ticker stocksemiconductors updated 2026-05-30

UMC — United Microelectronics Corporation

Thesis

Stance: Medium conviction. WATCH → STAGE IN ON PULLBACK. Not a buy-now idea. UMC is a capital-disciplined, mature-node foundry trading at a trough 8.8x EV/EBITDA after a three-year earnings compression cycle. The core case is value + income, not growth: a mid-single-digit-growth foundry with a 5.6% FCF yield and a 3–4% dividend yield priced as if earnings keep falling. The single most important balance-sheet development is the 2025 FCF normalization — FCF recovered from NT$5.3B (FY2024) to NT$52.1B (FY2025), driven by capex discipline (capex halved from NT$88.5B to NT$47.7B), not by earnings growth.

Bull case: 22/28nm secular demand (EVs, IoT, Wi-Fi 7, connectivity) drives utilization recovery; Singapore Phase 3 adds premium-priced automotive 22nm capacity from H2 2026; the Intel 12nm JDA (production 2027) is a quasi-asset-light US manufacturing beachhead; a ~10% H2 2026 ASP hike lifts gross margin off the ~29% trough toward 33–35%; the Qualcomm advanced-packaging win signals a move up the value chain; and the GlobalFoundries merger rumor ("Project Ultron") is unpriced optionality. At 8.8x EV/EBITDA and 5.6% FCF yield the stock prices in very little of this.

Bear case: SMIC's state-subsidized 28nm expansion permanently erodes commodity mature-node ASP and margin — SMIC cut 28nm from ~$2,500/wafer to ~$1,500/wafer and surpassed UMC in foundry market share (Q4 2024: SMIC ~5.5% vs UMC ~4.7%), a structural not cyclical shift; TWD appreciation structurally compresses USD-priced earnings reported in TWD (the Goldman Sachs Sell thesis); gross margins peaked at 45.1% (2022) and likely cap at ~35% even in a good cycle; there is no leading-edge / EUV pathway; and the Intel 12nm JDA produces no revenue before 2027 and could stall if Intel narrows its foundry ambitions.

What has to be true: the 22/28nm mix must keep rising (37% of FY2025 wafer revenue, record high) and gross margin must inflect above ~32% by H2 2026 — confirming SMIC pricing pressure is manageable rather than structural. The FundamentEdge read at evaluation: Gates 2 (second derivative), 3 (not-just-valuation), 4 (quality/moat, conditional) PASS; Gates 1 (growth primacy) and 5 (estimate revisions) FAIL. The timing problem at the April 2026 evaluation: the stock ran ~40% in four weeks (Mar 26 → Apr 20), trading above the ~$10.22 analyst consensus PT and near 52-week highs — FOMO is the primary behavioral trap. Target entry: $11.00–$11.30 (Tranche 1, 25–30%); size 1–2% of portfolio, tranche in.

Snapshot

One-liner: World's fourth-largest pure-play semiconductor foundry (~4.4% market share), specializing in mature nodes 22nm–90nm with no EUV or sub-14nm capability — the "workhorse foundry" for Wi-Fi chips, display drivers, power-management ICs, and automotive silicon.

Ticker / exchange: NYSE: UMC (ADR; 1 ADR = 2 ordinary shares) | TWSE: 2303 Sector / GICS: Information Technology — Semiconductors & Semiconductor Equipment HQ: Hsinchu Science Park, Taiwan | Founded: 1980 (Taiwan's first semiconductor company; pivoted to pure-play foundry 1995) | Website: umc.com

Valuation snapshot (as of April 2026):

Market cap ~$29.8B USD (~NT$955B); enterprise value ~$28.2B USD
P/E (TTM) 22.4x | Forward P/E 19.4x | EV/EBITDA 8.8x (trough) | EV/Revenue 3.7x | P/FCF ~18x
FCF yield 5.6% | Dividend yield ~3.0–3.8% | Short interest 1.3% of shares outstanding
52-week range (USD ADR): $5.71–$12.68 (checklist cites $6.56–$12.82 as the at-evaluation 52-wk range)
Shares outstanding ~12.59B ordinary (TWSE); ~2.52B on ADR-equivalent basis
Deep-dive cites current price ~$11.80 (April 2026); checklist evaluation price ~$12.33–$12.70

Key stats: FY2025 revenue NT$237.6B (~$7.9B USD); FY2025 gross margin 29.0%; FY2025 FCF NT$52.1B; FY2025 ROIC ~9.8%; 12 fabs; 400,000+ wpm capacity (12-inch equivalent); Q4 2025 utilization ~78%; 2026 capex guidance US$1.5B.

Business

Model: Pure-play contract foundry. Customers supply the chip design (GDSII tape-out); UMC supplies the manufacturing process, lithography, and physical fab, charging a per-wafer fee. Revenue = (wafers shipped) × (blended ASP). No subscriptions, no licensing, no product-IP revenue. Margins are purely operational — the spread between wafer selling price and the cost of running a fab (labor, chemicals, electricity, depreciation on billions in equipment). Utilization rate is the single biggest lever: a fab at 70% vs 80% utilization can swing gross margin 5–8 percentage points; below ~70% a fab is cash-flow negative, above ~85% it is highly profitable.

Node mix (FY2025): 22/28nm = 37% of wafer revenue (record high, up from ~34% in 2024); 40nm and below combined = 53%; legacy nodes >40nm = remainder. Specialty within 22/28nm: UMC's 28eHV+ and 22nm eHV (embedded high-voltage, up to 100V+ on the same die as logic) — UMC claims to be the first foundry to launch a 22nm eHV platform (June 2024), essential for display drivers, creating real PDK switching cost.

End markets (~FY2024): Communications (Wi-Fi ICs, LCD controllers, ISPs, RF/baseband) ~40%; Consumer/IoT (display drivers, MCUs, wearables, NAND controllers) ~25%; Computer/data processing (storage controllers, PC peripheral chipsets) ~20%; Automotive (ADAS sensors, EV power management, infotainment) ~10%; Industrial/other ~5%. (Estimated from quarterly color; UMC does not publish a clean end-market table.)

Customer type mix: Fabless ~81%; IDMs ~18%; other ~1%. Geographic revenue (FY2024): Asia-Pacific 61%, North America 25%, Europe 11%, Japan 3%.

Key customers (industry estimates — UMC does not disclose customer revenue %): MediaTek (2454.TW, ~10–15%; deep-dive cites ~12–15%; spun off from UMC in 1997; the anchor customer — premium Dimensity SoCs run at TSMC 4/5nm but Wi-Fi/Bluetooth/display lines stay at 22/28nm, a demand floor); Texas Instruments (TXN, ~8–12%, IDM analog/mixed-signal); Qualcomm (QCOM, ~5–8%, new advanced-packaging win); Broadcom (AVGO, ~5–8%, networking/connectivity); Novatek (3034.TW), Realtek (2379.TW), Himax (~8–12% combined, display driver ICs / 22nm eHV users); Intel (INTC, pre-revenue 12nm partner). No single customer confirmed above 15% — more diversified than TSMC's ~25% Apple concentration.

Process portfolio depth (the durable moat): 40–90nm is a high-margin specialty business in disguise — RF-SOI (RF switches in every smartphone, competes with GFS), embedded flash/eNVM (automotive MCUs, smart meters, secure elements; qualified at 55nm/40nm), high-voltage (EV BMS, gate drivers at 60–250nm), mixed-signal CMOS (ADCs/DACs). Legacy nodes >90nm (~10%) are very low growth but extremely sticky — once qualified at 0.18µm, automotive/industrial customers stay 10–15 years.

Moat (modest but real): (1) 22nm eHV specialty — UMC claims the only volume 22nm embedded-high-voltage platform globally (TSMC de-prioritizes; SMIC has not qualified automotive-grade eHV); (2) automotive qualification (IATF16949 at the USJC Japan fab; qualifying a new fab takes 2–3 years); (3) PDK lock-in — re-qualifying a chip at a new foundry takes 12–18 months (18–24 for specialty); (4) geographic diversification (Taiwan + Singapore + Japan + China) no other non-TSMC foundry matches at this node range; (5) scale at 400,000+ wpm. Moat limits: no EUV, no advanced node ≤14nm in volume without Intel; cannot serve AI accelerators, HBM, or leading-edge logic. The AI boom is largely a TSMC story; UMC benefits only indirectly as TSMC capacity fills with AI and pushes some mid-tier work down.

Fabs (12 fabs; 400,000+ wpm 12-inch equivalent):

Fab 12A — Tainan, Taiwan — 14nm–90nm flagship — ~87,000 wpm
Fab 8A/8D/8F — Hsinchu, Taiwan — 40nm–0.5µm — various (~120,000 combined)
Fab 12i P2 — Pasir Ris, Singapore — 28nm–65nm — ~50,000 wpm
Fab 12i Phase 3 (new) — Pasir Ris, Singapore — 22nm — up to 30,000 wpm — volume production H2 2026; grand opening April 2025; ~US$5B total fab investment; ~700 new jobs; automotive/specialty focus
USJC / Fab 12M — Mie Prefecture, Japan — 40nm–90nm automotive-grade — ~35,000 wpm — 100% owned since Oct 2019
United Semi / Fab 12X — Xiamen, China — 40nm–90nm — up to 50,000 wpm design cap
Intel Ocotillo (JDA, Chandler AZ, Fabs 12/22/32) — 12nm (Intel 12) — TBD capacity — development; customer pilot 2026, production 2027

Value chain: UMC sits at the highest-capital, highest-barrier fabrication layer but captures no design IP value. Key suppliers: ASML (DUV litho, no bypass, ~$350B mkt cap, ~12x UMC), KLA (metrology, no bypass), Applied Materials/Lam Research/Tokyo Electron (deposition/etch, partial bypass), Shin-Etsu/Sumco (300mm silicon wafer duopoly, partial bypass — the one input-cost risk worth monitoring). No upstream supplier is a small-cap underpriced secondary play; all are large-cap and fully valued.

Financials

All NT$ figures in billions TWD; USD equivalents at ~32 TWD/USD (the FX rate is itself a material earnings risk).

Income statement & margins (the gross-margin collapse and the recovery setup):

Metric	FY2022	FY2023	FY2024	FY2025	FY2026E
Revenue (NT$B)	278.7	222.5	232.3	237.6	~248–255 (est.)
Revenue growth YoY	+26%	-20.2%	+4.4%	+2.3%	+4–7%
Gross profit (NT$B)	125.8	77.7	75.7	68.9	~75–80
Gross margin	45.1%	34.9%	32.6%	29.0%	30–32% (guidance: high-20s in Q1)
Operating income (NT$B)	104.3	57.9	51.6	43.9	~50–58
Operating margin	37.4%	26.0%	22.2%	18.5%	~20–23%
Net income (NT$B)	89.5	59.7	47.2	41.7	~44–50
Net margin	32.1%	26.8%	20.3%	17.6%	~18–20%
EPS (TWD, basic)	NT$37.0	NT$24.60	NT$19.00	NT$16.70	~NT$18–20

The FY2022 peak (NT$278.7B) was a COVID-era boom that pulled forward 2–3 years of demand; 2023–2025 is the hangover. Gross margin fell 45.1% → 34.9% → 32.6% → 29.0% on three causes: utilization decline (~90%+ to ~70%), depreciation surge from Singapore Phase 3 capex (high-20% D&A growth in 2025), and TWD appreciation (~3pp headwind in Q2 2025 alone). FY2025 EPS NT$3.34 (NT$16.70 ÷ 5 ADR ratio) missed consensus but revenue beat; FY2025 wafer shipments +12.3% YoY; revenue in USD +5.3% YoY.

Cash flow & balance sheet (the FCF story):

Metric	FY2022	FY2023	FY2024	FY2025	FY2026E
Operating cash flow	145.9	86.0	93.9	99.9	~100–110
Capex	-80.1	-91.5	-88.5	-47.7	~-47 (US$1.5B)
Free cash flow	65.7	-5.5	5.3	52.1	~53–63
FCF margin	23.6%	-2.5%	2.3%	21.9%	~21–25%
Cash & equivalents	173.8	132.6	105.0	110.7	N/A
Total debt	53.2	80.2	81.5	79.0	N/A
Net debt	125.4	64.7	33.8	49.4	N/A
Net debt / EBITDA	~0.7x	~0.7x	~0.4x	~0.6x	N/A
ROIC	~25%+	~15%	~12%	~9.8%	~10–12%

The FCF recovery in 2025 (NT$52.1B vs NT$5.3B in 2024) is the most significant development — driven by capex discipline (halved YoY), not earnings growth. Net debt ticked up slightly in 2025 vs 2024 despite strong FCF, due to dividends and Singapore fab spending. Net debt of NT$49.4B is modest vs ~NT$100B operating cash flow; the balance sheet is clean — no covenant risk, no near-term refinancing issues. Normalized maintenance capex is ~NT$40–50B/yr, giving a non-expansion-year FCF ceiling of ~NT$50–60B. ROIC ~9.8% (FY2025) sits roughly at the ~10–12% WACC — a trough indicator, not a structural problem; if gross margin recovers to 33–35% ROIC re-crosses WACC.

Quarterly revenue and the second-derivative turn (NT$B):

	Q1'24	Q2'24	Q3'24	Q4'24	Q1'25	Q2'25	Q3'25	Q4'25
Revenue	54.63	56.80	60.49	60.39	57.86	58.76	59.13	61.81
QoQ %	—	+4.0%	+6.5%	-0.2%	-4.2%	+1.6%	+0.6%	+4.5%
YoY % (est.)	—	—	—	—	+5.9%	+3.5%	-2.2%	+2.3%
Gross margin	~35.5%	~33.8%	~33.8%	~30.4%	~26.7%	~28.7%	~29.8%	~30.7%

Growth decelerated through Q3 2025 (-2.2% YoY) then recovered in Q4 2025 (+2.3%). Q1 2026 net sales were NT$61.04B, +5.49% YoY — solidly in line with flat-shipment guidance; March 2026 monthly revenue +4.9% YoY. Q4 2025 revenue NT$61.81B annualized ≈ NT$247B exit rate.

Incremental margin analysis (the inflection signal): YoY incremental gross margins were deeply negative Q1–Q3 2025 (UMC losing margin dollar-for-dollar as revenue moved modestly — the Singapore depreciation effect, D&A rising ~high-20% YoY before capacity is utilized), then turned positive in Q4 2025 (+42% incremental GM on a modest revenue increase). The valley floor appears to have been Q1 2025 at 26.7% gross margin. At normalized 80%+ utilization UMC should generate ~$0.35–0.40 of incremental EBIT per dollar of incremental revenue, consistent with the 35–38% gross-margin ceiling and ~18–22% operating margins. Dilution: stable share count over 5 years, no material ATM/shelf/convertibles; self-funding — low dilution risk.

Industry landscape

Semiconductor foundry market ~$130–145B (2025E); mature-node segment (28nm+, ex-leading-edge) ~$55–65B; UMC SAM (22–90nm addressable) ~$30–40B; UMC ~$7.9B 2025 revenue ≈ ~20–25% of SAM, ~4.4% of total foundry. The industry is consolidated at the top (TSMC ~70%) and fragmented in the specialty tier. Foundry market share (Q2 2025, TrendForce): TSMC ~70%, Samsung ~7%, SMIC ~5%, UMC ~4.4%, GlobalFoundries ~4%, others ~10% (PSMC, Tower/TSEM, HHGrace). The defining structural shift: SMIC's state-subsidized 28nm expansion surpassed UMC in market share (Q4 2024: SMIC ~5.5% vs UMC ~4.7%; SMIC ~$9.36B 2025 revenue); Chinese foundries (SMIC, Hua Hong, HHGrace, Nexchip) are estimated at >25% of global mature-node capacity by end-2025. UMC and GFS are the "endangered middle" — too small to lead-edge compete, too valuable to disappear, vulnerable to SMIC commodity pricing. Cycle position (April 2026): mid-cycle recovery, not boom — utilization improved 70% → 78%, Q1 2026 guided mid-70%; oversupplied at commodity 28nm, tight at specialty 22nm eHV and automotive 40nm. Secular tailwinds: EV silicon content (a Tesla Model Y uses ~6,000 chips vs ~1,500 in a mid-range ICE car), Wi-Fi 7 ramp (22/28nm, volume 2025–2027), China+1 diversification, US CHIPS Act reshoring.

See sector page: semiconductors

Management

Overall management grade: B+ — competent, conservative, long-tenure, capital-disciplined team with a strong strategic track record; the primary gap is low insider equity ownership.

Leadership:

Stan Hung (洪嘉聰) — Chairman & Chief Strategy Officer (Chairman since 2008; joined UMC 1991, 35 years). Tam Kang University accounting graduate; rose through CFO and SVP Finance. The architect of UMC's defining 2012 strategic pivot — exiting the advanced-node R&D race at 28nm FinFET and below, ceding leading-edge to TSMC rather than spending tens of billions on EUV. Controversial then, vindicated since: it preserved UMC as a profitable independent foundry while Samsung/GFS retreated or absorbed sustained losses. UMC's founder is Robert Tsao (retired 2006); Hung is a company lifer, not a founder.
Jason Wang (王石) — CEO (effective February 25, 2026; Co-President from 2017; joined UMC 2008). San Jose State University; before UMC spent ~18 years at Trident Microsystems (VP Business Operations and Finance, Asia), credited with a 2003 PC-Graphics-to-Digital-Media pivot adding ~$2B enterprise value by 2006. Trident filed Chapter 11 in January 2012 — four years after Wang left; not attributable to him, but noted. At UMC: VP Corporate Marketing (2008), President UMC USA (2009–2014), SVP (2014–2017), Co-President (2017). Personally championed the Intel 12nm JDA. Sole-CEO accountability is new — the next 4–6 quarters are the validation window.
Ming Hsu — President & COO (effective February 25, 2026; added to Board). EVP prior; operational/manufacturing background; provides manufacturing continuity. Limited public profile — a less-known quantity.
Chitung Liu (劉啟東) — CFO & SVP, Head of Corporate Governance. Long-tenure; steady, credible communicator; managed the Singapore capex cycle and the 2025 discipline turn.
S.C. Chien (簡山傑) — former Co-President; departed Feb 25, 2026 to become Chairman of Unimicron Technology (3037.TW), an IC-substrate/PCB affiliate in the UMC group. Joined UMC 1989 (37 years); held 176 global patents; led technology/R&D/manufacturing. His move is to an affiliated, complementary (not competing) supply-chain layer — a manageable related-party dynamic, no self-dealing flag.

February 2026 restructuring: UMC formally ended its co-president model on Feb 25, 2026; Jason Wang named sole CEO, Ming Hsu elevated to President/COO and added to the board, framed as enhancing "decision-making efficiency." Stock fell ~3% on announcement — read as transition uncertainty, not a clear bullish signal. Net governance positive (clearer accountability).

Ownership & alignment (the standing structural concern): Aggregate insider ownership ~1.31% — low, typical of large Taiwanese corporates. Management is NOT "eating its own cooking" in a significant way; alignment runs through tenure and reputational capital, not equity concentration. US ADR institutional ownership ~5.98% (BlackRock, Vanguard — passive/index); ~92–93% widely held public float; bulk of economic ownership is Taiwanese institutional/retail on TWSE-2303. No controlling shareholder; activist risk low.

New US disclosure regime (effective March 18, 2026): the "Holding Foreign Insiders Accountable Act" (enacted Dec 18, 2025) now requires UMC directors/officers to file SEC Section 16(a) reports; Form 3 initial statements filed March 18, 2026 — the first time precise individual holdings appear in EDGAR (a governance improvement). Verified holdings: CFO Chitung Liu 6,162,917 total (4,012,917 direct + 150,000 spouse + 2,000,000 CTBC Bank Trust), incl. 1,210,000 RSAs vesting Dec 5 annually 2026–2029; Oliver Chang (SVP) 3,549,289; Jerry CJ Hu (VP) 2,395,280; Eric Chen (VP) 1,514,280; Francia Hsu (VP) 731,280; JT Lin (VP) 515,060; Wu Ling-Ling (Director) 0. Stan Hung and Jason Wang holdings not yet fully detailed in EDGAR (Hung est. 15–20M from 20-F). March 2026 insider selling: all six disclosed executives were net sellers — Liu sold 600,000 shares (~$5.1M at ~$8.50/share, a ~10% position cut), plus VP sales of 9,000–45,000 shares; no insider purchases in March or the prior 12 months. Interpretation: most likely routine estate/tax-planning sales newly surfacing under US disclosure (Q1 2026 revenue +5.49% YoY argues against a business-condition read), not conclusive — warrants ongoing monitoring. No 10b5-1 plans disclosed.

Governance: Board 9 members, 6 independent (two-thirds majority, Taiwan FSC compliant); three seats reserved for female directors; committees = Audit, Remuneration, Sustainable Development/Nominating; 2024 board self-evaluation rated "Excellent." No dual-class shares (one-share-one-vote), no poison pill; staggered 3-year terms (Taiwan norm). Clean litigation/enforcement record for the current team — no SEC actions, no restatements, no auditor resignations; historical TSMC trade-secret disputes (early 2000s) predate this management by 15+ years and were settled.

Capital allocation (grade B+): Standout decisions — the 2012 advanced-node exit (preserved ~$20B+ capex) and the 2025 capex discipline (halved to NT$47.7B, FCF NT$5B → NT$52B in one year). USJC Japan full acquisition (2019, good); USCXM Xiamen full buyout (CNY4.59B / ~US$695M, 2022–2023, neutral — consolidated a JV, timing tied to geopolitical risk management, arguably slight overpay for a geopolitically exposed asset). Dividends are the primary return vehicle (~28.8% 5-yr CAGR, funded by FCF, no cuts). The one gap: no buyback program despite low-valuation troughs — the 2022 window (P/E ~8x, TECC ~12.5%) was the ideal missed buyback opportunity; capital-allocation timing is Neutral (not value-destructive, not Malone/Buffett-grade). Compensation yellow flag: likely tied to revenue/EPS rather than ROIC/FCF (Taiwan 20-F aggregate-disclosure format limits hurdle visibility); SBC is low, dilution not a concern; no outsized change-of-control packages in the Feb 2026 reorg.

Corporate structure (clean): USCXM (Xiamen, 100% owned since July 2023 — was a ~35% JV with Chinese state partners until the 2022–2023 buyout), USJC (Japan, 100% since Oct 2019), Singapore Fab 12i P2/P3 (UMC-operated), plus a minority/indirect stake in Unimicron (3037.TW) and sales subs (Japan/Korea/Europe/USA). No shell SPVs, no IP shuffling, no concerning related-party transactions. (Note: a July 2025 director resignation at USCXM — Hsu Chih-Ching as Director/Vice Chairman — was an internal governance change, not a fraud/regulatory finding; new board member elected Aug 2025.)

Catalysts & risks

Near-term catalysts (0–12 months):

Q1 2026 earnings (late April 2026) — first read under sole-CEO Jason Wang; the key signal is Q2 2026 guidance (utilization toward 80%+ vs stall in mid-70s)
~10% H2 2026 ASP hike (announced via TrendForce April 17, 2026; pricing tied to mix/capacity/long-term deals) — a management action not yet in consensus estimates
Qualcomm advanced-packaging mass production (Q1 2026) — silicon interposer / WoW (Wafer-on-Wafer) hybrid bonding with 1500nF/mm² capacitors for Qualcomm HPC (Oryon) chips; won Dec 2024, breaks TSMC's near-monopoly on Qualcomm packaging; margin-accretive new revenue stream
Singapore Phase 3 ramp confirmation (H2 2026) — first customer orders for new 22nm automotive capacity (~30,000 wpm)
GFS "Project Ultron" merger news — any confirmation/denial; a confirmed merger would be a significant positive re-rating
Intel Foundry Direct Connect 2026 — 12nm milestone / customer tape-out updates

Medium-term catalysts (1–3 years):

Intel 12nm commercial production (2027) — first revenue-bearing milestone; volume depends on CHIPS Act-driven and defense/aerospace pull for US-made 12nm
UMC-Intel 6nm exploration (reported July 2025) — if formalized in 2026 would re-enter mid-range advanced nodes (Wi-Fi 7, Bluetooth, automotive radar) without funding a ~$30B EUV fab; exploratory only, no formal announcement
GFS-UMC merger if consummated — ~$37B combined entity, ~8–9% global foundry share; 18–36 month process; US national-security/CFIUS-equivalent review is a risk
Gross-margin recovery to 33–35% — requires utilization 82%+, specialty mix >40% of revenue, stable ASP; achievable by 2027 in base case

Risks (bear case):

SMIC structural 28nm pricing pressure (High likelihood): SMIC cut 28nm from ~$2,500 to ~$1,500/wafer; structural not cyclical; 28nm is ~10–15% of UMC wafer revenue. If 28nm ASP deflation spreads to 22nm, the gross-margin recovery thesis breaks. Mitigant: 22nm eHV specialty and automotive qualification SMIC cannot yet match.
TWD appreciation (Medium-High): UMC prices wafers in USD but reports in TWD; TWD strengthening from ~32 to <30 cuts USD-equivalent TWD earnings ~6.5% purely on FX (~10% TWD appreciation = ~2–3pp blended-margin compression). This is the Goldman Sachs Sell thesis; structural, driven by Taiwan's current-account surplus; only manageable, not eliminable.
Intel foundry strategy reversal (Medium): Intel Foundry lost $18.9B in FY2024; if Intel exits external foundry or pauses 12nm, UMC loses a differentiated demand anchor and the P/E-expansion narrative. Mitigant: the JDA is process development, not take-or-pay; UMC owns no Arizona capex; core revenue unaffected.
Taiwan-China geopolitical disruption (Medium): Singapore + Japan ≈ 25–30% of capacity partially mitigate, growing US exposure via Intel; never fully eliminable while Taiwan is the primary base.
Margin recovery delayed / gross-margin ceiling ~32–35% without a leading-edge pathway.
GFS merger integration risk (Low-Medium): only relevant if it proceeds.

What would make the thesis wrong: gross margin stays below ~29% in H2 2026 despite the announced price hike (customers push back or redirect to SMIC); Intel announces a 12nm JDA review/pause; TWD appreciates above 30/USD; SMIC achieves automotive-grade qualification by 2027 (erodes the specialty moat). Downside scenario: $7–9 per ADR (28nm ASP war escalates, gross margin troughs 25–27%, Intel JDA delayed, TWD headwinds — a 2023 re-run with no recovery catalyst). Portfolio note: high correlation with TSMC, Samsung, and the broad semiconductor sector — adding to existing TSMC exposure adds correlated risk; UMC offers Taiwan geographic + mature-node diversification vs TSMC's leading-edge focus.

Valuation / DCF

Versus 5-year history: UMC has traded ~6x–15x EV/EBITDA across the cycle; at 8.8x it sits at the lower end but not the absolute bottom (2023 trough was nearer 7x). At ~$12.50 the stock had already re-rated ~40% off the March 2026 low of ~$9.00 — no longer the bargain it was 30 days earlier.

Versus peers: UMC at 8.8x EV/EBITDA is the cheapest major foundry. TSMC ~16–19x (justified premium for EUV leadership, ~20%+ AI-driven growth, lower geopolitical risk per revenue dollar); GlobalFoundries ~9–12x (roughly in line, slightly richer); SMIC ~10–12x (China domestic, different risk profile). Peer comp table: TSMC ~19x EV/EBITDA, ~28x P/E, ~3% FCF yield, +20%+ growth; GFS ~10x, ~25x, ~4%, +4%; UMC 8.8x, ~22x, 5.6%, +5–7%.

Analyst sentiment: thin US ADR coverage (~5 analysts for a ~$30B company; primary coverage is Taiwan-side — KGI, Nomura/CICC/Morgan Stanley/JPMorgan Taiwan). Consensus 12-month PT ~$10.22 (range $8.60–$11.85); a separate average-PT read of ~$7.12–$8.17 (range $4.80–$10.60) sits notably below the trading price — either PT lag after the 2026 rally or a genuinely more bearish street. At ~$12.50 the stock traded ABOVE consensus PT — a meaningful caution signal for a new entry. Notable downgrades: BofA → Underperform (from Buy), PT NT$37 (down from NT$49); Goldman Sachs → Sell, PT NT$40.50 (TWD-appreciation thesis); BNP Paribas upgrade Underperform → Neutral, PT $8.60 (April 20, 2026, still below trading price). Overall consensus near "Reduce" (three sells, two holds, one strong buy per the at-evaluation data). Short interest 1.3%.

Simple DCF sanity checks (two framings, both preserved):

Checklist DCF: 5% revenue CAGR through 2030, margins recovering to 33–35% EBITDA by 2027, 65% FCF conversion, 10% discount rate → intrinsic value ~$10–13 per ADR (roughly fairly valued in base case at ~$12.50).
Deep-dive DCF: FY2026E FCF ~NT$55B (~$1.7B USD), 5% growth for 3 years → 3% terminal, ~11% WACC (geopolitical risk premium) → ~$14–16 per ADR, i.e. +15–35% upside from ~$11.80; wide range because +1% WACC cuts value ~10%.

Scenarios: Base case fair value ~$10–13 (deep-dive 12-month base target $13–15, +10–27% incl. dividend). Upside case ~$15–18 (GFS merger and/or 35%+ gross margin; bull exit at 12–14x forward P/E ≈ $15–18 on FY2027E EPS ~$1.33 ADR-adjusted). Downside case ~$7–9 (28nm ASP war, gross margin stuck 27–29%, SMIC wins 28nm, Intel JDA delayed; deep-dive bear PT $7–8 at 12% WACC, ~40% downside). At 19.4x forward P/E the market prices ~4–5% EPS CAGR; GFS optionality and the H2 2026 ASP hike are not fully in consensus models.

Discipline note: would NOT buy 10–15% higher (~$14): FCF yield drops to ~4.8% and forward P/E ~21.5x — paying for the catalyst rather than being paid to wait. Margin of safety is thin at ~$12.50; compelling again at $10.50–$11.50 (FCF yield ~6.5%, forward P/E ~17x).

Decision log

2026-04-26 — Profile + deep-dive initiated. No position. Conviction Medium. Monitoring Intel 12nm JDA progress, Singapore Phase 3 ramp, and Qualcomm packaging ramp as catalysts. Key watch: Q1 2026 earnings (late Apr 2026) — first read under new sole-CEO Jason Wang. GFS merger optionality is the upside-surprise scenario not in consensus. Suggested sizing 1–2%, tranche in. Exit triggers: gross margin <27% for 2 consecutive quarters; Intel JDA formal pause; TWD strengthens to <30/USD.
2026-04-26 (mgmt-dd) — Management grade B+ (unchanged). Refreshed with new Section 16(a) insider data. Key additions: CFO Liu holds 6.16M total shares; sold 600K (~$5.1M) in March 2026 — first month of mandatory US disclosure under the Holding Foreign Insiders Accountable Act. Five VPs also net sellers in March. Read as likely routine estate/tax-planning sales surfacing for the first time under the new US regime, not a business-condition signal (Q1 2026 revenue +5.49% YoY confirmed). Standing concern: ~1.31% aggregate insider ownership — alignment via tenure/reputation, not equity.
2026-04-26 (buy checklist) — Verdict: WATCH → STAGE IN ON PULLBACK. Evaluated at ~$12.50 ADR (at-evaluation 52-wk range $6.56–$12.82). Thesis valid (FCF recovery, H2 2026 ASP hike ~10%, Singapore Phase 3, Intel 12nm optionality, GFS merger upside). Key problem: stock ran +40% in four weeks (Mar 26–Apr 20), trading above the ~$10.22 analyst consensus PT, near 52-week highs — FOMO is the primary behavioral trap. FundamentEdge: Gates 2/3/4 pass; Gates 1 (growth primacy) and 5 (estimate revisions) fail. Technicals: do not chase the breakout; ideal entry was ~$9–10 four weeks earlier. Entry plan — Tranche 1 (25–30%) at $11.00–$11.30; Tranche 2 (30–40%) on Q1 earnings confirmation (Q2 2026 GM guide ≥29.5%, utilization ≥75%); Tranche 3 on Singapore Phase 3 volume or Intel 12nm PDK delivery. Do not buy above $13.50 (EV/EBITDA >10x, forward P/E >21x). Stop-think: gross margin <27% for 2 consecutive qtrs; Intel JDA suspended; TWD <30/USD; GFS merger definitively ruled out.
2026-05-30 — Consolidation. Five research fragments (profile, deep-dive, mgmt-dd, buy-checklist, prior canonical) merged into this single thesis-first page. No new research; stance unchanged: Medium conviction, WATCH → stage in on pullback toward $11.00–$11.30. Reconciliation flag: the deep-dive's claim that UMC owns ~35% of United Semi Xiamen is stale — superseded by the 100% acquisition (CNY4.59B/~US$695M) completed July 2023, corroborated in mgmt-dd and prior canonical; treated 100% ownership as truth.

Sources

Fragments folded into this canonical page (consolidation 2026-05-30):

umc-profile.md — Company Profile (skill: profile, 2026-04-26)
umc-deep-dive.md — Full Investment Deep-Dive (skill: deep-dive, 2026-04-26)
umc-mgmt-dd.md — Management Due Diligence (skill: mgmt-dd, 2026-04-26)
umc-buy-checklist.md — Pre-Buy Checklist (skill: checklist, 2026-04-26)
umc.md — prior canonical (date 2026-04-26, updated 2026-05-15)
(umc-filings.md empty placeholder — not merged.)

External sources cited across fragments: UMC IR press releases (BusinessWire — Q4 2025 results Jan 28 2026; Q4 2024 results Jan 2025); UMC Form 20-F for FY2024 (filed April 24, 2025); UMC 6-K filings (US insider disclosure rules March 18 2026; March 2026 insider shareholding changes + Q1 2026 investor call); UMC Form 3 (Wu Ling-Ling, March 18 2026); UMC management/leadership-change press releases (Feb 25 2026); UMC 22nm eHV launch (June 2024); UMC Singapore Phase 3 grand opening (April 2025); Intel-UMC Foundry Collaboration announcement (Intel Newsroom, Jan 2024); StockAnalysis.com; TrendForce (foundry market share Q2 2025; UMC H2 2026 price hike Apr 17 2026; UMC asks suppliers to cut prices 15% Oct 2025; Qualcomm advanced packaging Dec 2024; mature-node order drop 2H25; UMC mulls 6nm with Intel July 2025; Project Ultron / GFS-UMC merger Apr 2025); Nikkei Asia; Digitimes (SC Chien/Unimicron; Jason Wang CEO; UMC-Intel 12nm; USCXM buyout); Seeking Alpha (new CEO Feb 2026); Investing.com / Insider Monkey / Yahoo Finance (earnings-call transcripts); StockTitan (6-K / Form 3 filings); StocksToTrade (Q1 2026 +5.49% YoY); TradingKey (BNP Paribas upgrade Apr 20 2026); Benzinga (analyst consensus); HD In Research (SMIC surpasses GFS); Mordor Intelligence; mark lapedus / Mark Lapedus Substack; Trident Microsystems (Wikipedia, bankruptcy background).

Consolidation queue (merged 2026-05-30)

These fragment files were folded into this canonical page and stay live pending Pink's archive confirmation:

[ ] umc-profile.md
[ ] umc-mgmt-dd.md
[ ] umc-buy-checklist.md
[ ] umc-deep-dive.md
[ ] umc.md

Pages that link here (2)

Foundry & Semiconductor Materials Semiconductors (cross-cutting / geopolitics)

Open in Obsidian