United Microelectronics Corporation UMC Deep-Dive Research

Vertical Company History

UMC's starting point carries the mark of Taiwan's semiconductor era. Company materials and industry-institution materials define it as founded in 1980 and Taiwan's first semiconductor company. Public ITRI materials also state that UMC became Taiwan's first listed semiconductor company in 1985. The background was the institutionalization of professional manufacturing capacity after value-chain specialization began to mature, rather than a consumer-brand start-up. Design companies needed outsourced manufacturing, while manufacturing required scale, yield, process capability, and capital discipline. This became a key step in Taiwan's later formation of a complete semiconductor cluster. For today's investors, the most important point is that UMC has made money from the entrusted-manufacturing logic of "customers bring the design; UMC delivers yield and lead time" from the beginning, rather than from founder mythology. On the underwriting price and funds raised in 1985, currently accessible first-hand official archives do not provide clear, directly verifiable figures, so this report does not force in unverified details.

For a long period afterward, UMC grew with the entire foundry industry, but it ultimately followed a path clearly different from TSMC's. TSMC has gone deeper into advanced nodes, advanced packaging, and ecosystems. By Q1 2026, 3nm, 5nm, and 7nm together accounted for 74% of TSMC's wafer revenue, and quarterly gross margin reached 66.2%. UMC has instead shifted its center of gravity to mature and specialty processes. In current formal mass-production revenue, sub-14nm contributes zero. 22/28nm, 40nm, BCD, RFSOI, embedded memory, high-voltage processes, and other specialty technologies are its real home field. The 12nm push is being carried out through cooperation with Intel in the United States, not through self-built leading-edge lines. This difference is not simply a difference in "technology level." At root, it is a divergence in capital-allocation logic: TSMC bets on leadership and ecosystem lock-in, while UMC bets on repeatable mature demand, more controllable capital intensity, and the customer value of multi-region manufacturing.

This strategic divergence has not always looked correct. From 2018 to 2021, UMC also carried a heavy governance shadow. In 2020, the U.S. Department of Justice announced that UMC had entered a plea agreement in a trade-secret case and paid a US$60 million fine. In 2021, the company reached a global settlement with Micron, with both sides withdrawing litigation and UMC making a one-time payment. For the company's operations, the episode did not create a fatal cash hit, but it weighed on UMC's governance image and reminded the market that mature foundry is not a simple "low-tech, low-risk" industry. IP, compliance, cross-border cooperation, and geopolitics all genuinely affect valuation multiples. Today this issue is no longer the dominant share-price variable, but it remains a page that cannot be skipped when assessing management credibility.

What truly changed UMC's financial profile was the 2021-2022 industry supercycle. In 2022, revenue rose to NT$278.7 billion, gross margin reached 45.1%, operating margin was 37.4%, net income attributable to shareholders rose to NT$87.2 billion, full-year capex was US$2.7 billion, and management clearly wrote the 22/28nm portfolio, automotive, and specialty technologies into the growth axis. This round of high profitability did not come from nowhere. It came partly from pricing benefits created by industry supply-demand imbalance and partly from the company's earlier groundwork in product mix, customer structure, and specialty processes. Precisely because 2022 was so strong, the later decline needs to be understood correctly.

2023 was a year that tested UMC's quality. Industry destocking, weakening consumer electronics demand, high interest rates, and geopolitics all pressured valuation and orders. Company revenue fell to NT$222.5 billion, but gross margin still reached 34.9%, operating margin remained 26.0%, and net income still reached NT$61.0 billion. It was hurt, but it did not revert to original form like many companies that purely bet capacity on the cycle. In that year's shareholder letter, management repeatedly emphasized two things: structural profitability under weak demand, and the fact that 22nm specialty processes had entered mass production and were continuing to expand. The key signal is that UMC proved in a downcycle that it could live more decently than the industry average. That is one reason the market is willing today to give it a valuation above the traditional mature-node foundry average.

2024-2025 was a transition from "defending profit" to "resetting chips for the next phase." In 2024, revenue recovered to NT$232.3 billion, but gross margin returned to 32.6% and net income fell to NT$47.2 billion. The company also advanced joint development of a 12nm platform with Intel, explicitly linking its U.S. manufacturing foothold with customer supply-chain resilience. In 2025, revenue rose again to NT$237.6 billion, gross margin further compressed to 29.0%, operating margin was 18.5%, and net income attributable to shareholders was about NT$41.5 billion. The shareholder letter disclosed that Phase 3 expansion in Singapore had come online, Intel 12nm technology transfer was progressing smoothly, mass production was expected in 2027, and the company had formally written silicon photonics into its future roadmap. Put differently, 2024-2025 were not the prettiest profit years, but they were the years with the most future options.

By 2026, UMC had entered its second real pricing exam. The first exam was the 2023 industry trough, when the company proved it would not easily collapse. The second is now, when the company must prove that it can gradually turn 22nm specialty processes, selective price increases, U.S. 12nm, and a new optical-interconnect route into a higher profit center, rather than merely earning a bit more from an inventory-cycle rebound. Q1 2026 and Q2 guidance show that this path is indeed moving forward: utilization is rising, gross margin is returning to around 30%, and monthly revenue continues to set new highs for more than three years. But the stock market has already walked a long way down that path first. The real lesson of company history is that "when UMC looks its best, the stock is usually no longer cheap," rather than "whether UMC can recover."

Business Model and Moat

UMC's business model looks simple on the surface: customers hand chip designs to UMC, and UMC manufactures wafers using specified processes and earns foundry revenue. But real profit is squeezed out of "what type of orders, on which lines, and whether utilization can hold," not out of "taking a few more orders." The company's 2025 annual report defines its business scope as wafer manufacturing from 12nm to 0.6 micron and clearly states that its logic and specialty-process menu includes CMOS, 3D FinFET, RF CMOS, embedded non-volatile memory, embedded high voltage, BCD, RFSOI, silicon photonics, and 2.5D/3D advanced packaging. By Q1 2026, 22/28nm accounted for 34% of revenue, 40nm accounted for 18%, and 40-65nm also accounted for 18%. This shows that what truly feeds the company is the part of mature nodes with high requirements for process segmentation, not leftover old capacity.

This structure means UMC's profit sources have two layers. The first is fixed-cost absorption from utilization. Foundry depreciation is heavy. In Q1 2026, depreciation and amortization alone was NT$15.99 billion, close to one quarter of quarterly revenue. As utilization rose from 69% in Q1 2025 to 79% in Q1 2026, gross margin recovered from 26.7% to 29.2%, before fully reflecting Q2 volume-and-price guidance. The second layer is process mix. 22nm already contributed 14% of revenue in Q1 2026, and full-year 2025 22nm revenue rose 93% year on year. These products are usually tied to applications such as display drivers, power management, networking, and automotive control. Their ASP and gross-margin ceiling are below advanced nodes but clearly better than the most ordinary mature processes. In other words, UMC's operating leverage is refined leverage from "more expensive mature-node products with longer customer validation and slower replacement," not the explosive leverage seen in advanced nodes.

From a moat perspective, UMC has three real moats. The first is the depth of its specialty-process menu. Mature nodes are not something anyone can casually foundry. Many chips optimize for high voltage, low leakage, RF, embedded memory, temperature tolerance, and long-term reliability rather than minimum linewidth. In its 2025 annual report, UMC lists 14eHV, 28eHV-LP, 22eMRAM, 55BCD, 40/22nm RFSOI, silicon photonics, and other routes as R&D priorities. These processes cannot be migrated on a whim. Customer tape-out, validation, and certification cycles are long, and once passed they create stickiness. The second is geographic diversification in manufacturing. The company explicitly defines multi-site manufacturing in Taiwan, Singapore, Japan, and China, plus the U.S. 12nm cooperation with Intel, as differentiated value that gives customers supply-chain resilience. The third is manufacturing discipline within mature nodes. UMC's annual report itself lists operation, quality control, and cost management as core competencies. This matters especially in downturns, because mature-node industries often lose on yield, delivery, and cost, not on technology.

But it is also necessary to separate a real moat from a promotional moat. UMC has no network effect, no absolute pricing power created by advanced nodes, advanced packaging, and a design ecosystem like TSMC's, and no administrative and capital backing comparable to what Chinese domestic policy gives SMIC. Its moat is closer to "verifiable process capability plus manufacturing experience plus customer switching costs," not "others simply cannot enter." This directly produces two consequences. In good times, its cycle leverage is not as dramatic as TSMC's. In bad times, it will not rapidly go to zero like pure commodity capacity. But if Chinese mature-node competitors are willing to exchange subsidies for market share, UMC's moat will also be continuously eroded by price pressure. The company states this risk very directly in its own risk section.

On management and governance, UMC's current profile has strengths and shadows. The 2025 annual report shows that 6 of the 9 board seats are independent directors, so independent directors account for two thirds of the board. Chairman Stan Hung has held his current role since 2008, Jason Wang has served as CEO since 2014, the president and COO is SC Chien, and the CFO is Chitung Liu. The core management team is broadly stable. The issue is that insider ownership is not high: the chairman owns about 0.47%, the CEO about 0.23%, and the CFO about 0.04%. This is not necessarily bad, but it means ordinary shareholders and management are aligned more through compensation and reputation mechanisms than through large personal capital stakes. At the same time, the plea and settlement in the Micron trade-secret case mean the governance discount is no longer the main narrative, but it cannot be completely ignored.

In capital allocation, UMC's recent approach is to "significantly cut capex, maintain dividends, and avoid creating liquidity anxiety for itself." Cash capex in 2023 and 2024 was about US$3.0 billion and US$2.9 billion, respectively. It fell to US$1.6 billion in 2025, and the 2026 budget falls again to US$1.5 billion. At the same time, the annual report shows cash and cash equivalents of NT$110.66 billion at the end of 2025, and about NT$109.02 billion at the end of Q1 2026. The company is taking the path of gradually harvesting assets left by the major expansions of prior years, rather than using high leverage to bet on another new cycle. This capital discipline is one reason UMC deserves more respect than some mature-node peers.

Industry and Horizontal Peer Analysis

Putting UMC back into the industry makes it easier to understand why it makes money, and also why it should not be valued as a "direct AI beneficiary." According to Gartner data cited in UMC's annual report, global foundry market revenue was about US$173.7 billion in 2025, up 26.2% year on year. In 2026, it is expected to grow another 18.8% to US$206.4 billion. The main growth engines remain AI servers, high-performance computing, and advanced packaging. But the same page of the annual report also clearly states that utilization in the mature-process segment improved quarter by quarter in 2025, supported by stable consumer electronics demand, Chinese subsidy plans, and applications such as automotive, industrial control, and power management. In other words, most industry profit is still captured by TSMC at advanced nodes. The mature-process track is more about capturing the chain reaction of peripheral-chip demand after advanced-system expansion, not directly capturing large-model compute itself.

UMC's cycle is essentially a composite cycle created by the semiconductor inventory cycle, capital-expenditure cycle, and policy cycle. In an upcycle, the most beneficial variables are utilization, ASP, and specialty-process mix such as 22nm. In a downcycle, the most fragile variables are price and depreciation, because once mature-node supply is excessive, price wars first hit ordinary processes and then penetrate higher-value specialty processes through customer switching pressure. UMC's annual report even directly lists "Chinese manufacturers taking market share through price cuts supported by government subsidies" as a core risk, and specifically notes that this expansion overlaps with UMC's capacity and will affect orders, revenue, and profit. For this company, the cycle is a relationship between financial-statement lines, not an abstract concept.

Horizontally, I think the most representative deep comparison group is TSMC, GlobalFoundries, and SMIC. Vanguard International Semiconductor and Powerchip are certainly also important, but they are better used as secondary Taiwan mature-node comparisons rather than key anchors that determine UMC's long-term valuation ceiling. The table below puts the four companies on one page, and the differences are very clear.

The data in the table combines UMC, TSMC, GF, and SMIC's latest quarterly disclosures and market pages. The cross-company basis is not fully consistent, so it should be treated as a directional comparison rather than a fully comparable audited basis.

TSMC is a necessary comparison because it sets the valuation ceiling for the pure-play foundry leader, not because it looks most like UMC. TSMC's Q1 2026 revenue rose 35.1% year on year, gross margin was 66.2%, advanced nodes contributed 74%, and the market gives it more than 10x P/B. That valuation reflects AI/HPC demand, advanced packaging, customer ecosystem, and technology monopoly value. UMC trades at an approximate 62% P/B discount to TSMC, and that discount itself is reasonable. The problem is that UMC's current 3.87x P/B is already not low. It looks discounted only because TSMC is more expensive, not because UMC is "cheap." Translating "cheap relative to TSMC" directly into "UMC is cheap" is a classic comparison error.

GlobalFoundries is a closer international mature/specialty-process comparison for UMC, but it has developed into another type of company. GF's current narrative centers on U.S. manufacturing, communications infrastructure, automotive and data-center connectivity chips, and the localized value chain of silicon photonics and advanced packaging. Its Q1 2026 revenue was US$1.634 billion and gross margin was 27.6%, slightly lower in profitability than UMC, but the market gives it a higher PE. One reason is that capital markets are repricing the geopolitical value of "U.S. manufacturing plus AI connectivity infrastructure." Another is that GF recently launched a US$500 million buyback and started building a dividend framework. The competition between UMC and GF is about who can package mature and specialty processes as a long-term "non-commodity" asset, not about the most advanced logic.

SMIC is one of UMC's most realistic sources of pressure. Its market pricing and business logic are both different from UMC's. The market gives SMIC a premium as a "Chinese semiconductor localization platform," not as "discounted global mature-process cash flow." In Q1 2026, SMIC revenue was US$2.506 billion and gross margin was 20.1%. Q2 guidance calls for 14%-16% sequential revenue growth, while Reuters reported that SMIC planned to add about 40,000 12-inch wafers per month of capacity in 2026, with depreciation expected to increase by about another 30% year on year. This means that even if SMIC's earnings quality and valuation are not cheap, it will still create the most direct price and delivery-time competition for UMC in mainland China and for orders serving mainland customers. For UMC, the truly difficult fight is with a group of Chinese rivals willing to expand mature-node capacity over the long term and supported by relatively abundant policy and capital, not with TSMC.

Vanguard International Semiconductor and Powerchip are more like side mirrors. VIS's official IR page lists 2025 revenue at NT$48.591 billion, and Q1 2026 news language showed quarterly revenue of about NT$12.532 billion and EPS of NT$1.22. It is more purely exposed to 8-inch mature and specialty processes, so it can be used to observe roughly what pure mature-node assets are worth without U.S. 12nm or silicon-photonics options. Powerchip is more complicated. In addition to foundry, it has memory and asset-disposal issues. Micron's planned 2026 acquisition of its P5 fab actually shows that the capital-recovery speed and valuation logic of the mature/memory overlap track are not the same as UMC's. Including these two companies in the reference set makes UMC's niche clearer: it is a global challenger with specialty-process depth, multi-site manufacturing, and a U.S. option, but still unable to escape mature-node industry discipline. It is neither the leader nor the cheapest mature capacity.

Current Fundamentals and Valuation Analysis

Start with financials and operations. Over the past five quarters, UMC's revenue path has not been steep, but the recovery path is clear. 1Q25 revenue was NT$57.86 billion, gross margin was 26.7%, and utilization was 69%. 2Q25 revenue was NT$58.76 billion, gross margin was 28.7%, and utilization was 76%. 3Q25 revenue was NT$59.13 billion, gross margin was 29.8%, and utilization was 78%. 4Q25 revenue was NT$61.81 billion, gross margin was 30.7%, and utilization was 78%. 1Q26 achieved revenue of NT$61.04 billion, gross margin of 29.2%, and utilization of 79% in a traditional slow season. Looking only at this data, the conclusion is clear: demand is recovering, but it is a gradual mature-node recovery, not an explosive jump in orders.

The easily misread part sits at the bottom of the income statement. Q1 2026 EPS doubled year on year to NT$1.29, which looks very strong on the surface. But a breakdown shows that operating improvement was not as dramatic as EPS suggests. The company's pretax profit for the quarter was about NT$16.64 billion, non-operating income was about NT$5.37 billion, and income tax expense was only about NT$527 million. In other words, this "doubling" at the bottom line was amplified not only by operating leverage from utilization recovery, but also by obvious non-operating gains and a low tax burden. This distinction is very important for investors tracking UMC, because if the market directly extrapolates EPS doubling into the full year, it can easily overestimate the profit center.

Q2 and May revenue pushed the story another step forward. The company's Q2 guidance is 7%-9% sequential wafer shipment growth, a slight U.S. dollar ASP increase, utilization of 81%-83%, and a gross margin of about 30%. In May, UMC reported monthly revenue of NT$22.944 billion, up 17.78% year on year, with revenue for the first five months up 9.05% year on year. This shows that the first-half recovery continued into the middle of Q2, rather than stopping at a one-off Q1 rebound. The issue is that capital markets have already seen this. The stock closed at NT$133.5 on June 12, while the 52-week range had already stretched to NT$40.1-155.5. The financial statements have improved, but the stock is already "waiting for financials to exceed already high expectations," not "waiting for financials to get better."

The table below puts UMC's key operating figures over the past five years together, making it easier to see how it moved from a supercycle peak to today's moderate recovery phase.

Note: Revenue, margins, and net income are drawn from UMC's 2022-2025 annual reports and 2026Q1 capex review. 2021 gross margin and operating margin are back-calculated from that year's financial statements.

This string of numbers tells two stories. The first is that UMC's earnings peak already appeared in 2022, and today it is moving upward again from a lower step, not returning to a high-boom state. The second is that capex is clearly downshifting while profit has not yet fully recovered in sync. This means free-cash-flow leverage in 2026-2027 should, in theory, be better than in 2024-2025, because depreciation pressure is largely already in the accounts while new cash investment is no longer climbing. For mature-node foundries, lower capex is usually not bad news, provided it comes from the passing of an expansion peak rather than order problems. Based on UMC's guidance language and fab layout, it is closer to the former.

Valuation is the part of this report that most requires restraint. At the current price, Reuters/LSEG's figures for UMC are roughly P/B 3.87, adjusted PE 31.4x, and dividend yield 2.28%. Yahoo Finance gives a TTM PE of about 33.9x. For a mature-process foundry, this is no longer the range of a "cheap cyclical stock." It only still looks like a discounted asset because TSMC is expensive at more than 10x P/B. The problem is that the market will not automatically give UMC extra returns because it is "cheaper than TSMC." The real question is whether UMC's own operating quality is worth this price. My judgment is that if the market applies earnings multiples in the high 20s or even 30x to 2026 earnings recovery, it has already prepaid utilization recovery, 22nm improvement, and price adjustments quite fully.

Looking through cash flow makes this judgment clearer. In 2025, the company's operating cash flow was about NT$94.05 billion, compared with net income of about NT$41.53 billion, so OCF/net income was about 2.26x. In 2022, this ratio was about 1.66x. UMC's accounting profit is not weak, and cash conversion is more solid than at many semiconductor companies. The problem lies in the purchase price, not cash-flow quality. With 2025 capex already down to US$1.6 billion and the 2026 budget down again to US$1.5 billion, I prefer to view the company through "owner earnings" rather than reported net income. Conservatively assuming that roughly half to 60% of 2025 capex can be viewed as maintenance spending, 2025 owner earnings would land around NT$64.0 billion-72.0 billion, implying an owner-earnings yield of about 3.8%-4.3% on the current market cap. This yield is not bad, but it remains short of the margin of safety needed to heavily weight a mature foundry in a high-uncertainty cycle.

I use a hybrid "P/B plus owner-earnings multiple" approach for three scenarios rather than using PE alone, because UMC's profits fluctuate too much with the cycle, taxes, and non-operating items. Looking only at EPS can easily overstate the trend.

This is a framework-based projection of the range in which UMC may be repriced by the market over the next 12 months from the current price, not investment advice. In price-signal terms, I treat the conservative scenario as the "ideal buying zone," the base scenario as the "acceptable holding zone," and the upward result of the bull scenario as the "clearly overvalued line." The resulting price bands are: ideal buying around NT$80-88, acceptable holding around NT$96-112, and above NT$124, the market is basically already paying for the best version of UMC. At NT$133.5 on June 12, my margin-of-safety conclusion is "none."

Risks, Catalysts, and Zen Horizon Synthesis

The risks truly worth fearing for UMC are three variables that could change the long-term pricing logic, not short-term volatility. First, a Chinese mature-process price war has high probability and high impact. The company's own annual report states this very plainly. SMIC is maintaining high utilization while continuing to expand, indicating that mature-node supply discipline over the next two to three years will likely be much less healthy than in 2022. If mainland customers and customers serving the mainland market begin to place a larger share of orders with local fabs, UMC will first lose ASP, then utilization, and only then net income. The most dangerous part of a price war is that it hits both profit and valuation, because the market will quickly reclassify the "specialty-process upgrade" story as "mature-capacity commoditization."

Second, price adjustments may fall short of expectations, with medium-high probability and medium-high impact. Media discussion of a "broad 10% price increase in the second half" has been intense, but formal financial reports and investor-conference language have not given a uniform, unconditional number. Reuters and industry-chain reports are closer to descriptions such as "mature-node price increases, with some new orders or specific customer ranges at about 5%-10%." The easiest market mistake is to interpret "the company has started discussing price increases" directly as "profit will jump by the same percentage." Mature foundry is not SaaS. Whether price increases truly pass through to the financial statements depends on long-term agreements, new orders, customer structure, and competitor quotes. For investors, the items to watch are H2 ASP and gross margin, not the headline.

Third, Intel 12nm and the silicon-photonics option could lose speed, with medium probability and medium impact. This risk will not crush profit in the short term, because UMC's current revenue base does not depend on them. But it would hurt valuation. The market has already counted part of the future value of these two cards in advance. Official disclosures still put Intel 12nm on a path of qualification in 2026 and mass production in 2027, while silicon photonics is in 2026 risk production with later customer adoption still to be watched. If the timetable shifts out or customer validation slows, the financials may not immediately look bad, but the valuation premium for "mature foundry plus AI connectivity option" will retreat first. For a stock that has already risen a long way on expectations, that retreat would be painful.

Fourth, the income statement may be misread, with medium probability and medium impact. Q1 2026 EPS doubled year on year and looks attractive, but a low tax burden and non-operating income lifted the bottom line. If investors directly annualize Q1 EPS and then add H2 price-increase expectations, they will arrive at an overly optimistic full-year earnings picture. Once later quarters cannot sustain this bottom-line leverage, valuation compression can arrive quickly. This risk does not necessarily come from operating deterioration. It may simply come from the market realizing that it had previously calculated too much.

Positive catalysts are equally clear. First, if Q2 and Q3 continuously verify utilization above 81%-83%, gross margin around 30%, and a continued rise in the 22nm mix, the market will believe more strongly that "structural upgrade" is more than a headline. Second, if H2 ASP really moves up meaningfully and not at the expense of utilization, UMC's profit center will be recalculated. Third, if Intel 12nm produces more specific customer names, design adoption, and validation milestones, and if silicon photonics more clearly lands before mass production, the valuation story will be reignited. Negative catalysts almost match these point by point. If monthly revenue improvement stalls, ASP pass-through is not obvious, gross margin falls back below 27%, or competitors start winning orders through more aggressive pricing, the current high valuation will be hard to sustain.

The following tracking table contains the items I think are truly worth watching over the long term.

These indicators matter because they turn "story" back into "verification." UMC's variables are simple. The complexity is that the market will price them before they are realized. So tracking this company means watching whether utilization, ASP, 22nm share, and capex discipline are all turning together, rather than guessing at a larger grand narrative.

Putting vertical and horizontal analysis together, UMC's truly proven ability is "not spending recklessly in arenas where it is not strong, and running lines steadily in mature and specialty processes where it is strong," not "always running the fastest." Its past success has included both era dividends and management restraint in capital allocation; it has included the push of a semiconductor upcycle and the cushion of product-mix upgrades. These success factors still exist today, but their weights are changing. Cycle recovery is an open card, the 22nm upgrade is a half-open card, and Intel 12nm plus silicon photonics are hidden cards. The market's most likely mistake now is treating all three cards as if they will realize at the same time and with the same strength.

Looking at the key variables over the next one, three, and five years, the differences are large. Over one year, utilization, ASP, and gross margin matter most, because they decide whether the share price can digest valuation from a high level. Over three years, the most important issue is whether UMC can continue pushing up the revenue share of 22nm and higher-value specialty processes, avoiding a relapse into purely price-sensitive mature capacity. Over five years, the most important issue is whether Intel 12nm and silicon photonics are merely "good-sounding stories" or second curves that can form sustainable revenue. A true rerating must come from an affirmative answer to the third question, not from repeatedly telling the same cycle story through the first two questions.

Bull case:

• Utilization has recovered from 69% to 79%, and Q2 guidance moves further to 81%-83%, so mature-process operating leverage is still being released.

• 22nm revenue grew 93% year on year in 2025 and accounted for 14% of revenue in Q1 2026, so the structural upgrade is not empty talk.

• Capex has fallen from US$3.0 billion and US$2.9 billion in 2023-2024 to US$1.6 billion in 2025 and a 2026 budget of US$1.5 billion, creating better free-cash-flow conditions than in the previous two years.

• Multi-site manufacturing plus Intel 12nm's U.S. foothold provide a real selling point for customer supply-chain resilience.

Bear case:

• Q1 EPS doubling year on year was amplified by a low tax burden and non-operating income, so bottom-line leverage can easily be overestimated by the market.

• Chinese mature-process expansion and subsidy-backed competition are still intensifying, and UMC itself lists price war as a major risk.

• The current roughly 3.87x P/B and roughly 31x adjusted PE are already not cheap for mature-process foundry.

• Intel 12nm and silicon photonics remain options rather than the main profit line, but the stock has already priced in a considerable part of that imagination.

If this investment loses 50% three years from now, the first script I worry about most is this: from 2027 to 2028, Chinese mature-process manufacturers continue expanding 28/40/55/90nm capacity; customers serving Chinese domestic demand begin localizing orders more systematically; UMC is forced to cut prices in ordinary mature nodes and some specialty processes to preserve utilization; gross margin falls from around 30% to 22%-24%; the market recasts it as a "victim of mature-capacity oversupply"; P/B compresses from close to 4x back to around 2x; and the stock returns to the NT$60-70 range. The frightening part of this script is that it does not require the company to do anything wrong. It only requires industry discipline to worsen.

The second script is this: in 2027, Intel 12nm qualification keeps slipping, silicon photonics remains in risk production and customer testing, and the H2 price increase turns out to be more structural and selective than the market's imagined "broad price increase." In this case, UMC's core profit may not deteriorate sharply, but investors will suddenly realize that they had been valuing a mature foundry through advanced connectivity and AI infrastructure. If valuation falls from more than 30x earnings back to 18-20x, and P/B falls from close to 4x back to 2.5-3x, the stock price would be enough to suffer a sharp drawdown.

The final conclusion can be written plainly. UMC is a more stable and more disciplined mature and specialty-process foundry than many people assume. It has proven that it will not easily bleed too quickly in an industry downturn, and it is now proving that specialty-process upgrades such as 22nm can push it from a simple "mature-capacity seller" toward a higher-quality layer. But that does not automatically mean the stock deserves a heavy position at this moment. Today's question is how much "getting better" the share price has already discounted, not whether the company is getting better. My judgment is that the market has already capitalized utilization recovery, 22nm upgrades, H2 price adjustments, and even part of the Intel 12nm and silicon-photonics option in advance.

If the stock price falls to a more reasonable level tomorrow, I would be more willing to discuss UMC's long-term value rather than short-term risk. The truly interesting part of this company has always been whether it can slowly lift its profit center and cash returns on the seemingly unsexy mature and specialty-process track, then use U.S. manufacturing and silicon photonics to push the valuation ceiling a little higher. It is not about whether it will become an AI foundry hero. That path exists. The current price simply leaves investors too little room for error.

【Company Profile Scores】

• Fundamental quality: Medium

• Growth: Medium

• Moat: Medium

• Financial strength: Strong

• Management credibility: Medium

• Valuation attractiveness: Low

• Risk level: Medium

• Suitable investor type: Cyclical / Dividend

【Investment Rating】

• Rating: Watch

• One-sentence investment thesis: Earnings recovery is being realized, but the current price has already front-loaded utilization, price increases, and option narratives.

• Three price signals: Ideal buy price: NT$80-88

• Acceptable holding price: NT$96-112

• Clearly overvalued price: Above NT$124

• Current price category: Clearly overvalued

• Worth waiting for a better price: Yes; at minimum, wait for the stock to return below NT$96 while utilization can still be seen holding above 80%. The most ideal re-entry range is NT$80-88. The opportunity cost of waiting is that if H2 price increases exceed expectations and the stock continues rising on sentiment, one may miss a trend-driven gain. But the current price offers too little room for error.

• Target holding period: 3-5 years

• Expected annualized return: Conservative about -40% to -34%; base about -28% to -16%; bull about -7% to +3%

• Maximum loss risk: About 50%; the triggers are a Chinese mature-process price war, utilization falling below 75% again, gross margin sliding toward the low-20% range, and delays in the Intel 12nm and silicon-photonics options.

• Signals that trigger reassessment: Utilization below 75% for two consecutive quarters.

• Gross margin below 27% for two consecutive quarters.

• 22/28nm revenue share falls below 30%, or 22nm share declines instead of rising.

• H2 ASP fails to reflect price adjustments, and monthly revenue turns negative year on year again for consecutive months.

• Intel 12nm mass-production timing is delayed again, or silicon photonics remains in risk production without customer landing.

【Ideal/Fair Buy Price】80-88 TWD Basis: Corresponds to the conservative-scenario valuation, with an additional margin of safety of about 20%.

【Valuation Range】

• current: 133.5 (as of the 2026-06-12 close)

• bear (conservative · ideal buying zone): [80, 88]

• base (reasonable · acceptable holding zone): [96, 112]

• bull (optimistic · above the clearly overvalued line): [124, 138]

Research uncertainties:

• The underwriting price and fundraising size of the company's initial Taiwan listing in 1985 are not complete enough in currently accessible first-hand archives, so this report does not force in unverified figures.

• The market rumor of a "broad 10% price increase in the second half" is not formal financial-reporting language. This report has tried to distinguish formal disclosures from industry-chain reporting, but final verification still depends on H2 ASP and gross margin.

• Intel 12nm and silicon photonics are still closer to project and roadmap stages, with insufficient detailed customer-adoption data. The valuation section can only treat them as options, not as core-business forecasts.

• Peer valuation bases span markets and accounting standards. Horizontal comparison is more suitable for directional reading, and each multiple should not be treated as strictly comparable.

Reference sources:

• UMC 2025 annual report and 2024/2023/2022 annual reports.

• UMC Q1 2026 financial-results press release, monthly revenue announcements, and quarterly data pages.

• UMC's official disclosures on Intel 12nm and the silicon-photonics roadmap.

• Reuters/LSEG and Yahoo Finance market pages for 2303.TW.

• TSMC Q1 2026 disclosures and Reuters/LSEG market page.

• GlobalFoundries Q1 2026 and full-year 2025 disclosures.

• SMIC Q1 2026 results announcement, 2025 annual report, and Reuters reporting.

• Public materials from DOJ, UMC, and Micron on litigation and settlement.

• Vanguard International Semiconductor official IR and related quarterly news, and Reuters reporting on the Micron/Powerchip transaction.

Other Securities Mentioned in the Report

• 2330.TW - TSMC, the leader in advanced processes and advanced packaging, and the most important reference point for UMC's valuation discount.

• GFS.US - GlobalFoundries, an international mature and specialty-process comparable that directly mirrors UMC in U.S. manufacturing and silicon photonics.

• 00981.HK - SMIC, whose Chinese mature-process expansion and policy support are most likely to compress UMC's pricing and order room.

• 5347.TWO - Vanguard International Semiconductor, a Taiwan pure-play mature-node comparable used to observe the earnings center of 8-inch and specialty processes.

• 6770.TW - Powerchip, a mature-node and memory overlap-track comparison that helps frame capex discipline and asset value.

• INTC.US - Intel, UMC's 12nm partner and a key partner for UMC's U.S. manufacturing and long-term technology option.

• MU.US - Micron, both a historical litigation counterparty and an important reference point for Taiwan capacity integration and capital flows.

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