UMC’s beat is less about demand acceleration than mix discipline at capped capacity

The print matters because it argues against the bear case that United Microelectronics ADR would have to buy revenue with price at sub-80% utilization. What was priced in was modest revenue upside and no sharp inflection: the street had revenue at 1,942.0 million $ and EPS at 0.12, consistent with a company still working through mature-node digestion rather than entering a new upcycle. What actually surprised was not the size of the top-line beat, since revenue of 1,976.5 million $ was only +1.8% above estimate, but the EPS delivery at 0.13, +8.3% above estimate, while the company’s own reported accounts showed gross margin returning to 30.7% in Q4 FY2025 from 29.8% in Q3 FY2025. The variant perception is that this is a margin-quality beat, not a demand beat: revenue rose +4.5% QoQ on the company’s quarterly history, but utilization “stayed the same as the previous one, around 78%,” per Chi-Tung Liu, so the incremental gross margin came from mix, cost absorption, and price discipline rather than a factory loading snapback.

That distinction is important because the headline revenue trend could be misread as a cyclical acceleration. Q4 FY2025 revenue was 61,809.3 million $, up +4.5% QoQ and +2.4% YoY, after Q3 FY2025 revenue of 59,127.0 million $ grew only +0.6% QoQ and declined -2.2% YoY. The sequential improvement is real, but it arrived without the utilization improvement that would usually signal broad-based wafer demand. Chi-Tung Liu’s own framing captured the company-accounting basis: “Consolidated revenue was TWD 61.81 billion, with a gross margin around 30.7%.” That quote earns attention because management did not pair the revenue level with a loading recovery; the same call says utilization was around 78%. If investors pay for this as an early-cycle utilization story, they are paying for something the call did not deliver. If they pay for it as a mix and capacity-discipline story, the numbers support that view.

The financial trajectory reinforces that interpretation because gross margin has rebuilt while revenue has only crawled higher. Gross margin troughed at 26.7% in Q1 FY2025 on revenue of 57,859.0 million $, then improved to 28.7% in Q2 FY2025, 29.8% in Q3 FY2025, and 30.7% in Q4 FY2025. Over that same period, revenue moved from 57,859.0 million $ to 58,757.7 million $, then 59,127.0 million $, then 61,809.3 million $. That is not the profile of a foundry being rescued by surging end demand; it is the profile of a company regaining some margin control in a still-limited demand environment. The next quarter in the history already shows why the thesis must be about a defensible band rather than a straight-line recovery: Q1 FY2026 revenue is 61,037.9 million $, down -1.2% QoQ but up +5.5% YoY, with gross margin at 29.2%. The margin did not hold at 30.7%, but it also did not revisit 26.7%, which is the more relevant test for a mature-node foundry in an oversupplied market.

The reason the margin band has a floor is that the mix is moving toward nodes where UMC still has leverage. The call’s most important product detail was Jason Wang’s statement that “With the 22- and 28-nanometer segment, 22-nanometer's revenue increased 31% quarter-on-quarter to a record high, accounting for more than 13% of total fourth quarter revenue.” That matters because total 22 and 28 nanometers revenue in Q4 FY2025 represented 36% of the total revenue pool, while full-year 22 and 28 nanometers revenue increased by 3 percentage points and 14-nanometer increased by 2 percentage point on a year-over-year comparison. The company is not merely filling legacy capacity with lower-value wafer starts; it is shifting enough of the revenue base into 22/28nm and 14nm to offset a utilization rate that stayed around 78%. This is why the EPS beat carries more signal than the revenue beat: it says the revenue dollars added in Q4 FY2025 were not all low-quality dollars.

That mix point also explains why the demand composition is less negative than a broad mature-node label implies. Consumer increased by 3 percentage points to 31% from 28% in the previous year, IDM accounted for 19% for the full year after increasing by 3 percentage points to 19% in 2025, and IDM was still about 20% of Q4 revenue. North America represented about 21% in Q4 of last year, while the sequential regional change came mainly from Asia and Europe. These figures complicate a simple smartphone or PC-cycle read. Consumer’s 31% share matters for MediaTek and Novatek Microelectronics because both are tied to UMC nodes in the data pack, with MediaTek using 28nm and 40nm for communication and consumer chips and Novatek using 28nm for OLED DDIC wafers. IDM’s 19% full-year share and 20% Q4 share matter for Texas Instruments, which uses analog and specialty wafers, because UMC’s mix stability suggests specialty analog demand is not being displaced by a pure consumer restocking pattern.

The customer read-through is therefore selective rather than uniformly bullish. For Qualcomm, which uses UMC for connectivity and IoT chips, the relevant signal is the Q4 revenue pool’s 36% exposure to 22 and 28 nanometers and the 22-nanometer revenue increase of 31% quarter-on-quarter; connectivity and IoT content at mature nodes can benefit from UMC’s record 22-nanometer contribution without needing a broad utilization recovery. For Lattice Semiconductor, the 28nm/40nm FPGA wafer exposure points to a foundry partner that is protecting gross margin at 30.7% even with utilization around 78%, which is more important for supply assurance than for immediate price relief. For Faraday Technology, UMC’s subsidiary tied to ASIC wafers, the read-through is that custom silicon demand can ride the same mix migration, particularly as 14-nanometer gained 2 percentage point on a year-over-year comparison. For TSMC, the specific supply-chain linkage is LSI bridge chip fabrication for CoWoS-L advanced packaging; UMC’s 2026 capacity increase of around 1.2% year-over-year limits the chance that this becomes a volume flood, but the 22/28nm and 14nm mix data suggest UMC remains relevant in specialty steps around advanced packaging ecosystems.

The supplier implications are equally constrained by the capacity plan, which is exactly why the market should not extrapolate this into a broad equipment or materials boom. Capacity remained flat quarter-over-quarter and will decline by roughly 1% due to the annual maintenance schedule, while the 2026 capacity increase will be around 1.2% year-over-year. CapEx is around USD 1.5 billion for 2026 versus USD 1.6 billion in 2025, but because the essay must keep company money figures in TWD and the data pack gives these only in USD, the usable conclusion is directional from the verbatim values: the plan is lower, not higher. That means localized consumables and maintenance-linked suppliers have a cleaner read than expansion equipment. Everlight Chemical and San Fu Chemical map to mature-node lithography and wet-process materials, so a Q4 revenue base of 61,809.3 million $ and Q1 FY2026 revenue of 61,037.9 million $ support ongoing fab consumption, not a step-change. Mirle Automation, tied to fab AMHS, and Kinik Company, tied to CMP diamond disc conditioners and grinding wheels, get less of an expansion signal because capacity is only around 1.2% year-over-year. Photronics and Topco Scientific have a steadier mix read because 22 and 28 nanometers were 36% of Q4 revenue and 14-nanometer increased by 2 percentage point year-over-year.

The peer context sharpens the point: UMC is not trying to be TSMC, and the valuation argument should not require it. In the latest foundry peer table, TSM has revenue of 1,134,103.0 million $, gross margin of 66.2%, and revenue YoY of +35.1%, while UMC has revenue of 61,037.9 million $, gross margin of 29.2%, and revenue YoY of +5.5%. That gap is structural, not a quarterly issue. The more relevant comparison is that GFS shows gross margin of 27.6% with revenue YoY of +3.1%, and TSEM shows gross margin of 26.8% with revenue YoY of +15.5%. UMC’s latest 29.2% gross margin and +5.5% revenue YoY sit in a different trade-off: slower than high-growth specialty foundry names in the table, but with a gross-margin level above GFS and TSEM. The market may still value UMC as a mature-node price taker; the peer data support a narrower claim, that UMC is a mature-node operator with enough mix control to keep gross margin near the high-20s to low-30s despite limited capacity growth.

That said, the call delivery does not read like management is trying to sell a sharp 2026 acceleration, which is a feature of the thesis rather than a defect. The tone history shows Q4 FY2025 sentiment at 0.22, unchanged from Q3 FY2025, while guidance_tone declined to 0.28 from 0.31. Prepared_sentiment was 0.05, barely above Q3 FY2025 at 0.04, and qa_sentiment slipped to 0.27 from 0.28. Those are not promotional numbers. The constructive part is that tone_confidence rose to 0.61 from 0.50, ai_optimism rose to 0.42 from 0.28, uncertainty fell to 68.2 from 112.0, and qa_evasiveness fell to 69.6 from 114.3. In other words, management sounded less expansive on guidance but less uncertain and less evasive in Q&A. That combination fits a company with a better handle on a modest environment, not a company signaling a demand breakout.

The phrasing around capacity and utilization confirms that management is drawing a boundary around expectations. Jason Wang said, “Capacity utilization rate will be in the mid-70% range.” That wording matters because it takes the Q4 around 78% utilization and points investors to a lower utilization zone, not an upward inflection, even as revenue in Q1 FY2026 remains 61,037.9 million $ and gross margin remains 29.2%. The apparent conflict is that Q4 gross margin improved to 30.7% with utilization around 78%, but the following quarter has gross margin at 29.2% and capacity utilization guided in the mid-70% range. The reconciliation is mix and maintenance: capacity will decline by roughly 1% due to the annual maintenance schedule, while 22nm had just produced a 31% quarter-on-quarter revenue increase to more than 13% of total fourth quarter revenue. If that 22nm mix does not keep contributing at a record level, the margin band compresses quickly.

The income statement also contains one caution that should keep PMs from overpaying for the beat. Diluted EPS in the quarterly history was 5.95 in Q3 FY2025 and 4.05 in Q4 FY2025, even though revenue rose +4.5% QoQ and gross margin improved to 30.7% from 29.8%. On the company’s call basis, Chi-Tung Liu said net income attributable to the shareholder of the parent was TWD 10.06 billion and earnings per ordinary shares were TWD 0.81; elsewhere on the call he put net income at around TWD 10.05 billion or EPS of TWD 0.81 in Q4 of 2025. These call-basis figures should not be mixed with the street-comparison EPS of 0.13, but they do show that the earnings narrative is not a simple function of revenue and gross margin alone. For the full year, revenue grew by 2.3% to TWD 237.5 billion, gross margin rate was around 29% or TWD 68.9 billion, net income attributable to the shareholder of the parent was around TWD 41.7 billion or 17.6% net income rate, and EPS for 2025 was TWD 3.34, down from TWD 3.8 in 2024. The balance sheet gives room to manage through the cycle, with cash amounts still more than TWD 110 billion and total equity of TWD 379.8 billion at the end of 2025, but it does not change the fact that full-year EPS declined while revenue rose.

The investment conclusion is to own the restraint, not chase an imagined ramp. UMC beat because the street underestimated earnings resilience, not because the company proved a new mature-node demand cycle. The numbers that make the thesis defensible are EPS surprise of +8.3%, revenue surprise of +1.8%, Q4 revenue growth of +4.5% QoQ, gross margin of 30.7%, utilization around 78%, 22 and 28 nanometers at 36% of Q4 revenue, and 22-nanometer revenue up 31% quarter-on-quarter to more than 13% of total fourth quarter revenue. The numbers that keep it from becoming a broad bull case are Q1 FY2026 revenue down -1.2% QoQ, Q1 FY2026 gross margin at 29.2%, capacity utilization expected in the mid-70% range, and 2026 capacity increase around 1.2% year-over-year. That is a narrow but investable setup: margin protected by mix and supply discipline, with upside if 22nm strength persists, and downside if utilization deterioration overwhelms the mix gains.

What to watch next quarter is concrete. The thesis is confirmed if Q1 FY2026 revenue holds near 61,037.9 million $, gross margin stays near 29.2% rather than sliding back toward 26.7%, and management can keep 22 and 28 nanometers close to the Q4 FY2025 level of 36% of total revenue while 22-nanometer remains above 13% of total fourth quarter revenue as the last disclosed benchmark. It is strengthened if utilization in the mid-70% range does not prevent gross margin from staying within reach of 30.7%, because that would prove mix can offset lower loading. It breaks if Q1 commentary on the next call after 2026-01-28 points to utilization below the mid-70% range, if 22nm no longer looks like the category that rose 31% quarter-on-quarter in Q4 FY2025, or if gross margin moves materially away from 29.2% toward the Q1 FY2025 trough of 26.7%. The debate next quarter should not be whether UMC can match TSM’s 66.2% gross margin or +35.1% revenue YoY; it should be whether UMC can keep a 29.2% to 30.7% gross-margin band while capacity grows only around 1.2% year-over-year and the street stops treating sub-80% utilization as automatic price erosion.