Unimicron Technology(3037) · PCBs & Substrates

Unimicron Technology from a Long-Term Owner's Perspective

Other languages
Quick ReadPlain-language overview · read this first

Unimicron Technology is a leading Taiwanese manufacturer of high-end printed circuit boards. It supplies customers in chip packaging and servers, and is one of the named suppliers in the AI and server boom. The report's stance is clear: the company itself is solid, but the current price is far too expensive, so it recommends Avoid.

Its profits mainly come from process capability and scale. It makes technically demanding substrates and high-density boards, rather than earning a premium through brand power. Demand for this business exists over the long term, but the issue is that it rises and falls sharply with the industry cycle. It is not the kind of business that becomes steadily more stable as it grows. One figure captures the problem best: after costs, how much is left from every 100 in sales. In the strong cycle of 2022, nearly 36 was left; by 2025, less than 14 was left. Profitability has thinned substantially, showing that the strength in 2022 was more a matter of catching a good year.

More importantly, the earnings on the books have not turned into cash that can be distributed to shareholders. In both 2024 and 2025, the company did make money, but because it spent heavily on expanding plants, cash was a net outflow after everything was counted. In other words, the profits are real, but a large share of them does not stay with the company.

What about the current price? This is the report's biggest concern. Based on its current profits, buying the whole company would take more than a hundred years to earn back the purchase price, meaning many years of future good news have already been paid for upfront. The report estimates a reasonable purchase price at roughly 90 to 150 per share, while the current share price has already reached 911, far above that range. The biggest risk is not that the company goes bankrupt, but that once the excitement fades, the valuation returns to normal and the price falls sharply.

The above is only an explanation of this report and is not investment advice. The stock market involves risk; invest with caution.

Lead

Unimicron is a leading Taiwanese PCB and IC substrate manufacturer whose revenue mix has shifted toward high-end ABF substrates and HDI. The core thesis is that AI server demand may lift the cycle, but heavy capex, volatile margins, and weak free cash flow make the current price hard to justify. Research rating Avoid: a capable cyclical manufacturer, but the valuation leaves no acceptable margin of safety.

Full report

Prices in the article are as of publication; see the valuation band above for the live price.

Note: The analysis below relies mainly on the company's official investor briefings, TWSE company profile, TWSE monthly revenue tables, and Reuters/LSEG. Because some TWSE annual report PDFs could not be stably accessed in the current retrieval environment, I cross-checked the annual financial summaries in the company's official investor presentations against the TWSE company page and Reuters financial pages. Where an item cannot be confirmed, I state "unknown" or "additional information required." As of 2026-06-09, Reuters showed Unimicron's latest delayed quote at about NT$911/share, with market capitalization of roughly NT$1.447 trillion; the TWSE company profile showed share capital of about NT$15.889 billion.

Conclusion First

Investment Rating: Avoid

Core view: Unimicron is an understandable manufacturing business with real process and scale capabilities in high-end substrates and HDI, but it is not the Buffett-style great business that naturally earns high returns, consumes little capital, and has strong pricing power. Profitability and gross margin fluctuated sharply from 2020 to 2025. Although growth recovered in 2024 and 2025, free cash flow remained negative, which means accounting profit did not smoothly convert into distributable cash flow. More importantly, the valuation implied by the current share price is far above the range I can accept based on Owner Earnings, asset value, and relative valuation. This is a combination of a good-business fragment, a powerful industry narrative, and a very poor purchase price.

Is there a margin of safety at the current price: No

Suitable investor type: More suitable for industry/cyclical investors who can tolerate both business-cycle and valuation volatility and are willing to track the ABF/substrate cycle; less suitable for conservative long-term value investors whose core discipline is margin of safety.

Biggest uncertainties:

  • The extent to which AI/HPC-related demand can structurally lift Unimicron's profit profile, rather than merely creating a temporary boom.

  • In Q1 2026, non-operating income accounted for a large share of earnings, and its sustainability is highly questionable.

  • The current market valuation has already prepaid a large amount of future good news. The largest future risk looks more like valuation mean reversion than a small earnings fluctuation.

Direct answers to the seven core questions:

  • Is this a business I can understand? Yes.

  • Is it a good business? Above average, but highly cyclical and not elegant enough.

  • Does it have durable competitive advantages? It has some advantages, but not a very wide moat.

  • Is management trustworthy, and is capital allocation rational? Basically qualified, but capital allocation is conservative and steady with dilution; it is not excellent.

  • Can it generate real, distributable cash flow over the long term? It can generate real operating cash flow, but distributable cash flow is highly volatile and has been consumed by high capex in the past two years.

  • Is there enough margin of safety at the current price? No.

  • What facts would overturn the investment judgment? If Owner Earnings can stabilize at the NT$25 billion to NT$30 billion level over the next 2 to 3 years, capex intensity clearly declines, and gross margin can stay around or above 20%, then my view that the current valuation is too high would need to be reassessed.

Business Understanding

What is the core business? Unimicron is essentially an electronics manufacturing company centered on PCB, HDI, IC substrates, and FPC. The TWSE company profile defines it as a supplier focused on PCB and substrate manufacturing; Reuters also describes its main businesses as PCB, HDI, FPC, rigid-flex boards, substrates, and IC testing burn-in systems. By 2025, the company's revenue mix by technology had already shifted clearly toward higher-end products: substrates about 58% to 59%, HDI about 27%, PCB about 11%, FPC about 2% to 3%, and others about 1%. This structure remained broadly intact in Q1 2026, at around substrates 59%, HDI 27%, PCB 11%, and FPC 2%. In other words, Unimicron is no longer an ordinary multilayer-board factory. It is closer to a manufacturer of high-end substrates and advanced interconnect platforms.

Who are the customers, and how does the company charge? Publicly verifiable materials do not directly provide a full customer list, so precise customer concentration is unknown and would require the annual report or full investor Q&A transcripts. The company has disclosed its revenue mix by application: in Q1 2026, Computer accounted for about 65%, Communication about 22%, Consumer & Others about 9%, and Automotive about 4%; for full-year 2025, the mix was roughly Computer 61%, Communication 21%, Consumer 12%, and Automotive 6%. This indicates that Unimicron mainly supplies customers related to chip packaging, servers, communications equipment, consumer electronics, and automotive electronics. Revenue is essentially driven by unit price x shipment volume of high-complexity boards and substrates. Because products require design-in, process qualification, and yield ramp-up, I believe customer switching is not easy. That is an inference based on product characteristics, not a verbatim statement from management in the materials accessed.

Is revenue recurring, stable, and predictable? Long-term demand exists, but revenue is not smooth. From 2020 to 2022, the company rode an upcycle and revenue rose from NT$87.893 billion to NT$140.489 billion; in 2023, it fell back to NT$104.036 billion; in 2024 and 2025, it recovered to NT$115.373 billion and NT$131.241 billion. This means it is not subscription-like or consumer-staples-like stable revenue. It is typical manufacturing revenue driven by cycle + product mix + utilization rate. Q1 2026 revenue further rose to NT$37.446 billion, up 24.5% year over year, but that says more about an improving cycle than about the business model becoming inherently stable.

What does the cost structure look like? Its cost structure is very similar to that of other capital-intensive electronics manufacturers: materials, labor, manufacturing overhead, and depreciation are high, and profitability is highly sensitive to capacity utilization. In 2025, cost of goods sold accounted for 86.1% of revenue and operating expenses for 8.9%; in Q1 2026, cost of goods sold fell to 82.0%, while operating expenses rose to 10.7%. This means that even when demand improves, profit leverage depends heavily on product mix and yield, not simply on price increases. Put differently, Unimicron earns money from "process + scale + cycle," not from "brand premium."

Does the company depend on a few customers, suppliers, channels, policies, or key people? The precise share of the top five customers or suppliers cannot be confirmed from the materials available in this review. But based on the application mix, the company's dependence on the computing/server chain is clearly rising, with Computer already reaching 65% of revenue in Q1 2026. The chairman changed in 2026, with Chien Shan-Chieh, who has a UMC Group background, taking over; the largest shareholder remains United Microelectronics Corporation, with a stake of about 12.97%. This suggests Unimicron is not dependent on a founder-type entrepreneur. It looks more like a specialized manufacturing platform within a group system.

Is the business simple, transparent, and easy to understand? If the stock market closed for five years, would I be willing to hold it? From the standpoint of "what it does, whom it sells to, and how it makes money," the business is understandable. From the standpoint of whether it can compound easily and with high certainty over the next decade, it is not simple, because technology transitions, yield, utilization, industry capacity, and end demand all affect returns. If the stock market closed for five years, I would be willing to own this business at a reasonable price. But at the current price, I would not. What I am willing to own is a business, not the right to own a business at any price.

Business understandability score: 4/5.

Industry and Moat

Industry stage and long-term demand The PCB/substrate industry is not a sunset industry, and long-term demand is not fake demand. Prismark's PCB reports continue to track the global PCB market, and industry organizations generally believe that after the 2023 trough and a mild recovery in 2024, 2025 entered a more visible growth phase. The company's annual report summary also noted that, driven by AI servers, low-earth-orbit satellites, and automotive electronics, the global PCB industry recovered to about US$73.6 billion in 2024. The issue is not whether long-term demand exists. The issue is whether such demand can create stable and excess returns on capital for companies. The answer is usually no, because the industry remains meaningfully cyclical.

Main competitors and industry structure The broader PCB field is fragmented. Industry commentary and annual report materials both indicate that even large players usually have only single-digit market shares. In the narrower high-end substrate segment, concentration is higher. Taiwanese companies commonly compared with Unimicron include Nan Ya PCB (8046.TW) and Kinsus (3189.TW), while in broader global PCB competition, Zhen Ding, Nippon Mektron, TTM, and others are also important players. In terms of product similarity, Nan Ya PCB and Kinsus are closer direct comparables than ordinary PCB manufacturers.

Is the industry profit pool concentrated, and what is the company's position? Unimicron's strength is not consumer branding. It lies in high-end substrates, HDI, multi-site mass production capability, and customer qualification history. The technology revenue mix in 2025 and Q1 2026 shows that the company has shifted its revenue center of gravity to the substrate business, which has higher margins and stronger technical barriers. The problem is that although substrates are better than ordinary PCBs, this is still manufacturing. Margins are pulled by the cycle, upstream materials, and expansion timing. The company's gross margin reached 35.9% in 2022, fell to 19.5% in 2023, compressed further to 14.1% in 2024, and was still only 13.9% in 2025. This shows that although the company's position is not weak, the industry profit pool is not naturally stable.

Pricing power and inflation resistance Unimicron does have pricing power, but only episodic pricing power. When ABF/high-end substrate supply is tight, manufacturers with strong yield and supply capability can obtain better prices and higher utilization. But the sharp gross-margin swing from 2022 to 2025 shows that this is not the consumer-brand type of power that says "I can raise prices forever." It is more like a cyclical advantage: stronger bargaining power in a favorable cycle and a reversion to manufacturing normality in a downturn.

Moat assessment by item Brand advantage: Weak to medium. It has no brand in the minds of end consumers, but reliable delivery and process reputation among B2B customers are valuable. Cost advantage: Medium. Scale, yield, and product mix create cost differences, but not enough to make peers unable to catch up. Scale advantage: Medium. Multiple sites, multiple product lines, and a larger capital platform form an advantage. Network effects: None. Switching costs: Medium. This is an inference based on the design-in and qualification cycles of high-end substrates/HDI, especially in server and automotive applications. Channel advantage: Weak. It is one part of the manufacturing supply chain, not the controller of the channel. Patent/license/regulatory barriers: Weak to medium. The real barriers come more from process know-how, yield, and customer qualification than from administrative licenses. Data advantage: Weak. Culture and operating capability: Medium. Financial reports and investor briefings show the company can recover revenue and profitability after a cyclical trough, but they also show that earnings quality is highly sensitive to execution and the cycle. Capital allocation capability: Below medium. It looks more like conservative financing and expansion management than excellent capital allocation focused on per-share intrinsic value.

Is the moat widening, stable, or narrowing? I would define it this way: locally strengthening, but not meaningfully widening overall. It is locally strengthening because the revenue mix is shifting further toward substrates and HDI, raising the share of advanced products. It is not meaningfully widening overall because margins had not recovered in 2025 to anywhere near the 2022 level, and free cash flow had not improved in tandem. Companies with truly wide moats usually do not see gross margin fall from 35.9% to 13.9% in just 2 to 3 years. Unimicron's advantages are real, but they are closer to "high-quality cyclical manufacturing capability" than to a "perpetual franchise."

Industry attractiveness score: 3/5. Moat strength score: 3/5.

Management and Capital Allocation

Is management honest, rational, and long-term oriented? The confirmable facts are these: in official investor briefings, the company continuously discloses revenue, gross profit, cash flow, balance sheet data, and revenue mix by technology and application. Disclosure quality is acceptable among Taiwanese manufacturing companies. In 2026, the chairmanship passed to Chien Shan-Chieh from the UMC system, reflecting governance continuity under group leadership rather than a sudden loss of control. Reuters' latest leadership information also shows that the company remains a fairly typical combination of professional management and group governance. On "candor," I have not seen obvious promotional disclosure that talks only about the story and ignores cash flow. But I also have not seen the public style of an excellent capital allocator with unusually strong self-demands around mistakes, capital efficiency, and per-share value. My judgment is: credible and steady, but not outstanding.

Ownership structure and alignment with shareholders Verifiable public information shows that United Microelectronics Corporation owns about 12.97% of Unimicron and is the most important shareholder. This structure has pros and cons. The benefit is group backing, with relatively stable governance and resource coordination. The drawback is that Unimicron is not a company where a high-ownership founder behaves like an owner and distributes cash accordingly. Alignment between investors and management/major shareholders is more like "group governance alignment" than Buffett's preferred model of "management with high ownership deeply tied to ordinary shareholders."

Is capital allocation excellent? How does the company use cash? For the past two years, I cannot call it excellent. I can only call it relatively prudent, but not friendly enough to per-share value. In cash-flow terms, the company generated NT$14.967 billion of operating cash flow in 2025, but capex was NT$25.617 billion, giving rough free cash flow of -NT$10.650 billion. To cover expansion and debt arrangements, the board approved a cash capital increase and convertible bond financing in 2025. The stated uses were repayment of maturing debt and replenishment of working capital. By Q1 2026, the cash-flow statement already showed NT$3.629 billion of cash capital increase proceeds. From the standpoint of keeping the balance sheet safe, this is rational. From the perspective of existing shareholders, it does involve dilution.

Dividends, buybacks, M&A, and debt repayment The company does pay dividends, but dividend intensity has clearly contracted. Annual cash-flow data show cash dividend payments falling from NT$12.190 billion in 2023 to NT$4.575 billion in 2024, and then to NT$2.295 billion in 2025. In the materials I accessed, I did not see buybacks worth treating as a core capital-allocation highlight. The more visible pattern is dividend contraction + cash capital increase + convertible bond issuance. On M&A, I did not find enough public evidence in the current materials to evaluate how much value acquisitions have created, so this should be marked as additional information required.

Management and capital allocation score: 3/5. The reason is simple: management does not appear irresponsible, but capital allocation looks more like serving the industry cycle and capacity expansion than being tightly centered on "maximizing per-share intrinsic value growth." For long-term owners, those two are different.

Financial Quality

Start with the most important conclusion: Unimicron's financial statements do not show obvious signs of accounting fraud, and operating cash flow has been positive over the long term, which is healthier than many concept stocks. But it is far from a "high-quality cash machine," because large amounts of cash have been consumed by capex in the past two years, leaving very limited cash truly available for shareholders. More importantly, the volatility in profitability fully exposes the cyclical nature of this industry: 2022 was a cyclical peak, and 2024 to 2025 was a recovery period, not a smooth compounding period.

Key financial metrics Scope: consolidated; unit: NT$100 million; rough FCF = operating cash flow - purchases of property, plant, equipment and investment property; EPS is basic earnings per share attributable to owners of the parent.

Year Revenue Gross Margin Operating Margin Net Margin Operating Cash Flow Capex Rough FCF EPS
2020 878.9 14.6% 4.6% 6.1% 181.6 145.5 36.1 3.74
2021 1,045.6 22.6% 12.6% 12.9% 429.4 231.7 197.7 8.98
2022 1,404.9 35.9% 27.2% 22.2% 610.8 320.7 290.1 20.08
2023 1,040.4 19.5% 8.6% 11.8% 309.0 229.4 79.6 7.88
2024 1,153.7 14.1% 4.4% 4.8% 102.9 261.3 -158.3 3.34
2025 1,312.4 13.9% 5.1% 5.8% 149.7 256.2 -106.5 4.38
2026Q1 374.5 18.0% 7.4% 14.4% 58.3 54.8 3.6 3.28

The raw data in the table come from annual and quarterly financial summaries published in the company's official investor presentations. Margins and rough FCF are calculated by me using a consistent approach.

How should we read these numbers? First, revenue grew rapidly from 2020 to 2022, plunged in 2023, and recovered again in 2024 to 2025. This is a strong company in a cyclical industry, not a compounder that becomes steadier over time. Second, the margin collapse is very clear: gross margin fell from 35.9% in 2022 to 13.9% in 2025, while operating margin fell from 27.2% to 5.1%. This tells me that the high profit in 2022 looked more like a cyclical windfall than a stable structural advantage. Third, Q1 2026 statements improved significantly, but non-operating income and expenses were NT$3.542 billion, even higher than operating income of NT$2.757 billion. That means Q1's high profit cannot simply be annualized and treated as "core operating profit."

Operating cash flow, free cash flow, and profit matching If we only compare operating cash flow with net income, Unimicron is not bad: from 2020 to 2025, operating cash flow was positive every year and exceeded net income in most years. That does not fit the typical pattern of inflated profits. The real problem is capex intensity. From 2023 to 2025, capex intensity was roughly 19.5% to 22.6% of revenue, causing rough free cash flow to be negative in both 2024 and 2025. In other words, the profit is real, but much of it cannot be freely distributed to shareholders. This matters greatly to long-term value investors: a company can have real earnings and still fail to be a high-quality cash-distribution machine.

Returns on capital and asset efficiency Based on public data, Unimicron's ROE was very strong at the 2022 cyclical peak, but it had fallen significantly by 2023 to 2025. Using net income attributable to the parent and average equity as a rough calculation, ROE was about 12.7% in 2023, 5.2% in 2024, and 6.5% in 2025. Using a simplified ROIC estimate, 2023 to 2025 was roughly 10% -> 4% -> 5%. These are not bad figures, but they are clearly below the level of a business with long-term excellent capital returns. Put differently, the market is currently giving it something closer to a high-growth-stock valuation, while its historical return quality looks closer to that of an above-average manufacturing leader.

Balance sheet and survivability Unimicron's financial safety is actually decent. Reuters/LSEG showed total debt of about NT$60.28 billion in 2025 and year-end cash of about NT$54.87 billion, leaving very low net debt. In Q1 2026, the official balance sheet showed cash of NT$60.179 billion, shareholders' equity of NT$116.953 billion, and no loss of balance-sheet control. Based on 2025 EBITDA of NT$28.555 billion, net debt/EBITDA was only about 0.2x. Financial leverage itself is not the main risk. To me, Unimicron's bigger risk is not "it cannot survive," but "if bought at a high price, investors may wait a long time before returns match the price paid."

Working capital and accounting quality From the end of 2025 to Q1 2026, inventory rose from NT$17.802 billion to NT$18.936 billion, and inventory days increased from 53 days to 55 days. Net accounts receivable rose from NT$27.565 billion to NT$28.976 billion, but receivable days improved from 71 days to 69 days. This shows that the company is indeed building inventory and increasing shipments during the recovery, but there is no visible uncontrolled deterioration for now. Overall, I do not see significant evidence pointing to aggressive accounting or profit manipulation. Still, non-operating income had a relatively high weight in the most recent quarter's profit, so some caution on near-term earnings quality remains necessary.

Share count changes, dividends, and buybacks The TWSE company profile showed share capital of about NT$15.889 billion in June 2026. Public news and announcement summaries show that in January 2026, the company completed a cash capital increase, issuing 46 million shares at NT$116/share, alongside convertible bond financing approved in 2025. Roughly speaking, this created about 3% dilution for existing shareholders. For a company I would want to own over the long term, such financing is not unacceptable. But it does show that value creation here in the past two years has not come from "buybacks + per-share value increase," but from "expansion + financing + waiting for the cycle to recover."

Owner Earnings and Valuation

Core conclusion on Owner Earnings If you ask in the Buffett sense, "How much money can this company truly leave to owners in a year?" my answer is: far below the optimistic expectations implied by the current share price. In 2025, the company reported consolidated net income of NT$7.550 billion and net income attributable to the parent of NT$6.673 billion; depreciation and amortization were NT$18.773 billion; operating cash flow was NT$14.967 billion. The problem is that 2025 spending on property, plant, equipment and investment property reached NT$25.617 billion. If all capex is treated as necessary spending, then 2025 Owner Earnings on a strict cash basis were actually negative. Clearly, part of that capex was for expansion rather than maintenance, but even so, it cannot be beautified into a high-quality cash cow.

A conservative Owner Earnings estimate I use a conservative but explainable framework:

  • Starting point: 2025 operating cash flow of NT$14.967 billion;

  • Maintenance capex: because the company is in a high-end substrate and expansion cycle, I do not accept a very low estimate. Conservatively, I treat 50% to 60% of total capex as maintenance capex, or about NT$12.8 billion to NT$15.4 billion;

  • Working capital: since 2025 to 2026 is a recovery and expansion period, I conservatively reserve another NT$1 billion to NT$2 billion for working capital absorption.

On this basis, Unimicron's current normalized Owner Earnings are only about NT$8 billion to NT$10 billion. This already gives some leniency by treating part of expansion capex as "growth capex." Based on the current market capitalization of roughly NT$1.447 trillion, the current price corresponds to about 145x to 181x conservative Owner Earnings. Even if we crudely annualize Q1 2026 parent-company net income of NT$5.043 billion, the current market capitalization still implies an annualized P/E of about 72x. This is not value-stock valuation. It is highly optimistic growth-stock valuation.

Method 1: Owner Earnings DCF The three scenarios below already give relatively generous growth assumptions for AI/high-end substrates. The discount rate is the return I require as a conservative shareholder, not the market's "story discount rate."

Dimension Bear Base Bull
Starting Owner Earnings NT$10 billion NT$18 billion NT$22 billion
Growth rate over next ten years 3% 8% 10%
Discount rate 11% 10% 9.5%
Terminal growth rate 2.5% 3.0% 3.5%
Equity value NT$124.9 billion NT$383.4 billion NT$622.8 billion
Per-share value about 79 about 241 about 392

These valuations are scenario estimates based on public financial data, not management guidance. Their implied assumptions are:

  • Bear case: the Q1 2026 improvement is largely cyclical repair, and normalized cash-generation capacity returns to around NT$10 billion;

  • Base case: AI/HPC demand lifts normalized Owner Earnings to around NT$18 billion, but margins remain clearly below 2022 levels;

  • Bull case: the strong conditions seen in Q1 2026 last for a relatively long time, and capex efficiency improves.

Even in my relatively generous bull case, intrinsic value is only about NT$392/share. Compared with the current NT$911, the gap remains enormous.

Method 2: Relative valuation Unimicron's current valuation is roughly TWSE official PE about 134x, Reuters forward P/E about 143x, P/B about 12.8x, P/S about 10.4x, and P/CF about 46.4x. Among directly comparable Taiwanese substrate peers, Nan Ya PCB's market capitalization in early June was about NT$546.0 billion, with a Reuters-specific P/E also around 180x and P/B about 11.5x to 12.2x; Kinsus had a market capitalization of about NT$373.6 billion and a Yahoo P/E of about 211.6x. This tells me two things. First, Unimicron is not the only expensive company in the sector. Second, expensive peers do not make it cheap. More likely, the entire ABF/substrate sector is being intensely priced by the AI narrative.

Method 3: Asset or liquidation value In Q1 2026, the company had cash and cash equivalents of NT$60.179 billion, funds and investments of NT$12.645 billion, property, plant, equipment and investment property of NT$126.148 billion, total assets of NT$266.881 billion, and shareholders' equity of NT$116.953 billion. Based on share capital, book value per share was only about NT$73.6/share. Even if you believe replacement value of fixed assets or some land value is above book value, the asset approach still struggles to lift value into the several-hundred-NT-dollar range, let alone NT$911. Therefore, investing in Unimicron at the current price is almost entirely a bet on "many years of sustained high future profitability," not on an "asset discount."

Overall valuation conclusion

  • Conservative intrinsic value range: NT$80 to NT$120/share

  • Reasonable intrinsic value range: NT$140 to NT$240/share

  • Optimistic intrinsic value range: NT$250 to NT$390/share

  • Current price relative to intrinsic value: There is no discount to any of my three scenarios; it is even above the upper end of the optimistic scenario.

  • Required margin of safety: Given industry cyclicality and capex intensity, I would require at least a 35% to 40% discount.

  • Ideal buy price range: NT$90 to NT$150/share

  • Acceptable hold price range: NT$150 to NT$250/share

  • Clearly overvalued range: above NT$350/share

This set of ranges does not mean Unimicron can only be worth a few dozen NT dollars. It means that if you insist on buying from a long-term owner's perspective and require returns commensurate with risk, you cannot pay a nearly "infinitely optimistic" price for a capital-intensive and clearly cyclical manufacturer.

Margin of safety judgment The current price not only has no margin of safety, it has already entered a zone where the company must beat expectations for many consecutive years before investment returns can become barely reasonable. The most fragile assumption in the valuation is that the market assumes the uplift from AI/server demand will remain with the company for a long time, at high margins and low capital consumption. The historical financial record does not provide enough evidence to support that. For value investors, this is the classic case of a good company at a bad price.

Risks, Checklist, and Final Judgment

The most important risks and the opposing view The strongest counterargument is not complicated: bulls see the AI narrative; bears see cyclical manufacturing plus a huge valuation bubble. Bulls will say Unimicron is an indispensable ABF/HDI beneficiary of AI/HPC infrastructure. Bears will say 2022 already proved that profits in this type of company can surge and collapse, while the price the market is assigning in 2026 is no longer buying a "recovery" but buying the "perpetualization of multi-year super-cycle conditions." If, over the next 2 to 3 years, gross margin cannot return to around 20%, Owner Earnings cannot rise, and high capex plus financing remain necessary, valuation contraction will become the main source of permanent capital loss.

On specific risks, I focus most on the following categories. Competition risk: Nan Ya PCB, Kinsus, and broader global PCB/substrate manufacturers are all benefiting from the same AI narrative. No one in the industry can rest easy. Technology substitution and product migration risk: If packaging routes, customer architectures, or material systems change, current advantages may weaken. That is normal in advanced manufacturing. This risk is more an industry-characteristic inference than a specific disclosed event from the company. Cyclical risk: The margin changes from 2022 to 2025 already show that cycle swings can significantly change earnings. Customer and application concentration risk: Computer-related revenue reached 65% in Q1 2026, and dependence on the AI server/computing chain is rising. Capex risk: Operating cash flow is positive, but free cash flow was negative in 2024 and 2025, which means expansion mistakes would directly erode shareholder returns. Overvaluation risk: This is the core risk today, not a side issue. TWSE official PE has reached about 134x. Financing and dilution risk: The company has actually carried out a cash capital increase and convertible bond financing in recent years. FX, interest-rate, supply-chain, and geopolitical risks: As an export-oriented, cross-region manufacturer, these risks cannot be absent. They are just not the first balance-sheet-level risk at present.

What facts would make me admit that the current "Avoid" judgment is wrong? I would be willing to reassess if the following conditions occur:

  • Over the next 8 quarters, the high substrate/HDI share continues, while gross margin can stay above 20% rather than spike for one quarter and then fall back;

  • Normalized Owner Earnings can stabilize above NT$25 billion to NT$30 billion, and capex intensity declines significantly;

  • Operating cash flow and free cash flow improve together on a sustained basis, without relying on financing to raise the cash balance;

  • The market price falls sharply back near my reasonable value range.

Conversely, the following facts would strengthen my bearish view:

  • Gross margin falls back into the 13% to 15% range;

  • Inventory days keep rising while revenue growth slows;

  • Significant equity financing appears again, or high capex continues without improving cash returns;

  • The latest-quarter profit structure, where "non-operating income is materially higher than operating income," repeats.

Comparison with other opportunities Compared with broad-market indexes, Unimicron's advantage is upside sensitivity, while its disadvantages are single track, single company, and single valuation assumption. Compared with a diversified index such as the S&P 500, outperforming with Unimicron at the current price requires being right simultaneously on AI demand, substrate supply, company execution, the pace of capex, and the absence of a large valuation pullback. That is obviously much harder than buying an index. Compared with risk-free yield, the Reuters page showed the U.S. 10-year Treasury yield at about 4.57%. For a company whose valuation is already extremely high and whose historical free cash flow is not stable, you should demand expected returns well above that level. Based on my valuation work above, Unimicron does not currently offer such compensation. If I could hold only 5 assets, I would not put it in the portfolio.

Investment Checklist

Question Conclusion
Can I understand this business? Pass
Does it have long-term stable demand? Pass
Does it have a durable moat? Uncertain
Does it have pricing power? Fail
Can it generate stable free cash flow? Fail
Are its returns on capital excellent? Uncertain
Is management trustworthy? Pass
Is capital allocation rational? Uncertain
Is the balance sheet solid? Pass
Is valuation below intrinsic value? Fail
Is the margin of safety sufficient? Fail
Would I feel comfortable holding it long term? Fail
What key facts would make me sell? Gross-margin decline, continued dilution, free-cash-flow deterioration, worsening application mix
Am I only tempted to buy because the share price has risen or market sentiment is strong? Very likely; requires serious self-check

The conclusions in the table come from the integrated assessment above of business model, earnings volatility, capex, financing actions, and current valuation.

Final investment conclusion

【Final Rating】 Avoid

【One-sentence investment thesis】 Unimicron is a reasonably competitive high-end substrate/HDI manufacturer, but the current price has prepaid too much multi-year AI strength and high profitability. For conservative long-term value investors, the odds are clearly unattractive.

【Core bullish arguments】

  • The business is understandable, and the revenue mix has clearly concentrated toward higher-end products such as substrates and HDI.

  • Fundamentals improved visibly in Q1 2026, with revenue up about 24.5% year over year and gross margin recovering to 18.0%.

  • The balance sheet is not fragile. Net leverage is low, and short-term debt repayment is not the main concern.

  • The long-term demand direction from AI/servers/high-performance computing is real.

【Core bearish arguments】

  • The industry and the company's earnings are highly cyclical. The margin swings from 2022 to 2025 already prove this is not a stable compounder.

  • Free cash flow was negative in 2024 and 2025, showing that real distributable cash flow is unattractive.

  • Non-operating contribution was high in the latest quarter, so earnings quality cannot be simply annualized.

  • The current valuation is extremely expensive. TWSE/Reuters PE is about 134x to 143x, far beyond my acceptable range.

  • There has been a cash capital increase and convertible bond financing in recent years, diluting shareholders.

【Key assumptions】

  • AI/HPC demand can structurally lift substrate profitability rather than merely create a short-term boom.

  • Capex will not remain for a long time at a level that suppresses free cash flow.

  • The competitive landscape will not quickly deteriorate because of new capacity expansions.

  • Within the current earnings improvement, the contribution from core operations will increase rather than continuing to rely on non-operating items.

【Fair Buy Price】 NT$90 to NT$150/share. Basis: I apply a 35% to 40% margin-of-safety discount to my NT$140 to NT$240/share reasonable value range. Given the industry's strong cyclicality and high capital consumption, I am unwilling to pay a high premium for a "manufacturing leader with a medium moat."

【Target Holding Period】 If the price enters a reasonable range, I think this type of asset is suitable for a 5 to 10 year holding period; at the current price, I do not recommend establishing a long-term value position.

【Expected Annualized Return】 Based on the three scenarios above and assuming that long-term market price reverts toward fundamentals:

  • Bear case: about -20% to -22%/year

  • Base case: about -12% to -17%/year

  • Bull case: about -8%/year

These returns assume purchase at the current price of about NT$911 and reversion to my estimated intrinsic value range after ten years. They sound harsh, but that is exactly why the main problem today is not that the company is bad. The price is too high.

【Maximum Loss Risk】 If the market eventually reprices Unimicron as an "excellent but cyclical manufacturer" rather than a "perpetual high-growth pure AI beneficiary," a share-price reversion toward NT$150 to NT$250 or even lower is not unimaginable. Relative to the current price, that would imply a 70% to 85% level of capital loss. If an industry downturn coincides with a valuation derating, the decline could be larger. This risk is not bankruptcy. It is valuation mean reversion after buying at a high price.

【Tracking Indicators】

  • Whether substrate revenue share remains above 58% to 60%

  • Whether Computer/AI/HPC application share continues to rise and is not a short-term spike

  • Whether gross margin can stabilize above 18% to 20%

  • Whether operating margin improves at the same time, rather than relying only on non-operating items

  • Whether operating cash flow and free cash flow strengthen together

  • Whether annual capex clearly falls from the NT$25 billion level

  • Whether equity financing or convertible bond expansion appears again

  • Whether inventory days and receivable days deteriorate

  • Whether monthly revenue stays high without obvious inventory-clearance pressure

  • Whether net debt/EBITDA remains low.

【Signals That Would Trigger Reassessment】

  • Gross margin remains above 20% for 4 to 8 consecutive quarters, with operating margin rising at the same time

  • Free cash flow turns positive and stays positive continuously

  • Capex falls, while revenue and profit still grow

  • Management stops dilutive financing and shifts toward greater emphasis on per-share value

  • Valuation falls materially back into my reasonable range

  • Conversely, if non-operating contribution continues to be much higher than operating profit, inventory builds rapidly, or another large financing occurs, I would strengthen the Avoid judgment.

【Final Recommendation】 Put calmly, Unimicron is not a company that "cannot be bought"; it is a stock that "should not be bought at the current price." If you are willing to act as a long-term business owner, your attention should be on whether the company can continuously produce high-quality Owner Earnings, whether it has a truly widening moat, and whether the purchase price leaves enough room for mistakes. At today's price, I do not see that room. The most prudent approach is not to chase strength, but to wait for valuation to return to a range where fundamentals can work for you.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.

804631892303

PCB SubstratesAI Server InfrastructureAdvanced ManufacturingCyclical StocksTaiwan Technology
Reader Q&A10

Baillie Framework · Ten Questions for Growth Investing

10

Hunting ten-year five-baggers among great growth stocks — pressing the upside question: "Can it get much bigger?"

Baillie Framework · Ten Questions for Growth Investing — score profile: 40/100 total Ceiling 5/10 · Revenue 2x 5/10 · Next engine 4/10 · Moat 5/10 · Reinvention 5/10 · Management 4/10 · Customer need 5/10 · Unit economics 3/10 · 5x path 2/10 · Blind spot 2/10 0510 How large is its market ceiling? Is it expanding an existing market, or creating an entirely new one? — 5/10 Ceiling 5 Can its revenue at least double over the next five years? Will growth mainly be driven by volume, price, or new businesses? — 5/10 Revenue 2x 5 Five years from now, what will take over as the next growth engine? Does this "second curve" exist today? — 4/10 Next engine 4 What is its core competitive advantage? Will this moat widen or narrow over the next three to five years? — 5/10 Moat 5 If its core business is disrupted, does it have the DNA to reinvent itself? How does it handle mistakes and bad news? — 5/10 Reinvention 5 Does management, especially the founder, have a long-term perspective and deeply aligned interests with the company? Is it willing to sacrifice current profits for five to ten years from now? — 4/10 Management 4 If it disappeared tomorrow, how much would customers miss it? Is its growth model sustainable and not dependent on harming society or exploiting regulation? — 5/10 Customer need 5 What are the unit economics of this business, in terms of gross margin and incremental returns? Does it improve or deteriorate with scale? Where does the money it earns go? — 3/10 Unit economics 3 What conditions must hold simultaneously for it to rise fivefold in ten years? Are those conditions realistic? What expectations are embedded in today's share price? — 2/10 5x path 2 Why has the market not recognized all this yet? Is it because investors do not understand it, look down on it, or cannot see far enough? What will become the "narrative inflection point"? — 2/10 Blind spot 2
  • How large is its market ceiling? Is it expanding an existing market, or creating an entirely new one?5/10

    Conclusion: Unimicron has a high market ceiling, but it is more about expanding and upgrading an existing electronics manufacturing market than creating a brand-new platform market. AI servers, HPC, and advanced packaging will raise the usage, layer counts, and unit prices of ABF substrates and high-end HDI. The essence, however, remains a high-value manufacturing segment within the PCB/IC substrate supply chain, not a new market like cloud, operating systems, or ecosystem platforms.

    The ceiling comes from the industry pool itself being large enough. Prismark estimates that the global PCB market will be over USD 85 billion in 2025 and about USD 123.3 billion by 2030, with growth driven by AI infrastructure, data centers, networking equipment, automotive electronics, and advanced computing. Compared with Unimicron's 2025 revenue of TWD 131.241 billion and EPS of 4.38, that external market space is still substantial. If the company keeps raising its share and unit prices in ABF, HDI, and AI server substrates, there is still room for revenue to move to another level.

    Unimicron is already positioned on this long slope. The product mix in the report shows substrates at about 58-59% and HDI at about 27%, with Computer/server-related products at about 65% in Q1 2026. External data also show that the upcycle is strengthening: Q1 2026 revenue was TWD 37.446 billion, gross margin was 17.96%, and EPS was 3.28; May 2026 revenue was TWD 14.060 billion, cumulative revenue for 1-5 months was TWD 65.439 billion, up 26.75% YoY. This shows AI/HPC/ABF/HDI is not an empty story, but real demand already reflected in revenue and capacity investment.

    But this ceiling is not "infinite." The current price is about TWD 969, with a market cap of about TWD 1.45 trillion. A fivefold outcome would imply about TWD 4,845 per share and a market cap of about TWD 7.3-7.7 trillion. To support that scale, Unimicron cannot merely follow industry growth. It must prove over the long term that it can capture a larger profit pool, sustain more stable gross margins, and build stronger bargaining power. The report's gross margin decline from 35.9% in 2022 to 13.9% in 2025 reminds us that this is still a capital-intensive manufacturing business with clear supply-demand cycles.

    So the answer is: Unimicron's market ceiling is high in the sense of being a leader in high-end electronics manufacturing. AI is expanding the existing PCB/substrate market and pushing value toward ABF, HDI, and advanced-packaging-related segments. But it is not creating a brand-new market. Its growth ceiling depends on industry expansion, customer qualification, capacity ramp, yield, and the competitive landscape, not on a software-like platform that can be replicated indefinitely.

    Jun 9, 2026
  • Can its revenue at least double over the next five years? Will growth mainly be driven by volume, price, or new businesses?5/10

    Conclusion: It has a chance, but a revenue doubling is not "locked in." Starting from 2025 revenue of TWD 131.24 billion, doubling in five years means about TWD 262.0 billion, equivalent to about a 15% CAGR. That is not an absurd target during an upswing in substrates/HDI, but it requires AI/HPC/server demand, ABF tightness, and Unimicron's capacity ramp to hold up for several consecutive years. Short-term data do support strong momentum: Q1 2026 revenue was TWD 37.446 billion, up +24.5% YoY; May 2026 monthly revenue was TWD 14.060 billion, up +32.37% YoY, and 1-5 month revenue was TWD 65.439 billion, up +26.75% YoY. Simple annualization of the first 5 months already points to roughly TWD 157.0 billion, but it still needs another large step to reach TWD 262.0 billion.

    The main growth driver should be "volume + mix," not consumer-brand-style price hikes. The company's revenue center of gravity is already in IC substrates/ABF and HDI. The report's 2025/Q1 2026 framing puts substrates at about 58%-59% and HDI at about 27%. Reported company commentary also mentioned that AI and high-performance computing orders are lifting utilization, the AI product share is rising, and 2026 capital budget has increased to about TWD 34.0 billion, with about 70% allocated to ABF capacity and process capability. This shows the core path is more high-end ABF/HDI shipments, a higher server/AI product mix, and improved utilization and yield. The price side has cost pass-through and high-end product ASP support, but this is not strong brand pricing power where prices can rise at will.

    New businesses are more accurately high-end manufacturing extensions, not a second and completely different company. Smooth trial production and qualification at the Guangfu ABF plant, plus trial production and small-volume shipments at the Thailand plant, will contribute incremental growth, but these still sit inside the same capital-intensive electronics manufacturing model, where profits come from equipment, customer validation, yield, and capacity utilization.

    So my judgment is: a five-year revenue doubling can be treated as an optimistic but discussable upside scenario, not the base case. If that scenario materializes, the primary drivers would be capacity expansion and product mix upgrade, with price improvement only secondary. The risks are that ABF/HDI supply expands too quickly, AI server pull-ins slow, or gross margin returns to the 13%-15% range. Especially around the current price of about TWD 969 and market cap of about TWD 1.5 trillion, the market is already clearly betting on this path. Revenue doubling alone may not be enough; what matters is whether it converts into more stable operating profit and free cash flow.

    Jun 9, 2026
  • Five years from now, what will take over as the next growth engine? Does this "second curve" exist today?4/10

    Conclusion: Five years from now, the most likely successor is not a brand-new business, but the second stage of higher-end AI/HPC substrates and high-end interconnect manufacturing: higher-layer-count/larger-size ABF, advanced packaging substrates, high-end HDI, and supporting overseas capacity. It already exists today, and is even part of the current main growth line. But it is not an independent new S-curve. It is more like Unimicron pushing the same high-end manufacturing capability into harder, more expensive, and more capital-intensive segments.

    The evidence is that this curve is not a paper concept. Company commentary indicates that the overall AI product share is expected to exceed 60% in 2026, and the board also raised 2026 capital budget to about TWD 34.0 billion, with 70% used to expand ABF capacity and improve process capability. In recent operations, May 2026 revenue was TWD 14.06 billion and cumulative 1-5 month revenue was TWD 65.439 billion, up 26.75% YoY, showing that demand from AI server ASICs and high-end ABF has already entered the financial statements. There are also signs of execution on the capacity side: trial production and qualification at the Guangfu ABF plant have been smooth, while the Thailand plant has entered trial production and already has small-volume shipments/customer qualification.

    But I would apply a discount: this "second curve" is more an upgrade of the main curve than a business-model leap. Automotive, low-earth-orbit satellites, communications, and overseas capacity may all contribute incremental growth, but based on current information, they are not yet independent engines strong enough to replace AI/HPC ABF. The true successor remains advanced packaging and specification upgrades in ABF and HDI driven by AI ASIC/GPU/HPC demand, which bring both volume and price uplift. In other words, if Unimicron's growth engine five years from now works, it will not be because there is "one more new story." It will be because customers keep requiring larger sizes, higher layer counts, higher yields, and shorter lead times, and the company can turn that difficulty into higher utilization and gross margin.

    So my judgment is: the second curve exists today, but its quality is only moderately favorable, not the kind of clear new platform curve seen in top-tier growth stocks. Around the current TWD 969 share price and about TWD 1.45 trillion market cap, the market has already priced the AI/ABF succession story aggressively. The next proof that this curve can truly take over is not the single statement that "AI share keeps rising," but whether gross margin can stabilize around 20% or above, whether free cash flow after capex can turn positive, and whether overseas/high-end new capacity can ramp without obvious dilution to shareholders. If it cannot, this looks more like a strong cyclical upgrade than a new growth engine that independently takes over five years from now.

    Jun 9, 2026
  • What is its core competitive advantage? Will this moat widen or narrow over the next three to five years?5/10

    Conclusion: Unimicron's core competitive advantage is genuine high-end manufacturing capability, not monopoly-style pricing power. Over the next three to five years, the moat is likely to widen in parts, but its width is limited. Its advantages mainly come from process know-how in ABF substrates and HDI, yield ramping, customer qualification, multi-site mass-production capability, and a product mix that keeps concentrating in high-end substrates/HDI. The company is no longer an ordinary PCB manufacturer. In its product mix, substrates account for about 59% and HDI about 27%, while AI/HPC demand is driving high utilization and expectations for product price increases.

    The strongest part of this moat is the ability to "make it, make it consistently, and earn enough customer confidence for volume allocation." High-end ABF/HDI is not a standardized component that customers can casually move to another factory. Design-in, qualification, reliability testing, and yield stability all take time. Once a supplier enters server, HPC, communications, or automotive chains, customers do face some switching costs. The progress of capacity at Unimicron's Guangfu plant, Thailand plant, and other sites also shows that it is using capital, process capability, and customer qualification to deepen supply capability. The improvement in the company's Q1 2026 revenue of TWD 37.446 billion, gross margin of 17.96%, and EPS of 3.28 also supports the judgment that when the cycle turns up, the company can convert high-end capacity into profit.

    But this is not a wide moat where "the bigger it gets, the steadier it becomes, and the longer it lasts, the more profitable it is." Direct competitors include Nan Ya PCB and Kinsus, while the broader PCB/substrate chain also includes Zhen Ding, TTM, and others. Customers will also maintain multi-supplier strategies over the long term. More importantly, the margin evidence is restrained: Unimicron's gross margin fell from 35.9% in 2022 to 13.9% in 2025, and only recovered to about 18% in Q1 2026. This shows it has process barriers and customer barriers, but pricing power is heavily affected by supply-demand cycles, utilization, material costs, and peer capacity expansion. It cannot raise prices unilaterally for long.

    So my judgment is: over the next three to five years, the technical moat will widen, while the economic moat will widen only modestly. If AI/HPC demand for ABF and high-end HDI remains tight, Unimicron's yield, qualifications, and capacity will make it stronger than ordinary PCB manufacturers. But once peer capacity catches up and customers complete dual-sourcing, excess profits will be pushed back toward manufacturing norms. At the current valuation of about TWD 969 per share and about TWD 1.5 trillion market cap, the market is already rewarding this moat in advance rather than underestimating it.

    Jun 9, 2026
  • If its core business is disrupted, does it have the DNA to reinvent itself? How does it handle mistakes and bad news?5/10

    Conclusion: Unimicron has the ability to reinvent itself, but mainly through engineering-led moves into adjacent arenas, not through a change in business-model species. If its core business is disrupted, for example if ordinary PCB or traditional substrates are replaced by higher-end packaging, new materials, or customer-owned capacity, its most likely response is to keep migrating toward interconnect manufacturing with higher technical thresholds. It has already moved from ordinary PCB into IC substrates, HDI, high-end ABF, and AI/server applications. The company has also disclosed that the overall AI product share is expected to exceed 60% in 2026, with capital budget of about TWD 34.0 billion, of which about 70% will be used to expand ABF capacity and improve process capability. This shows it has the organizational ability to reallocate capacity along with customers' next-generation products.

    But the boundary is also clear: it is still the same capital-intensive manufacturing model. The Guangfu ABF plant and Thailand plant moving through trial production/qualification and volume ramp are examples of high-end capacity migration. The company disclosed that Guangfu plant qualification is proceeding smoothly, while the Thailand plant has entered trial production and the small-volume shipment/customer qualification stage. That is good, but it does not turn the company into software, a platform, IP licensing, or an asset-light business. If the disruption is simply "low-end PCB being replaced by high-end HDI/ABF," Unimicron has a chance to keep up. If the disruption comes from a major change in packaging architecture, customers internalizing the key substrate value, or ABF/HDI again becoming low-margin capacity competition after expansion, the evidence for reinvention is much weaker.

    Its attitude toward mistakes and bad news is "adequate disclosure, but not excellent capital allocation reflection." Public materials do not only tell the AI story: 2025 revenue was about TWD 131.24 billion, EPS was 4.38, and dividend was TWD 2, and these results were disclosed together with the company outlook. Although Q1 2026 net profit of TWD 5.043 billion and EPS of 3.28 reached a 13-quarter high, during the same period the company also disclosed cash capital increase, convertible bond, and overseas GDR financing arrangements. Combined with 2025 gross margin of about 13.9% and negative free cash flow, this suggests management is willing to put cyclical volatility, margin pressure, and financing needs on the table. Transparency is not poor.

    The real deduction is that disclosing bad news is not the same as high-quality error correction. As of the delayed quote on June 9, Unimicron was about TWD 969, with a market cap of about TWD 1.45 trillion, and the market was already pricing it as an "AI substrate winner." If future reinvention still depends on continuing capital raises, debt/GDR issuance, and heavy-capex trial and error, common shareholders bear dilution and cyclical risk. My judgment is: Unimicron has an engineering-led self-renewal gene and can migrate toward adjacent higher-end processes if the core business is disrupted. But it has not yet proven that it has top-tier DNA for cross-business-model reinvention, rapid admission of capital allocation errors, and putting per-share value first.

    Jun 9, 2026
  • Does management, especially the founder, have a long-term perspective and deeply aligned interests with the company? Is it willing to sacrifice current profits for five to ten years from now?4/10

    Conclusion: management has a long-term perspective, but the interest alignment is not top-tier. Unimicron looks more like a professional manufacturing platform under UMC group governance than an owner-operator with high founder ownership and personal wealth deeply tied to the common stock. At the current valuation of about TWD 969 per share and about TWD 1.45 trillion market cap, this point is especially important.

    On ownership and personnel, public information shows that United Microelectronics holds about 12.97%, while Chien Shan-Chieh, as the legal representative, personally holds 0 shares. In 2026, UMC co-president Chien Shan-Chieh became chairman. This brings governance stability, group resources, and semiconductor experience, but it is not a structure where the founder family or core operator holds a high ownership stake. In other words, the alignment between ordinary shareholders and management comes more from professional reputation, group governance, and performance assessment than from a strong linkage between personal wealth and long-term compounding of the share price.

    The willingness to invest for the long term is clear. The company raised 2026 capex to TWD 34.0 billion, with about 70% invested in ABF substrate capacity expansion and process improvement, and the new chairman also publicly emphasized that resources would be more focused on advanced technology areas such as substrates and PCB. This shows management is willing to sacrifice short-term free cash flow, dividend flexibility, and asset turnover efficiency for the AI/HPC substrate opportunity, betting on capacity and customer positioning five to ten years out.

    But the quality of capital allocation deserves a discount. In recent years, Unimicron has not returned capital to shareholders through buybacks or by raising intrinsic value per share. It has supported expansion through a cash capital increase of 46 million shares at TWD 116 per share, TWD 4.0 billion of convertible bonds, and a subsequent GDR financing plan; meanwhile, it proposed a 2025 cash dividend of TWD 2 per share. From an industrial operating perspective, these financings may be rational and can preserve the balance sheet and expansion pace. But for long-term minority shareholders, they mean dilution and dividend contraction, showing that when "expanding capacity" and "improving per-share value" compete, management leans toward the former.

    So my judgment is: Unimicron's management is credible, stable, and willing to invest for the long term, but it is not the ideal Baillie framework type of "deep founder alignment + anti-dilution + per-share value first." It can sacrifice current cash returns for its industrial position five to ten years from now. Whether that sacrifice ultimately returns to per-share value depends on post-expansion ABF gross margin, free cash flow, and whether financing dilution continues, not merely on management's long-term perspective.

    Jun 9, 2026
  • If it disappeared tomorrow, how much would customers miss it? Is its growth model sustainable and not dependent on harming society or exploiting regulation?5/10

    Conclusion: Customers would miss Unimicron, but more as "qualified capacity that is indispensable in the short term" than as "the only irreplaceable supplier over the long term." Its growth model is broadly a healthy expansion of AI/semiconductor infrastructure, not a way to make money by harming society or exploiting regulatory arbitrage. But sustainability is constrained by geopolitics, supply-chain dual-sourcing, overseas capacity validation, environmental protection/energy/water resources, and customer concentration. Based on the delayed quote on 2026-06-09, Unimicron's share price was about TWD 969, with a market cap of about TWD 1.45 trillion, meaning the market has already fully priced its importance in high-end ABF/HDI and AI server substrate.

    If Unimicron disappeared tomorrow, the customers hurt most would be high-end ABF substrate, HDI, and AI server substrate customers. The issue would not merely be losing one supplier. These products require design-in, process validation, yield ramp, and stable delivery. CNA reported that the company's Q1 2026 production-line utilization was expected to remain above 90%, the AI product share was estimated to exceed 60%, and about 70% of the TWD 34.0 billion capital budget would be used for ABF expansion and process capability, showing that customers are indeed pulling qualified capacity. Cnyes also reported that its Guangfu ABF plant trial production and qualification are proceeding smoothly, while the Thailand plant has entered trial production and customer qualification/small-volume shipment stage. Such qualification and ramping are themselves customer switching costs.

    But "being missed" does not mean "being irreplaceable." Unimicron's moat is mainly process know-how, yield, scale, and a history of customer validation, not patent monopoly or network effects. Nan Ya PCB, Kinsus, and global PCB/substrate manufacturers can still serve as alternatives and dual-source candidates. The more realistic situation is that customers would face short-term disruption in delivery, yield, qualification, and product mix, while over the medium to long term they would mitigate the impact through dual-sourcing, requalification, order transfer, or changes in packaging/design paths. So Unimicron is important in the supply chain, but it is not a single point of scarcity for the entire industry.

    On social and regulatory sustainability, I would rate it positively but not perfectly. Its growth comes from AI servers, HPC, semiconductor packaging, and high-speed interconnect infrastructure. This demand itself is not a business model like tobacco, gambling, predatory finance, or data-abuse monetization. The company's ESG pages also list water resources, greenhouse gases, energy, waste, and chemical safety as environmental KPIs, and disclose product-side green product compliance requirements such as RoHS/REACH. The real risks are that PCB/substrate manufacturing is naturally water-intensive, power-intensive, and chemical-intensive, while overseas expansion must deal with local environmental, labor, permit, and customer audit requirements, plus geopolitical and supply-chain restructuring risks. Therefore, Unimicron's growth is not built on harming society, but whether it can continue depends on whether it can maintain environmental compliance, customer diversification, and capex discipline while expanding.

    Jun 9, 2026
  • What are the unit economics of this business, in terms of gross margin and incremental returns? Does it improve or deteriorate with scale? Where does the money it earns go?3/10

    Conclusion: Unimicron's unit economics are "improving during a cyclical repair, but not yet proven to be a high-return compounding machine." High-end ABF/HDI gives gross margin upside, and Q1 2026 revenue of TWD 37.446 billion, gross margin of 17.96%, and EPS of 3.28 improved clearly from the 2025 trough. But 2025 revenue was TWD 131.24 billion and net profit attributable to the parent was TWD 6.673 billion, with full-year gross margin only about 13.9% and operating margin about 5.1%. This shows profits mainly depend on utilization, product mix, and supply-demand, not permanent pricing power.

    Scale does not automatically make the business better. Scale brings customer qualification, yield learning, and fixed-cost absorption, and when ABF/AI orders are tight it improves gross margin. But if capacity expansion runs ahead of demand, depreciation, materials, labor, and idle capacity can also amplify profit pressure in reverse. In particular, Q1 2026 net profit cannot be directly annualized into stable compounding. Based on the report's framing, Q1 operating profit was TWD 2.757 billion, while non-operating income and expenses were TWD 3.542 billion, higher than operating profit. That means the core business has indeed improved, but not all profit came from core operations.

    The money it earns mainly goes back into capacity, rather than returning in large amounts to shareholders. In 2025, operating cash flow was about TWD 14.967 billion, capex was about TWD 25.617 billion, and free cash flow was about TWD -10.65 billion. At the same time, the company raised 2026 capital budget to about TWD 34.0 billion, with 70% used for ABF expansion and process capability. This explains where capital goes: cash is used for ABF capacity, process upgrades, factory automation, and working capital rather than buybacks. The company also uses cash capital increase, convertible bonds, GDR, and other financing to fund capital needs, creating dilution pressure for existing shareholders.

    So the incremental return of this business is still "unproven." If TWD 34.0 billion-level capex can truly translate into long-term gross margin above 20%, positive free cash flow, and higher ROIC, scale will make the business better. If AI/ABF supply-demand is only cyclical tightness, post-expansion depreciation and financing dilution will absorb a fair amount of growth benefits. At the current share price of about TWD 969 and market cap of about TWD 1.45 trillion, the market is already pricing it like a high-quality growth stock, but the unit economics that have been proven look more like high-end cyclical manufacturing: flexible, process-barriered, and also constrained by heavy capital and slow cash recovery.

    Jun 9, 2026
  • What conditions must hold simultaneously for it to rise fivefold in ten years? Are those conditions realistic? What expectations are embedded in today's share price?2/10

    Conclusion: Starting from the current price, a fivefold gain in ten years requires a nearly perfect set of conditions to hold simultaneously, and the realism is very low. Starting from a share price of about TWD 969 and a market cap of about TWD 1.5 trillion, a fivefold outcome means about TWD 4,845 per share and a market cap of about TWD 7.3-7.7 trillion. Working backward at 25x PE, that would require about 290-310B TWD of net profit, or TWD 290.0-310.0 billion, not TWD 29.0-31.0 billion. The scale is jarring by comparison: 2025 net profit attributable to the parent was only TWD 6.673 billion, while Q1 2026 single-quarter net profit was TWD 5.043 billion, which annualizes to about TWD 20.17 billion. The target net profit is about 44-46 times 2025, and 14-15 times Q1 annualized. Q1 profit also had a high non-operating contribution, so it cannot be treated simply as a clean long-term operating baseline.

    To get there, at least five things must happen at the same time. First, demand for AI/HPC/ABF substrates must not be a two- or three-year shortage cycle, but a structural expansion lasting a decade. Second, Unimicron must keep gaining high-end share amid expansion by Nan Ya PCB, Kinsus, and global substrate capacity, without being pulled back to manufacturing norms by the supply cycle. Third, gross margin and net margin must return to, or even exceed, levels near the cyclical peak for the long term, instead of falling from 35.9% gross margin to 13.9% as they did from 2022-2025. Fourth, high capex must convert into real free cash flow, without continued dilution through cash capital increases, convertible bonds, or GDR. Fifth, ten years from now the market must still be willing to give a cyclical manufacturing company a 20-25x, or even higher, valuation multiple. The hardest part is not any single condition, but that all of them must hold simultaneously.

    These conditions are not completely without factual basis: the company is indeed positioned in the chain of rising demand for AI servers, HPC, and ABF substrates, and management has also said that the overall AI product share is estimated to exceed 60% in 2026, with capital budget rising to TWD 34.0 billion and about 70% allocated to ABF capacity and process capability. But there is a huge earnings gap between a "good cycle" and a "fivefold return in ten years." If one assumes future net margin as high as 15%-20%, earning TWD 290.0-310.0 billion of net profit would still require about TWD 1.5-2.1 trillion of annual revenue, roughly 11-16 times the TWD 131.24 billion revenue in 2025. For a capital-intensive manufacturing leader with expanding supply and unstable pricing power, that is an extremely demanding assumption.

    The expectations embedded in today's share price are already very full: StockAnalysis shows a 2026-06-09 share price of TWD 969, market cap of about TWD 1.45 trillion, PE of 129.95x, and Forward PE of 55.85x. In other words, what the market is buying now is not "recovery from the 2025 trough," but "AI substrate tightness becoming long-lasting, margins structurally rising, capex efficiency improving, and valuation staying high for a long time." This also explains why it is already far above the report's optimistic value range of TWD 250-392. The current price leaves almost no margin for gross margin retreat, ABF overcapacity, persistently weak free cash flow, or further financing dilution.

    Jun 9, 2026
  • Why has the market not recognized all this yet? Is it because investors do not understand it, look down on it, or cannot see far enough? What will become the "narrative inflection point"?2/10

    Conclusion: The market has not failed to recognize it. It has recognized it very clearly. Unimicron now looks more like a company where the "AI/ABF substrate story has already been priced with high intensity" than a typical Baillie-style "overlooked long-term winner." As of 2026-06-09, the share price was about TWD 969, the market cap about TWD 1.45 trillion, and PE about 129.95x. The analyst consensus is Buy, but the target price is below the current price. This shows the market does not misunderstand or dismiss it; it has already bought forward the narrative of AI servers, ABF tightness, and high-end HDI capacity.

    The real disagreement is this: the market sees "long-term shortage in AI infrastructure," while skeptics see "a cyclical peak in capital-intensive manufacturing that may be mistaken for structural compounding." The positive facts are real. The company reported Q1 2026 revenue of TWD 37.446 billion, gross margin of 17.96%, and EPS of TWD 3.28, with profit reaching a 13-quarter high. May revenue was TWD 14.060 billion, up 32.37% YoY, and 1-5 month revenue grew 26.75% YoY. Management also said AI/HPC demand is lifting utilization and product mix, and raised 2026 capital budget to about TWD 34.0 billion, with about 70% allocated to ABF capacity and process capability. But these facts are already in the price; they are not a hidden discount.

    So the answer is closer to this: the market is looking far ahead, but perhaps assuming too smooth a path. It is willing to believe that ABF/HDI supply-demand tightness, AI pull-ins, and overseas capacity ramp will last for years, but it has not yet fully proven that these incremental gains will become high-quality, distributable, low-dilution free cash flow. The key risks in the report remain: 2025 gross margin was only 13.9%, rough free cash flow was negative, and Q1 2026 non-operating income was higher than operating profit. This means current single-quarter EPS cannot yet be treated as "clean long-term operating profit."

    A positive narrative inflection point would be several consecutive quarters with gross margin stably above 20%, operating profit rather than non-operating gains becoming the main driver of earnings, free cash flow turning positive, capex as a percentage of revenue falling, new ABF/HDI capacity ramping without diluting margins, and overseas/new capacity such as the Guangfu plant and Thailand plant passing qualification smoothly and bringing real orders. Only then could the market narrative upgrade from "AI substrate cyclical stock" to "high-end interconnect platform-type growth stock."

    The negative inflection points are also clear: AI server pull-ins slow, ABF/HDI supply catches up or even becomes excessive, gross margin returns to 13%-15%, inventories and receivables start rising, the company continues to use cash capital increases, GDR, and convertible bonds to fund capex, or profit remains propped up by non-operating contributions. At that point, the market would reprice Unimicron from an "AI scarcity asset" back to an "excellent but cyclical manufacturing company," and valuation compression would hurt more than a modest swing in earnings.

    Jun 9, 2026
Ask about this report

Members can ask about this report; once answered it appears under "Reader Q&A" on this page. You can also highlight a passage in the text to ask about it directly.