
| AI semiconductor dedicated chip | KRW 14.73bn | Jan 2026 – Aug 2027 |
| Fully homomorphic encryption accelerator (Niobium Microsystems, US) | KRW 9.90bn | Jan 2026 – Aug 2027 |
(Source: single sales/supply contract disclosures)
Customer names are withheld in most cases as trade secrets. The three that are public are Niobium Microsystems (US), ICY Tech (China) and FotoNation (Europe). (Receipt numbers for these three disclosures were not confirmed and are listed as missing — see section 16.)
The encouraging side and the not-yet-arrived side are set out separately, at equal weight.
| 2023 | 2024 | 2025 | H1 2026 | |
|---|---|---|---|---|
| Revenue | 71.31 | 111.81 | 120.95 | 94.68 |
| Cost-of-sales ratio | 68.8% | 70.3% | 84.1% | 68.4% |
- Estimated total costs on contracts in progress rose by KRW 15.58bn in 2025 and KRW 10.06bn in the first half of 2026 — a cumulative KRW 25.64bn over 18 months. No half-year within that window showed a decline. (The KRW 0.40bn movement recorded in 2024 covers a different measurement period and is excluded from this cumulative figure.)
- On a parent-only basis, gross profit in Q2 2026 was −KRW 4.28bn, and the consolidated cost ratio for Q2 was 79.42%.
This reads two ways
| Read as improvement | Read as too early to say |
|---|---|
| The consolidated cost ratio fell 15.7 percentage points, from 84.1% in 2025 to 68.4% in H1 2026, returning to the 2023 level of 68.8% | Within that same half, the Q2 consolidated cost ratio was 79.42%, higher than Q1. A low half-year average may reflect favourable recognition concentrated in Q1 |
| The company states that reuse of design assets produces economies of scale | Estimated total costs on contracts in progress rose by a cumulative KRW 25.64bn over 18 months, and that is not a forecast but a movement that already occurred on contracts already signed |
Where this splits next — if the Q3 consolidated cost ratio is confirmed below 70%, that adds one piece of evidence that the first-half improvement is a continuing trend rather than a one-quarter recognition effect.
Established facts — new production orders of KRW 42.3bn in the first half of 2026, against production revenue of KRW 8.00bn in the same period. Orders are 5.3 times revenue.
- Design projects turn into revenue 1–1.5 years after the order, so a substantial part of this gap can be explained as timing.
- Production's share of quarterly revenue fell from 19.0% in Q2 2025 to 9.6% in Q2 2026. Per the yardstick above, however, that share falls whenever IP revenue is strong.
Where this splits next — if quarterly production revenue exceeds KRW 10bn twice in a row, there is evidence that the character of revenue is shifting from project work toward repeat shipment.
Established facts — Q1 2026 consolidated operating profit of KRW 0.11bn (KRW 105mn) and net profit of KRW 1.34bn. In the following quarter the consolidated cost ratio rose to 79.42%.
Where this splits next — two consecutive quarters of operating profit would provide grounds to treat profitability as a state rather than an accident.
Established facts — first-half 2026 segment operating results were semiconductors −KRW 20.46bn and IP +KRW 9.53bn; the parent-only operating loss was KRW 21.38bn and the consolidated operating loss KRW 10.93bn.
- Goodwill recognised on the Analog Bits acquisition is KRW 81.26bn, equal to 36.5% of total equity (KRW 222.76bn).
- This document does not judge whether that goodwill is impaired. Impairment is an accounting determination.
Where this splits next — parent-only gross profit turning positive would reduce the degree to which consolidated profit rests on the IP segment.
Established facts — 60.3% of 2025 revenue arrived in the second half (39.7% in the first). 2025 is the only year for which the split has been disclosed.
The author's estimate assumes that this single year's first-half/second-half split repeats. With a sample of one, neither a mean nor a deviation can be calculated. This is a place where the range, not the direction, is exposed. Where this splits next — if Q3 2026 revenue shows the same split, the sample becomes two.
- Unexercised stock options 2,822,726 shares = 8.30% of 34,025,440 shares outstanding (of which 2,032,076 are currently exercisable)
- Treasury shares nil → options can only be settled by issuing new shares
- Convertible bonds, bonds with warrants and exchangeable bonds: all nil
- No dividend (no cash or stock dividend in the 5th, 6th or 7th fiscal years); accumulated deficit KRW 615.23bn
- Lock-up releases scheduled: 2,739 thousand shares on 2026-09-29 and 3,130 thousand shares on 2026-12-29 (further releases scheduled in December 2027 and 2028)
To be recorded alongside — a substantial part of the KRW 615.23bn accumulated deficit arose from the accounting remeasurement of convertible preferred shares issued before listing (2024 finance costs of roughly KRW 269.0bn, with no cash outflow). Under Korean commercial law, however, the deficit must be resolved before dividends can begin.
For this company AI is both the force that expands its market and a force that may substitute for its tools. Both directions are set out at equal weight.
| Axis | Expanding direction | Contracting direction |
|---|---|---|
| Demand and pricing | Rising demand for AI-specific chips brings design work in. First-half 2026 revenue was KRW 94.68bn | Pricing terms are withheld from disclosure by agreement with counterparties and cannot be verified |
| Cost and productivity | The company states that design automation means payroll does not rise proportionally as project count rises | Total payroll rose from KRW 63.38bn in 2024 to KRW 74.79bn in 2025 (+18.0%), against revenue growth of +8.2% over the same period |
| Substitution and moat | Official partner status with Samsung and GlobalFoundries | AI-based design tools may widen the scope of automation (not an established fact to date — treated as a signal to watch) |
Signals to watch — a design-tool company (Cadence, Synopsys and the like) announcing automated coverage of the full custom-chip design flow; or, in the opposite direction, SEMIFIVE's cost ratio staying below 70%.
| Revenue line | What it is | H1 2026 share | When recognised |
|---|---|---|---|
| Design services | Turnkey chip design | 57.2% | Spread over the period the work progresses |
| Production products | Supply of chips printed under exclusive mask rights | 8.5% | In a lump at the point of shipment |
| IP fees | Licensing of Analog Bits design IP | 34.4% | In a lump at contract signing (94%) |
(Note: this mix and the IP revenue figure in the text (KRW 30.58bn) do not reconcile exactly — see section 0-2, item 3.)
IP fees are booked all at once in the quarter a contract closes, so quarterly variation is large. IP fee revenue fell 17.2%, from KRW 33.29bn in 2024 to KRW 27.57bn in 2025, then reached KRW 30.58bn in the first half of 2026 — 10.9% above the full-year 2025 figure. The half-year and full-year readings point in opposite directions. For that reason the author did not simply double the first-half 2026 IP result.
| Market | Starting point | 2029 forecast | CAGR |
|---|---|---|---|
| System semiconductors, total | USD 528.2bn (2024) | USD 971.9bn | 9.0% |
| AI semiconductors | USD 53.7bn (2023) | USD 302.3bn | 29.4% |
(Infiniti Research data as cited by the company · reference date 23 March 2026; the original publication title and date were not confirmed)
- In-house chip programmes at the largest technology companies are spreading to mid-sized firms. The pattern begun by Google, Amazon, Apple, Tesla and Meta has moved down to mid-sized fabless companies and set makers.
- Japan and the US produced actual revenue. Japan revenue went from nil in 2024 to KRW 21.68bn in 2025 and KRW 16.12bn in the first half of 2026; US revenue went from KRW 8.74bn in 2024 to KRW 26.94bn in 2025.
- US export controls on advanced semiconductors bound for China. Within the range the author checked, the US Bureau of Industry and Security withdrew a design-tool licensing requirement in July 2025, and no primary source was found showing the controls subsequently extended to design services. This company's China revenue was KRW 7.20bn in the first half of 2026 (7.6% of the total).
- A regulator review process. In its semi-annual report the company stated that certain past new-share issues and secondary sales may fall within the scope of "public offering" under the Financial Investment Services and Capital Markets Act, that it self-reported the failure to file a securities registration statement to the Financial Supervisory Service, and that a review process has been under way since it submitted a written explanation in January 2025. In the same report the company stated that "the type and level of any sanction cannot be predicted." This document does not judge whether any law was breached or whether a sanction will be imposed.
The company described the structure of this market as close to perfect competition. Foundry capacity is concentrated at TSMC and design IP at Arm, Cadence and Synopsys, while the design-services layer in between is split among many firms.
| Competitor | Strength | Relative position | Market share |
|---|---|---|---|
| GUC (Taiwan) | Top-tier TSMC partner; many leading-edge track records | Relative market position unverified directly due to undisclosed market shares. | Not disclosed |
| Alchip (Taiwan) | Many large-die and HPC volume production track records | Relative market position unverified directly due to undisclosed market shares. | Not disclosed |
| Faraday (Taiwan) | Named as a comparable at the time of listing | Relative market position unverified directly due to undisclosed market shares. | Not disclosed |
| Gaonchips (Korea) | Fellow Samsung partner; strength in layout | Relative market position unverified directly due to undisclosed market shares. | Not disclosed |
None of the four discloses market share. "Where does this company rank?" is therefore a question this document cannot answer. What can be established is direction — the export share on a parent-only basis rose from 1.4% in 2024 to 29.5% in 2025. Evidence that this expansion has translated into margin improvement has not yet been established.
The two largest sources of competitive pressure are not competitors: ① customers bringing design in-house, and ② AI-based design automation.
| Defence | Verifiable from filings | What is established |
|---|---|---|
| Technology and design IP | Established | Items the company disclosed: subsystem library, design automation tools, Analog Bits design IP |
| Economies of scale | Judgement suspended | Revenue +69.6% from 2023 to 2025 while the cost ratio moved 68.8% → 84.1%; 68.4% in H1 2026 |
| Customer lock-in (switching costs) | Not established | See below |
| Barriers to entry / certification | Partly established | Samsung partner and GlobalFoundries partner status; not exclusive (multiple partners exist) |
| Network effects | Not established | — |
| Brand | Company statement only | The company's statement that it is "first in system semiconductor revenue among Samsung partners" — not cross-checked against filings |
In the four-year major-customer table of the annual report, the top customer of 2022 (30.45%) and the top customer of 2023 (25.48%) each stood at 0% in 2025.
The fact pointing the other way is recorded too. On the consolidated financial statement notes, top-customer concentration fell from 25.2% in 2025 to 24.7% in the first half of 2026. This reads as customers rotating while new ones also come in.
Summary — of six defences, one is verified from filings (technology and design IP), one is partly verified (partner status), one has judgement suspended (economies of scale), two are not established, and one rests on a company statement alone. It is not zero — partner status is a real threshold, and the IP segment did produce operating profit.
The path by which the moat weakens — if AI-based design tools come to offer "IP integration plus block assembly" as a standard feature, this company's design platform ceases to function as a differentiator.
| Item | 2023 | 2024 | 2025 | H1 2026 |
|---|---|---|---|---|
| Revenue | 71.31 | 111.81 | 120.95 | 94.68 |
| Cost of sales | 49.07 | 78.60 | 101.76 | 64.78 |
| Cost-of-sales ratio | 68.8% | 70.3% | 84.1% | 68.4% |
| SG&A expenses | 54.16 | 56.14 | 72.59 | 40.83 |
| Operating profit (loss) | −31.91 | −22.93 | −53.39 | −10.93 |
| Net profit attributable to owners | −86.18 | −290.90 | −56.22 | −10.89 |
| Total equity | −254.28 | 139.89 | 220.52 | 222.76 |
| Total assets | 193.30 | 209.21 | 309.28 | not recorded |
| Borrowings | not recorded | not recorded | not recorded | 0 (fully repaid) |
(Source: Annual Report (filed 2026-03-23) · Semi-Annual Report (filed 2026-08-14). Year-by-year cash balances and H1 2026 total assets are not in the author's records and could not be included.)
Separating the nature of the losses — the KRW 290.90bn loss in 2024 did not arise from operations. It came from a non-cash remeasurement of convertible preferred shares issued before listing (2024 finance costs of roughly KRW 269.0bn) and was resolved on listing. The negative total equity of KRW 254.28bn in 2023 stems from the same accounting treatment. The losses actually generated by operations are the operating line: −KRW 31.91bn (2023), −KRW 22.93bn (2024), −KRW 53.39bn (2025), −KRW 10.93bn (H1 2026).
| Item | Value (as at 30 June 2026) |
|---|---|
| Cash and cash equivalents | KRW 31.73bn |
| Short-term financial instruments | KRW 70.09bn |
| Total cash on hand | KRW 101.82bn |
| Borrowings and bonds | Nil |
| Lease liabilities (stated separately) | KRW 9.32bn |
| Runway | c. 5.4 quarters |
Runway is calculated from cash actually consumed in the first half of 2026 (−KRW 37.75bn for the half). Measured over a longer window (the most recent twelve months), the figure comes to about 10.2 quarters. The author recorded the shorter of the two.
Three things soften this number: ① KRW 130.6bn was raised in the December 2025 listing and borrowings were repaid in full; ② Q1 2026 operating profit of KRW 0.11bn; ③ order backlog moved from KRW 109.4bn to about KRW 140.0bn.
| Item | 2024 | 2025 | Change |
|---|---|---|---|
| Total employee benefits (cost of sales + SG&A) | 63.38 | 74.79 | +18.0% |
| Outsourcing costs (foundry and back-end) | 31.68 | 56.56 | +78.5% |
| R&D expensed in the period | 9.98 | 19.86 | +99.0% |
(KRW bn · Source: cost-by-nature note, Annual Report (filed 2026-03-23))
The two largest blocks are payroll and outsourcing, because what this company sells is engineering time plus production placed outside. R&D roughly doubled in a year and was expensed in full (no development costs capitalised).
| Revenue | Operating profit | Net profit to owners | Cost ratio applied | |
|---|---|---|---|---|
| Upper path | 412.34 | +42.41 | +42.50 | 68.42% (H1 2026 actual) |
| Middle path | 345.70 | +1.69 | +1.78 | 74.11% (last twelve months actual) |
| Lower path | 289.00 | −41.94 | −41.85 | 84.13% (2025 actual) |
(KRW bn · all three paths apply the same SG&A of KRW 87.81bn)
All three cost-ratio assumptions are values that were actually observed — they are not invented, but the cost ratios the company actually recorded in three periods. The upper path is simply the case where the first-half 2026 cost ratio holds.
- Upper path — second-half orders hold at first-half levels, production revenue attaches, and the cost ratio holds at the first-half 2026 level of 68.4%
- Middle path — the last-twelve-months cost ratio of 74.11% persists. Even with revenue at 2.86 times the 2025 level, the operating margin is 0.5%
- Lower path — IP revenue reverts to 2025 levels, a large customer departs, and the cost ratio returns to the 84% area
Upside room is +KRW 40.72bn and downside room −KRW 43.63bn — similar in size, with the downside slightly larger.
| To reach the upper path | To reach the lower path | |
|---|---|---|
| What must happen | ① Second-half orders hold at first-half levels ② Production revenue rises in absolute terms for two consecutive quarters ③ Cost ratio holds in the 68% area | ① IP revenue reverts to 2025 levels ② A large customer departs ③ Cost ratio returns to 84% |
| Where things stand | ③ is already met on a half-year basis (68.4%) but was 79.42% in Q2 alone. ② moved the other way on a share basis (9.6%) | All three have each occurred once in the past |
| Basis for "profitability arrives late" | Nature | Basis for "profitability arrives" | Nature |
|---|---|---|---|
| Estimated total costs up a cumulative KRW 25.64bn over 18 months | Established fact | New production orders of KRW 42.3bn | Established fact |
| Q2 consolidated cost ratio 79.42% | Established fact | H1 2026 consolidated cost ratio 68.4% | Established fact |
| Parent-only Q2 gross profit −KRW 4.28bn | Established fact | Q1 operating profit KRW 0.11bn | Established fact |
| Two large customers at 0% in 2025 | Established fact | Japan production order of around KRW 11bn | Company announcement (supply still to come) |
| Operating margin of 0.5% on the middle path | Author's estimate | Repeat shipment over 4–7 years on exclusive mask rights | Company statement |
| — | — | The company's plan to reach profitability in 2026 | Company statement (primary document not checked) |
Where this review is least certain — no allowance at all was made in the middle path for the production transfer. Production orders of KRW 42.3bn and the Japanese order of around KRW 11bn exist, but with no disclosure to pin down when they are recognised as revenue, they were set to zero. If that piece switches on, the middle path is rebuilt upward.
The variable that moves the outcome most is gross margin. On the middle path, revenue at 2.86 times still leaves an operating margin of 0.5%, so what decides the result is cost, not revenue.
| # | Variable | Assumption used here | If wrong, where it goes | Checkpoint |
|---|---|---|---|---|
| 1 | Cost-of-sales ratio | Last-twelve-months actual of 74.11%, held flat | At 80%, still an operating loss in 2028 | Q3 consolidated cost ratio below 70% / above 80% — November 2026 |
| 2 | First-half/second-half revenue split | 2025 actual (39.7% in H1) | If it was chance, the 2026 estimate is overstated | Q3 revenue — November 2026 |
| 3 | Production transfer | Increment of zero (already inside backlog) | If it switches on, the middle path is rebuilt upward | Quarterly production revenue above KRW 10bn for two consecutive quarters — November 2026 |
| 4 | Durability of profitability | First full-year profit in 2028 | Two consecutive quarters makes profitability a state | Two consecutive quarters of operating profit — November 2026 |
| 5 | Timing of IP revenue recognition | Three-year average of −1.3% | Below KRW 30bn per half, losses widen again | Half-year IP fee revenue below KRW 30bn — March 2027 |
| 6 | Lock-up releases | Share-count change only | Released shares change the free float, not shares outstanding | 2026-09-29 / 2026-12-29 |
| 7 | Regulator review process | No related cost included | The company states the type and level of any sanction cannot be predicted | Related disclosures — as filed |
| 8 | US controls on China | Not extended to design services | If extended, KRW 7.20bn of China revenue is affected | US Commerce final rules — Federal Register, as issued |
| 9 | Substitution by AI design tools (not an established fact — signal only) | Not included | Design platform differentiation narrows | A design-tool company announcing full-flow coverage |
| Signal | Where to find it | Upper / lower | Frequency |
|---|---|---|---|
| Order backlog | Order status, annual and semi-annual reports | Above KRW 170bn / below KRW 110bn | Half-yearly |
| Change in estimated total costs (cumulative) | Progress notes, annual and semi-annual reports | Increase halts / four consecutive half-years of increase | Half-yearly |
| New production order intake | Annual and semi-annual reports plus company announcements | Above KRW 40bn per half / below KRW 15bn | Half-yearly |
There are grounds to treat all three as leading indicators: design projects become revenue 1–1.5 years after the order; the change in estimated total costs flagged the Q2 2026 cost-ratio rise in advance; and new production orders currently stand at 5.3 times production revenue.
| Strengths — verified from filings | Weaknesses — not yet established |
|---|---|
| Cash on hand KRW 101.82bn, zero borrowings | Estimated total costs on contracts in progress up a cumulative KRW 25.64bn over 18 months |
| IP segment operating profit +KRW 9.53bn (H1 2026) | Parent-only operating loss KRW 21.38bn; accumulated deficit KRW 615.23bn |
| Official partner status with Samsung and GlobalFoundries | Segment profit concentrated in IP alone, with related goodwill at 36.5% of equity |
| Orders up in three consecutive fiscal years (87.2 → 123.8 → 168.4) | Unexercised options 8.30%; no dividend |
| Cost ratio 84.1% (2025) → 68.4% (H1 2026) | No half-year operating profit established |
| Opportunities — from outside | Threats — from outside |
|---|---|
| Repeat shipment over 4–7 years after production transfer, per the company's statement | AI design tools may widen the scope of automation (not an established fact — a signal) |
| Demand shifting toward AI-specific chips | US controls on China may extend to design services |
| Actual revenue in Japan (KRW 21.68bn) and the US (KRW 26.94bn) in 2025 | Lock-up releases increase the free float (September and December) |
| New production orders of KRW 42.3bn (H1 2026) | The company states the outcome of the regulator review process is undetermined |
One-sentence summary — while cash of KRW 101.82bn and IP segment profit hold the company up, it is at the stage of establishing whether the parent business can generate operating profit on its own.
Market consensus hypothesis (no specific source is named): the bottleneck in the shift of demand toward AI-specific semiconductors is design; SEMIFIVE stands at that bottleneck as a Samsung Foundry partner; production revenue attaches in 2026 and the company turns profitable.
What this document establishes — demand did arrive and revenue did grow. The cost ratio also fell, from 84.1% in 2025 to 68.4% in the first half of 2026. What has not yet been established is a profitable half-year at the operating line and the arrival of production revenue. Even accepting the actual growth rate and the actual seasonal split in full, so that revenue reaches 2.86 times the 2025 level in three years (KRW 120.95bn → KRW 345.70bn), a persisting last-twelve-months cost ratio of 74.11% leaves 2028 net profit to owners at KRW 1.78bn.
The difference between the two views comes from timing, not revenue — the company presented profitability in 2026 (primary document not checked); the middle path in this document places it in 2028. Those two years matter more than the −12.8% difference in the revenue forecast. (Note: the author's 2026 revenue estimate, which underlies that −12.8%, is not presented in the body of this document — see section 16.)
| Item | Summary |
|---|---|
| Direction of profit | The operating loss narrowed from KRW 53.39bn in 2025 to KRW 10.93bn in the first half of 2026. That direction is nonetheless tied to two numbers: a seasonal split with a one-year sample, and the cost ratio |
| Where verification stands | One profitable quarter occurred and did not continue; the production transfer is verified through orders but not yet in revenue. The segment with positive operating profit remains IP |
| Next checkpoint | Q3 report, November 2026 (prior date: lock-up release on 2026-09-29) |
- Upper path confirmation — quarterly production revenue above KRW 10bn twice in a row, together with two consecutive quarters of operating profit (check date: Q3 report, November 2026 · one profitable quarter recorded so far)
- Lower path confirmation — the Q3 consolidated cost ratio returning to 80% or above, or half-year IP fee revenue falling below KRW 30bn (check dates: November 2026 · March 2027)
Closing — first-half 2026 revenue of KRW 94.68bn equals 78.3% of full-year 2025 revenue, and the operating result for the same period was −KRW 10.93bn. No case has yet been established in which revenue growth carried through to a profitable half-year at the operating line.
- Company target not checked against the primary document — the 2026 revenue target of KRW 269.0bn was confirmed only through reporting on the listing filings; the outlet and date are not in the author's records
- 2026 revenue estimate not presented — verifying the −12.8% in the body requires the author's 2026 revenue estimate, but this document presents only the 2028 figures
- Both cross-checks failed to run — ① the check that tests the authenticity of profit through cash flow cannot compute a ratio, because profit was negative in all of the last three years; ② the check against an industry benchmark table could not be run, because no such table exists
- The author's standard revenue method was not applied — this company does not disclose unit volumes, so the author's usual "volume × price" method could not be used; the estimates were built instead from actual segment growth rates and the actual seasonal split
| What | How |
|---|---|
| Unit volumes undisclosed for all segments | Revenue could not be built from volume × price and was estimated from value ratios |
| Only three years of history | The listing was in December 2025, so there is no five-year record; there are only three half-year observations |
| Seasonal sample of one year | 2025 is the only year with a disclosed half-year revenue split |
| Market size cannot be verified | With no statistics for the size of, or share within, the design-services layer, there was no way to test whether the estimates are large relative to the market |
| Two Taiwanese competitors not researched | GUC and Alchip are listed in Taiwan and fell outside the scope of this work |
| Sensitivity of the 2028 profit estimate | The +KRW 1.78bn on the middle path changes sign if revenue moves 5%. The apparent precision is an illusion |
| SG&A assumption | Holding SG&A at KRW 87.81bn assumes the ratio to revenue falls from 60.0% (2025) to 25.4% (2028) |
| No record of how this review began | There is no record of why this company became a subject of review, so it cannot be tested whether the starting point pulled the conclusion |
For the items below, primary source identifiers — receipt numbers, document numbers, publication dates — are not in the author's records. Until they are re-checked, this document records them as missing.
| # | Item | What is missing |
|---|---|---|
| 1 | Annual Report (filed 2026-03-23) | Receipt number |
| 2 | Semi-Annual Report (filed 2026-08-14) | Receipt number |
| 3 | Three single sales/supply contract disclosures (KRW 18.23bn · 14.73bn · 9.90bn) | Receipt numbers and disclosure dates |
| 4 | 2026 revenue target of KRW 269.0bn | Outlet name and publication date |
| 5 | Japan AI chip production order of around KRW 11bn | Disclosure/announcement document identifier |
| 6 | Infiniti Research market size data | Original publication title and date (only the company's citation was seen) |
| 7 | CEO name and head office address | Source receipt number |
| 8 | July 2025 action by the US Bureau of Industry and Security | Federal Register document number |
| 9 | Company statement on the regulator review process | Location of the item within the semi-annual report |
| 10 | Parent-only Q2 gross profit −KRW 4.28bn · Q2 consolidated cost ratio 79.42% | Source location for the quarterly figures |
| 11 | Top-customer concentration of 25.2% / 24.7% | Note number |
| 12 | Last-twelve-months cost ratio 74.11% · first-half share 39.7% | Source location behind the calculation |
| 13 | US revenue KRW 8.74bn → KRW 26.94bn | Year labels (read as 2024 → 2025, primary document not checked) |
What this document answered and what it did not — it answered "how this company's business is structured, and what stage its profitability has reached." It does not answer questions about corporate value, share price, or whether any level is appropriate.
(Source: Annual Report (filed 2026-03-23) · Semi-Annual Report (filed 2026-08-14))
| Item | Value |
|---|---|
| Shares outstanding | 34,025,440 |
| Unexercised stock options | 2,822,726 (8.30%) |
| Largest shareholder | SiFive, Inc. 14.69% (20.35% incl. related parties) |
| 2025 revenue / operating profit / net profit to owners | KRW 120.95bn / −53.39bn / −56.22bn |
| H1 2026 revenue / operating profit / net profit to owners | KRW 94.68bn / −10.93bn / −10.89bn |
| Cost-of-sales ratio (2023 → H1 2026) | 68.8% → 70.3% → 84.1% → 68.4% |
| Cash on hand (2026-06-30) | KRW 101.82bn · zero borrowings |
| Runway | c. 5.4 quarters |
| Order backlog (2026-06-30) | c. KRW 140.0bn |
| H1 2026 order intake | KRW 118.9bn (of which production KRW 42.3bn) |
| Segment operating result (H1 2026) | Semiconductors −KRW 20.46bn / IP +KRW 9.53bn |
| Goodwill | KRW 81.26bn (36.5% of total equity) |
| Dividend | None (accumulated deficit KRW 615.23bn) |
| Next checkpoint | Q3 report, November 2026 (prior date: lock-up release 2026-09-29) |
This is an individual's business review based on publicly available regulatory filings. It does not recommend buying or selling any security and does not constitute investment advice. The author is not a licensed financial investment business operator or investment adviser, and provides no individual consultations.
All data and figures in the text were compiled from public sources including electronic regulatory disclosures, but their accuracy and completeness cannot be guaranteed. Items whose source identifiers could not be confirmed are set out openly in section 16. Estimates reflect the author's assumptions and may differ from actual outcomes; all figures are as at the reference dates stated in the text.
This document contains no content on corporate value, share price, or whether any level is appropriate, and makes no judgement as to whether the company breached any law or will face any sanction. Investment decisions and their consequences belong to the investor; loss of principal is possible; and this material may not be used as evidence in any determination of legal liability.
Filings and results after the date of writing are not reflected.
The author does not hold shares in this company.
After this report is published, the author may buy or sell without prior notice.
The analysis in this document is an opinion offered for the public purpose of investors' right to know and the sharing of information, with no intent to damage the reputation of, or interfere with the business of, any company.
This English version is a translation of the Korean original. Where the two differ, the Korean version governs. The two sentences in bold above correspond to the Korean originals: "필자는 동 주식을 보유하고 있지 않습니다." and "보고서 발행 이후에는 예고 없이 매수 또는 매도할 수 있습니다."
댓글
댓글 쓰기