Isu Petasys (007660): 97% of Sales Are Exports – and H1 Operating Cash Flow Went Negative

Almost every Korean company you have heard of sells something to Koreans. This one barely does. Isu Petasys books 3.0% of its revenue in Korea and the other 97.0% in Thailand, Mexico, Taiwan and a bucket the filing calls “other countries” — and the boards it ships end up inside AI accelerators and 800G network switches. It is, on the numbers, one of the purest export businesses listed in Seoul. It also just posted its best half-year ever and watched its operating cash flow go negative.

🏢 What Is Isu Petasys, and What Does It Actually Sell?

Isu Petasys is a Korean maker of high-layer-count printed circuit boards — the thick, densely stacked boards that carry signals inside AI accelerators, network switches and servers — and it sells almost all of them outside Korea.

The product is called an MLB, or multi-layer board. It is not a semiconductor and not a package substrate; it is the large motherboard-class board those parts sit on. The company describes itself in its own filing as producing “printed circuit boards (PCB), the core component of electronic products,” and says it operates on a 100% build-to-order basis — nothing is made for stock. (Source: DART half-year report, rcpNo 20260814002283, filed Aug 14, 2026)

The basics, all as of June 30, 2026 unless noted:

Note the two dates. The company was incorporated in 1972 as a shipping business and only entered PCBs in 1989 — the two are often collapsed into a single founding year in English-language summaries. Its own filing keeps them separate.

📊 Where Does Isu Petasys’s Revenue Actually Come From?

From four places, none of which is Korea: Thailand (29.1%), Mexico (18.6%), Taiwan (14.9%) and an unnamed “other countries” bucket (34.4%) — with Korea at just 3.0% of H1 2026 consolidated revenue.

This is the geography note in the consolidated segment disclosure, and it is the single most useful table in the filing. English-language coverage of Isu Petasys almost always describes it as an “AI PCB play.” It is that. It is also, structurally, a company whose customers assemble hardware in Southeast Asia and North America.

Two things move in that table. Thailand shrank as a share — from 39.0% to 29.1% — not because it fell, but because Taiwan (+87.6%) and the “other” bucket (+78.6%) grew far faster. Read the shares alone and you would think Thailand was lost. Read the dollars and it grew 8.8%.

The mirror image is where the assets sit. Non-current assets by region at June 30, 2026 were $270.9M in Korea (89.0%), $29.7M in Asia and $3.9M in North America. Isu Petasys manufactures at home and bills abroad — an inversion worth holding on to, because it is what the rest of this article turns on.

🗓️ How Did a 1972 Shipping Company Become an AI-PCB Maker?

Through five renamings, one group acquisition and four plant builds — the company has been called Daeyang Shipping, Daeyang Trading, Namyang Precision, Isu Electronics and Petasys before settling on Isu Petasys in 2002.

🎯 Who Are Isu Petasys’s Customers?

The filing does not say — it identifies them only as “Company G,” “Company N,” “Company M” and “Company A” — but it does disclose that a single customer accounted for 40.1% of consolidated revenue in H1 2026.

This is worth being precise about, because English-language coverage routinely attaches specific American names to this company. Those names do not appear in the filing. What appears is this, verbatim in structure:

• In the switch market: the company says it has secured “Company G and Company M” as customers among the major players that hold “60% or more of the global market,” and is in discussion with “Company A” about new entry.
• In the AI accelerator market: it says it currently has “Company G and Company N” as principal customers, and that both will participate in next-generation models.

We are not going to guess which company each letter is. What is disclosable and disclosed is the concentration:

And the number-two customer changed identity between the two years. Read the separate-basis columns across periods and the second-largest slot swaps hands: Customer B went from 5.71% of parent revenue to 10.38% — crossing the 10% disclosure threshold — while Customer C went the other way, from 10.27% down to 6.77%. In dollar terms B more than doubled, from $17.8M to $45.1M (2.5×), while C slipped from $32.0M to $29.4M. The company’s disclosure of “two customers above 10%” is identical in both years and hides the fact that they are not the same two. (Source: rcpNo 20260814002283, separate-basis revenue note; percentages are our calculation against separate-basis half-year revenue of ₩584,593,160k / ₩419,309,339k.)

The direction on the largest customer is mildly reassuring: top-customer concentration fell from 42.2% to 40.1% while revenue grew 45.8%. It is still a business where one buyer is two-fifths of the top line. That is the same structural position Samsung Electro-Mechanics occupies inside the AI hardware chain — a component supplier whose fortunes track a handful of buyers’ order books.

🌍 Is Isu Petasys Actually a Global Leader?

We did not find a market-share figure for Isu Petasys in any of its filings — the company states a goal of “expanding market share,” which is a plan, not a disclosure.

This matters because the deep-dive convention is to quote a share number here, and there is a real temptation to manufacture one. Korean annual reports frequently do disclose market share; this one does not. The exact sentence in the half-year report reads that the Plant 5 investment “will contribute to expanding market share and increasing profitability” — future tense, no base, no source. We are not going to convert a company’s stated ambition into a share statistic.

What the filing does support, and what we can therefore say:

• The company describes its own ambition as “Global Top Tier PCB,” not as market leadership held.
• Its disclosed Korean capacity is running flat out. Capacity was 119 thousand m² in H1 2026 (against 228 thousand m² for full-year 2025 and 200 thousand m² for 2024), reported under a single site code, “Daegu Plant,” at a 96.8% utilisation rate. Annualised, H1 capacity runs about 4.4% above the 2025 full-year figure — very close to flat out. Read the basis carefully: this table is separate (parent-only), so “Daegu Plant” is the whole Nongong complex — four plants reported as one site — and it excludes the China manufacturing subsidiary entirely. It is not a group capacity figure.
• Output grew far less than revenue. Production was 115 thousand m² in H1 2026, against 214 thousand m² for full-year 2025 and 188 thousand m² for 2024 — annualising to roughly 7.5% growth on 2025, on the same separate basis.
• Its pricing is going up. Average export selling price was ₩4,875,174 per m² ($3,622) in H1 2026, against ₩4,477,422 ($3,326) for full-year 2025 (+8.9%) and ₩3,739,531 ($2,778) for 2024 (+19.7% into 2025). This series is separate-basis and covers exported product only. The company attributes this to a rising mix of high-value products.

All figures in this section: DART half-year report rcpNo 20260814002283, “생산 및 설비에 관한 사항” and “주요 제품 등의 가격변동추이.” Capacity, output, utilisation and ASP are disclosed on a separate (parent-only) basis — they describe the Daegu plant, not the consolidated group.

📈 Is the Business Getting Better or Worse?

Better on every income-statement line, and by a wide margin — H1 2026 operating profit rose 60.8% on revenue up 45.8%, lifting the operating margin to 20.0%.

Two lines in that table disagree with each other, and both are worth explaining.

EPS grew slower than profit — 45.3% against net profit’s 58.1%. The gap is the April 2025 rights offering. The weighted-average share count used for H1 2025 EPS was 67,457,524; for H1 2026 it was 73,409,219, up 8.8%. On a full-year basis the dilution is larger still: shares outstanding went from 63,246,419 at end-2024 to 73,409,219, an increase of 16.1%. Anyone comparing this half’s profit growth to last year’s needs to divide by a bigger number.

And operating cash flow went the other way entirely. That gets its own section.

🔍 Why Did Operating Cash Flow Go Negative in a Record Half?

Because working capital absorbed $138.9M and cash taxes nearly tripled — the company earned $81.8M of net profit and still ended the half with $36.9M less cash from operations than it started.

This is the one number in the filing that no press release leads with, and it is not in the headline table the company puts in its own business review. It is four lines down in the consolidated cash-flow statement:

The working-capital swing has two identifiable drivers on the balance sheet, both consolidated and both measured from Dec 31, 2025 to Jun 30, 2026:

• Trade receivables rose 63.7% — from $149.3M to $244.4M. Revenue over the comparable half rose 45.8%. Receivables grew faster than sales.
• Inventory rose 25.3% — from $222.0M to $278.2M.

Together those two accounts absorbed $151.3M of the $138.9M working-capital outflow (other operating liabilities offset part of it). Add cash tax of $30.5M — up from $11.9M, because last year’s profit is this year’s tax bill — and the record half turns cash-negative.

How was it funded? Not by borrowing much. The company ran down its short-term financial instruments — the parked proceeds of the 2025 rights offering — by $65.5M net ($141.4M disposed against $75.9M reacquired), drew $29.7M of new long-term debt, and spent $58.7M on property, plant and equipment. It also received $5.4M of government subsidy in cash during the half. Investing activities were therefore net positive for the period, which is unusual for a company mid-way through a $297.2M plant program, and tells you where the cash for the working-capital build actually came from.

None of this is a solvency question. Consolidated equity was $638.8M against $450.9M of liabilities at June 30, 2026, a debt-to-equity ratio of 70.6% — essentially unchanged from 70.5% at end-2025 and far below the 141.2% at end-2024, when the rights offering had not yet landed. It is a quality-of-earnings question, and it is the kind that only shows up if you read past the income statement. The same discipline applies here that we used when SK hynix’s non-operating income matched its operating profit: a record number is a starting point for a question, not the answer to one.

🏗️ What Is “Plant 5,” and When Does It Actually Arrive?

Plant 5 is a new Korean factory inside a ₩400bn ($297.2M) capacity program running from 2025 to 2028 — and the half-year report’s own dividend-policy section dates its completion to 2028, not 2026.

This distinction matters because the filing contains two statements about Plant 5 that read very differently:

• In the business review, a forward-looking sentence: 2026 growth is “expected” on the basis of, among other things, “the reflection of Plant 5 expansion effects.” The verb is 예상됩니다 — is expected. This is projected information, not a disclosed fact.
• In the dividend-policy section, a dated commitment: large-scale investment including “the new Plant 5 investment” is scheduled through 2028, and the company says it will pursue a more aggressive shareholder-return policy “after 2028, when the Plant 5 expansion is complete.”

The capital-spending schedule matches the second version, not the first:

So the honest reading is: 61% of the money is still unspent, the company itself dates completion to 2028, and the “Plant 5 effect in 2026” line is a forecast the company is entitled to make and a reader is entitled to discount.

One genuinely unusual detail sits underneath the program. Isu Petasys was designated a reshoring company under Korea’s Act on Assistance to Korean Off-shore Enterprises in Repatriation, and will receive a total of $29.7M in subsidy for domestic plant construction, plus a further $2.2M regional-investment subsidy. The money comes with conditions: it must complete the investment and new hiring within the commitment period, keep operating for three years from settlement, maintain its existing Korean sites, and — until May 2031 — cannot use, transfer, exchange, lend or pledge the subsidised assets contrary to purpose without ministerial approval. (Source: rcpNo 20260814002283, “정부보조금 수령에 대한 공시”)

A company that bills 97% of its revenue abroad is being paid by the Korean government to build at home. That is the Korean industrial-policy trade in one sentence, and it is invisible in English coverage.

💰 What Do You Actually Pay for Isu Petasys Today?

About 41 times trailing earnings and 9.5 times book for a $6.06bn company — the largest and by far the most profitable of Korea’s listed PCB makers, and the only one of the four whose shares are down year-to-date.

Against the other listed Korean PCB names — a group worth defining carefully, because they do not all make the same thing:

The table says something the individual numbers do not. Isu Petasys earns roughly four times the operating margin of the next-best name in the group, and it is the only one of the four whose shares have fallen this year while the other three re-rated hard. That is not automatically an opportunity — a stock at 41× trailing earnings has priced in a lot regardless of what its peers did — but it does mean the “Korean PCB stocks went up on AI” story is not the story here. Something specific to this name has been going the other way. The negative operating cash flow, the 16.1% share increase and the 2028 capex tail are all candidates; the filing does not tell us which the market is weighing.

🏷️ Which Isu Ticker Are You Actually Buying?

Only 007660 is Isu Petasys — three other Isu group companies trade on Korean exchanges, the holding company that controls all of them is not listed at all, and the closest look-alike ticker, 007810, is a different PCB maker entirely.

There is no American Depositary Receipt for Isu Petasys. Foreign investors buy the Korean line or nothing — see our complete list of Korean stocks available as US-listed ADRs for which Korean names do have one, and our guide to buying Korean stocks as a foreign investor for the account mechanics.

🌏 Is Foreign Money Buying or Selling Isu Petasys?

It depends entirely on the window you choose: foreigners were small net buyers over the last five sessions and substantial net sellers over the last twenty and over the full quarter — and domestic institutions were on the other side of all three.

The signs flip between the five-day and twenty-day columns, and we are going to call that what it is: a reversal, not accumulation. A five-session buy of $3.1M against a twenty-session sale of $71.2M and a quarter-long sale of $383.5M is a rounding error inside a sustained exit. Publishing only the five-day figure would let you write “foreigners are buying Isu Petasys again.” Publishing all three windows makes clear that the marginal buyer over the quarter has been the domestic institution, not the foreigner. If that reading pattern is new to you, we set out the mechanics in how foreign and institutional flows move Korean stocks, and track the weekly picture in the Foreign Flow Watch series.

🏛️ Who Controls Isu Petasys, and What Do Minority Holders Get?

An unlisted holding company, ISU Corporation, controls 21.47% directly and 26.13% with the founding family — and it has pledged 7.15% of Isu Petasys as collateral for ₩188bn ($139.7M) of borrowings, most of it for the benefit of a different group company.

The pledge — where the shares actually are

On September 4, 2026, ISU Corporation filed a routine-looking amendment to its 5% ownership report. Its stake had not changed by a single share. The reason for filing was, in the form’s own words, “change in contract terms on held shares (loan extension).” Buried in that filing is the detail:

The first three lines are ordinary: a holding company borrows against its main asset. The fourth line is not. Read the footnote to it and the structure becomes clear:

“This is a case in which [ISU Corporation] provides collateral for the principal-and-interest obligation borrowed from Shinhan Investment by Shinhan Insight No.1 Co., Ltd. — the underwriter of the hybrid capital securities issued by ISU E&C on 2026.03.06 — to raise the funds required for that acquisition. Should the collateral ratio fall short due to share-price movement, additional collateral shares may be provided.”

— DART rcpNo 20260904000407, footnote (*1), our translation

Put plainly: 4.49% of a listed AI-hardware company is posted as security for a ₩150bn loan whose proceeds funded a capital raise by the group’s unlisted construction arm. The obligation carries a 130% maintenance ratio and an explicit top-up clause tied to Isu Petasys’s share price. The filing does not disclose a trigger price. As an illustration only: 130% of the ₩150bn ($111.4M) principal is ₩195bn ($144.9M), which spread over 3,293,994 shares is ₩59,199 per share — but the obligation is defined on principal and interest, so the real threshold sits above that, and the contractual remedy named in the filing is posting more shares, not selling these ones.

The same affiliate, twice more

ISU E&C appears in two other places in the half-year report, and neither is in the ownership section:

• It is building the plant. Isu Petasys has a construction contract with ISU E&C for the new factory, with an unbilled balance of $91.5M at June 30, 2026. (Source: rcpNo 20260814002283, “건설공사 도급계약에 대한 공시”)
• Isu Petasys guarantees its obligations. The listed company provides a joint guarantee to the Korea Trade Insurance Corporation for ISU E&C’s export guarantee insurance, sized at $42.3M at June 30, 2026, up from $40.5M at end-2025. (Source: same filing, “특수관계자를 위하여 제공한 지급보증내역”)

So the shareholders of Isu Petasys are, at the same time: paying an affiliate to build their factory, guaranteeing that affiliate’s trade insurance, and watching 4.49% of their company sit as collateral for that affiliate’s capital raise. Each of the three is individually legal, disclosed and unremarkable in a Korean group. Read together they describe how much of a minority shareholder’s balance sheet is doing work for someone else — which is, in miniature, exactly the mechanism behind the Korea Discount.

Two smaller related-party items complete the picture. Isu Petasys paid $11.7M in outsourced processing fees to DNP Corporation, a related party, in H1 2026 — the single largest related-party expense. And it holds a real-estate fund investment that has gone backwards: an equity-method stake in the Gravity Private Real Estate Investment Trust No. 11, acquired for $22.3M in H1 2025 and carried at $17.8M at June 30, 2026 — a 20% markdown from cost in fourteen months. The movement note is explicit: the trust opened the half at ₩26,499,794k, took a ₩2,485,342k ($1.85M) equity-method loss, and closed at ₩24,014,451k. The structure includes a put and call between the two beneficiary-certificate classes at a guaranteed 9.5% internal rate of return, and the company values its stake only on the Class-2 certificates that carry residual profit rights. A separate real-estate trust, No. 24, is consolidated as a subsidiary. (Source: rcpNo 20260814002283, related-party and equity-method notes)

What minority holders actually receive

Very little today, with a dated promise of more. FY2025’s dividend was ₩230 per share, up from ₩150 — a total of $12.5M and a yield of 0.2% at the current price. The payout ratio was 10.7% on a separate (parent-only) basis and 10.5% on a consolidated basis — the same ₩16,884M dividend divided by separate net profit of ₩157,855M or consolidated net profit of ₩160,482M. Two numbers for one dividend, and the company’s target is set on the separate basis, so that is the one to track.

The direction is easy to misread. The dividend per share rose 53%, from ₩150 to ₩230 — but the consolidated payout ratio fell, from 12.8% for FY2024 to 10.5% for FY2025, because profit grew faster than the dividend did. Shareholders got more won per share and a smaller slice of the earnings. (Source: rcpNo 20260814002283, “주요배당지표” — the filing’s own table.)

In its Value-Up filings — the April 2025 report and its April 2026 progress update (rcpNo 20260430800992) — the company says plainly that cumulative investment through 2028 is expected to exceed cumulative net profit, and commits to raising the separate-basis payout ratio to 25–30% by 2029, with “a more aggressive shareholder-return policy” only after Plant 5 completes in 2028. That is an unusually honest way to say “not yet,” and it is a genuine Value-Up participation rather than a boilerplate one. It is also three years away. Our guide to Korea’s Value-Up program explains what these filings are and what they do and do not commit a company to.

One governance fact for the file: the board has four directors — three inside and one independent. All three inside directors hold zero shares and are recorded in the filing as affiliated-company executives in relation to the largest shareholder. With consolidated assets of $1.09bn the company sits below the ₩2tn threshold at which Korean law requires a larger independent contingent; assets have nearly doubled since end-2024. (Source: rcpNo 20260814002283, “임원 및 직원 등의 현황”)

📚 Lingo Check

🎯 Why It Matters for K-Export Stars

This site exists because a specific kind of information asymmetry is still open. Isu Petasys is the cleanest illustration of it we have written about.

In English, this is “a Korean AI PCB stock.” Every quantitative fact that makes it interesting — that 97.0% of its revenue leaves the country and 89.0% of its fixed assets do not; that its best-ever half produced negative operating cash flow; that the Korean government is paying it to build at home while restricting what it can do with the result until 2031; that 4.49% of the company is pledged against a construction affiliate’s borrowing — sits in a Korean-language filing that has no English version. None of it is secret. All of it requires reading the primary document.

It also sits precisely on this site’s thesis. K-Export Stars is about Korean companies whose fortunes are decided outside Korea, and there is no purer example on the exchange than a Daegu factory that books three percent of its sales at home. The same reading applies across the AI hardware chain we cover: Korea’s chip supercycle is usually told through memory, but the boards, substrates and passives underneath it are the same export bet with different margins and much less coverage — see also what SK hynix actually makes and who buys it.

This company is one of ten in our K-Semiconductor Filing Map — a supply-chain map where every cell is marked by what the filing itself supports, and each one links to the receipt number it came from.

Disclaimer: This article is for informational and educational purposes only and is not investment advice, a recommendation, or an offer to buy or sell any security. Figures are drawn from public filings on DART and market data from the Korea Exchange as of the dates stated, and may be superseded by later disclosures. Prices, valuations and flow figures are point-in-time and go stale quickly — verify against a live source before acting. Do your own research and consult a licensed adviser. See our corrections policy.