Announcement

The three largest trailing‑edge foundries – GlobalFoundries (GloFo), United Microelectronics Corp. (UMC), and Semiconductor Manufacturing International Corp. (SMIC) – reported FY 2025 revenues of $6.79 bn, $7.63 bn, and $9.33 bn respectively. Together they account for roughly 13.5 % of global foundry sales, yet each follows a markedly different roadmap shaped by geography, policy, and technology choices.

---

Technical Specs

GlobalFoundries – Specialty FD‑SOI and GaN

• Node portfolio: 12 LP FinFET → 180 nm CMOS, plus FD‑SOI (22 nm) and GaN‑on‑Si (45 nm).
• Flagship process: 22FDX – 22 nm fully depleted silicon‑on‑insulator with embedded MRAM. Performance: 1.2 V operation, sub‑10 µA/µm static power, suitable for automotive radar and millimeter‑wave 5G.
• Key platforms: 45RFSOI (RF front‑end), 28 nm bulk logic, 40/55 nm BCDLite for power ICs, SiGe BiCMOS for high‑frequency analog, GaN‑on‑Si for power amplifiers.
• Capacity: Fab 8 (Malta, NY) – 300 mm, Trusted Foundry Category 1A; Fab 1 (Dresden) – expanding to 1.5 M wafers/yr by 2028; Fab 7 (Singapore) – 450 k wafers/yr; two 200 mm sites for GaN and legacy nodes.
• Capital spend: FY 2026 capex guided at 15‑20 % of revenue, driven by oversubscribed demand for 22FDX, silicon photonics, and SiGe.
• IP addition: Acquisitions of Advanced Micro Foundry (silicon photonics) and MIPS/Synopsys ARC (RISC‑V & AI inference IP) enable a bundled fab‑plus‑IP offering unique among mature‑node players.

UMC – 12 nm FinFET Co‑Development with Intel

• Node mix: 14 nm (14FFC) bulk logic, 22 nm and 28 nm HKMG for display drivers, 40‑65 nm analog/power, plus a 12 nm FinFET slated for 2027 mass production at Intel’s Chandler, AZ fab.
• 12 nm metrics (projected vs 14FFC): +10 % performance, –20 % power, –10 % die area, and three fewer mask layers, translating into a ~15 % cost advantage per wafer.
• Capacity: 12 fabs, >400 k 12‑inch‑equivalent wafers/month. Fab 12i Phase 3 (Singapore) – 30 k wafer starts/month, ready for 22/28 nm volume.
• Specialty processes: eHV (embedded high‑voltage) for OLED drivers, eFlash (350 nm‑28 nm), RFSOI, and BCD for mixed‑signal.
• Financials: Q1 2026 revenue NT$61.04 bn ($1.93 bn), net income up 108 %, gross margin 29.2 %, utilization 79 % (target low‑80 % Q2).

SMIC – High‑Volume 28 nm and DUV‑Based 7 nm‑Class

• Node spread: 350 nm‑7 nm; >90 % of output at 28 nm and above.
• Advanced nodes: N+2 (7 nm‑class) – DUV multi‑patterning, 20 k WSPM, ~60‑70 % yield; N+3 (7 nm/6 nm‑equivalent) – also DUV, limited to Huawei Kirin 9030.
• Capacity: Four new 12‑inch fabs (Shanghai, Shenzhen, Beijing, Tianjin) adding ~340 k wafer starts/month of 28 nm‑plus capacity.
• Utilization: 93.5 % average 2025, wafer shipments 9.7 M 8‑inch‑equiv.
• Capex: >$7 bn FY 2025, pushing gross margin down to 21 % (guidance 18‑20 % Q4 2025) as depreciation from new fabs weighs on profitability.
• Export constraints: On U.S. Entity List, no EUV tools, DUV and etch equipment increasingly restricted; Taiwan blacklist adds permit requirements for high‑tech shipments.

---

Market Implications

1. Supply‑Side Diversification – With TSMC occupying ~70 % of global foundry revenue, the three trailing‑edge players collectively provide a critical buffer for automotive, industrial, and defense silicon. GloFo’s FD‑SOI platform, for example, offers a low‑power alternative to bulk CMOS that is attractive for safety‑critical radar where leakage must be minimized.
2. Pricing Pressure and Floor – Mature‑node pricing fell sharply in 2024‑25 as Chinese capacity surged. TrendForce now reports price hikes of up to 10 % for 28‑55 nm wafers as TSMC reallocates 40‑90 nm capacity to AI‑centric CoWoS interposers. The floor appears to be stabilizing around $1,500‑$1,800 per 28 nm wafer.
3. Policy‑Driven Capital – GloFo’s $1.575 bn CHIPS Act award and $3.1 bn DoD contract tie its growth to U.S. defense spending, ensuring a steady pipeline of radar and secure‑element orders. SMIC’s expansion is state‑backed, insulating it from pure market cycles but leaving it vulnerable to export bans that limit advanced‑node yields.
4. Consolidation Signals – The “Project Ultron” internal memo that floated a GloFo‑UMC merger would create a ~28 % market share, but cross‑border regulatory hurdles (U.S., China, Taiwan) make such a deal unlikely in the near term. Still, the mere existence of the study indicates that margins are tightening and scale is becoming a survival factor.
5. Technology Trade‑offs – UMC’s 12 nm FinFET, built on Intel’s mature‑node platform, gives it a foothold in the low‑power, high‑performance segment without the cost of a full EUV line. However, the node still lags behind GloFo’s 22 nm FD‑SOI in terms of leakage, making the two processes complementary rather than directly competitive.
6. Future Capacity Outlook – By 2028, GloFo’s Dresden expansion and Malta capacity tripling will push its mature‑node output above 2 M wafers/yr. SMIC’s four new fabs will add ~340 k wafer starts/month, but export limits keep its 7 nm‑class capacity in the low‑tens‑of‑thousands range, far below the hundreds of thousands needed for a true high‑volume advanced‑node competitor.

---

Bottom Line

The trailing‑edge segment is not a monolith. GlobalFoundries leans on specialty FD‑SOI and GaN to capture automotive and defense niches, backed by U.S. policy money. UMC bets on a 12 nm FinFET partnership with Intel to bridge mature‑node services and low‑power logic, while expanding its dominant display‑driver market. SMIC pours state capital into massive 28 nm capacity and pushes DUV‑based 7 nm‑class nodes despite tightening export controls. Their differing strategies will shape where automotive, AI‑server power ICs, and high‑frequency RF components are sourced over the next five years.

---

Key figures

Company	FY 2025 Revenue	Global Share	Primary Growth Node
GlobalFoundries	$6.79 bn	3.9 %	22FDX FD‑SOI
UMC	$7.63 bn	4.4 %	12 nm FinFET (2027)
SMIC	$9.33 bn	5.3 %	28 nm & N+2 (7 nm‑class)

---

All financial figures are sourced from company SEC filings, quarterly reports, and TrendForce market data.