Semiconductor Manufacturing International Corp
Pure-play foundry: manufactures logic/mixed-signal/CIS/power/RF wafers for fabless customers on 0.35um-7nm-class nodes; revenue per wafer x utilization; capacity expansion co-funded by China state funds (Big Fund) via fab JVs
Earnings, margins, COGS & capex
FY2025 was a record year: revenue $9.33B (+16.2% YoY in USD) on domestic-substitution demand, 93.5% annual capacity utilization and improved mix; gross margin recovered to 21.0% (+3pp) though reported 7nm-class yield struggles and depreciation from $8.1B capex cap the ceiling (Q4 2025 GM fell to 19.2% on higher depreciation). Net profit attributable to owners was $685M (+39% YoY in USD; +36% in RMB) - thin for the revenue base because depreciation and JV minority interests absorb much of the P&L, and the FY2025 annual dividend was skipped to fund the buildout. Q1 2026 held roughly flat sequentially ($2.506B, GM 20.1%) with a strong Q2 2026 guide of +14-16% QoQ on improved pricing and product mix. The structural story is volume and sovereignty, not margin: capex near revenue means free cash flow stays negative through the buildout.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~79¢ is cost of goods and ~11¢ operating expense, leaving ~10¢ of operating profit (~7¢ net).
Revenue trend
Margins
recovering from the 2024 trough (~18%); capped by new-fab depreciation and reported 7nm-class yields (Q4 2025 dipped to 19.2%)
thin; depreciation-heavy
improving (+39% net profit YoY in USD) but structurally low vs TSMC ~40%+
stays negative while expansion capex ~$8B/yr continues; FY2025 dividend skipped
COGS structure
Dominated by depreciation from aggressive fab buildout (four new 12-inch fabs ramping: Beijing, Shanghai Lingang, Shenzhen, Tianjin), equipment costs inflated by export-control workarounds and extended tool lead times, plus wafer materials and labor. DUV multi-patterning at 7nm-class burns far more mask layers, tool time and yield than EUV flows - a structural COGS penalty vs TSMC/Samsung at equivalent geometry.
Capex
$8.10B FY2025 (above initial plan on strong demand and stretched equipment lead times); 2026 guided roughly flat with 2025 (per the March 2026 action plan). Capex ~87% of revenue vs ~30-35% at TSMC - only sustainable because Beijing (Big Fund, municipal JVs) co-funds capacity as strategic infrastructure.
Latest earnings
Roughly in line on revenue with margin above some expectations; the story was the guide, not the print - and the H-shares later jumped +10% on 2026-07-09 amid the AI/price-hike narrative
Q2 2026: revenue +14-16% QoQ, gross margin 20-22% (improved pricing and mix); FY2026 above-industry growth with capex held in line with 2025 (per March 2026 action plan)
- Capacity utilization
- 93.5% (FY2025)
- FY2025 capex
- $8.10B
- FY2025 net profit (attributable)
- $685M (+39% YoY in USD)
- TTM P/E (H-share)
- ~123x; forward ~71x (stockanalysis.com, 2026-07-09)
Growth drivers
- Domestic substitution — Chinese fabless (HiSilicon, GigaDevice, Will Semiconductor/OmniVision, UNISOC) reshoring wafer demand from overseas foundries under US export-control pressure
- Improved pricing and product mix cited in the Q2 2026 guide (+14-16% QoQ), with AI-adjacent domestic demand (power management, domestic accelerator ecosystems) supporting loadings
- 93.5% capacity utilization (FY2025) plus reported price increases into 2026 - pricing power from being the only scaled domestic advanced-logic option
- New 12-inch capacity ramps (Beijing/Shanghai/Shenzhen/Tianjin) adding tens of thousands of wpm on 28nm-and-above mature nodes
- 7nm-class (N+2) output reportedly for Huawei Kirin-class SoCs and domestic accelerators - small but strategically priced volume
Bull & bear
SMIC is the irreplaceable infrastructure of China's chip sovereignty push: a state-underwritten monopoly on domestic advanced logic with demand growing faster than it can add capacity, pricing power emerging (Q2 2026 guide +14-16% QoQ), and margins past trough - the market pays for strategic scarcity, and that scarcity is deepening.
- Captive demand: every US export-control tightening pushes more Chinese fabless volume to SMIC; utilization already 93.5% with price increases reportedly sticking
- Record FY2025 ($9.33B, +16.2%) with net profit +39% (USD) shows operating leverage as new fabs fill
- State co-funding removes the balance-sheet constraint that kills normal foundries at ~87% capex intensity - SMIC gets a TSMC-scale buildout without having to fund it from its own cash flow
- 7nm-class output reportedly for Huawei-class SoCs and domestic AI accelerators is strategically priced, and domestic AI demand is growing with no scaled alternative local supplier
- Toolchain localization (NAURA, AMEC, SMEE) is a slow-burn call option: each domestic tool qualified reduces the binding constraint on SMIC's roadmap
SMIC is a structurally low-margin, FCF-negative utility for Chinese industrial policy trading at ~123x earnings: the technology ceiling is hard (no EUV), yields at 7nm-class are reportedly poor, mature-node overcapacity is building, and one more turn of export-control escalation can impair existing operations - the equity prices in a scarcity premium the cash flows never validate.
- ~123x TTM / ~71x forward P/E for ~7% net margins, negative FCF, ~$2.2B net debt and a skipped FY2025 dividend - the valuation is a sovereignty narrative, not an earnings stream
- DUV multi-patterning at 7nm-class is an economic dead end: reported low yields already weigh on margins, and 5nm-class is effectively out of reach while TSMC ships N2
- Capex ~87% of revenue indefinitely; minority/JV structures and state priorities mean shareholders may never see the cash
- Entity List exposure is operational, not just roadmap: spares, servicing and consumables for installed US/NL/JP tools are a standing chokepoint
- China-wide mature-node capacity glut (Hua Hong, Nexchip, others) threatens the 28nm+ business that actually pays SMIC's bills
- US investors largely cannot own it (NYSE exit 2019, OTCQX removal and ADR termination 2021 after the CMC designation) - permanently impaired global investor base and index inclusion
What it is worth
Multiples vs foundry peer set + reverse-DCF sanity check; a conventional DCF is dominated by policy-driven capex, so scarcity/strategic premium must be named explicitly
Export-control escalation hits tool servicing or a mature-node price war compresses GM back under 15%; the scarcity narrative cracks and the multiple derates toward global mature-node peers - HK$30-45 zone (below the HK$44.40 52-wk low, toward early-2025 levels)
Revenue grows above-industry (low-to-mid teens) per the 2026 action plan, GM 20-22%, FCF stays negative; the stock tracks earnings growth off a rich base - fair value arguably well below spot on cash-flow math (~40-50x forward would imply roughly HK$47-60) but policy flows keep a floor
Sovereignty premium sustains or expands: domestic AI demand + price increases lift GM toward mid-20s%, revenue compounds mid-to-high teens, multiple holds - H-shares retest/exceed the HK$93.50 52-wk high
At HK$83.55 (2026-07-09) the H-shares trade ~123x TTM / ~71x forward earnings (stockanalysis.com) vs TSM ~25-30x and UMC/GFS ~15-25x - and the SSE STAR A-shares (688981.SS) trade at a further large premium to the H-shares. A reverse-DCF at ~$92B (H-share basis) requires either TSMC-like margin convergence (blocked by the EUV ceiling) or decades of state-subsidized growth with eventual FCF harvest; neither is visible in current economics (~7% net margin, negative FCF, net debt, dividend skipped). The price is carried by domestic-substitution scarcity, mainland/HK flows, and the AI narrative, not discounted cash flow. US persons face ownership restrictions stemming from the 2020 CMC designation and 2021 ADR termination; this is analysis of a covered ecosystem name, not an ownership recommendation.
SWOT
Strengths
- Only scaled Chinese foundry with near-advanced (7nm-class) logic capability - de facto monopoly on domestic advanced-node supply
- World No.3 foundry by revenue behind TSMC and Samsung (passed GlobalFoundries and UMC); $9.33B FY2025 revenue, 93.5% utilization
- State backing — Big Fund and municipal JV equity underwrite ~$8B/yr capex a company this size could never self-fund
- Captive, growing demand base as Chinese fabless firms are pushed (by US controls and Beijing procurement policy) to source domestically
- Broad mature-node platform (28nm+, CIS, PMIC, RF, power) where most global unit volume actually sits
Weaknesses
- No EUV access under current Dutch/US export controls — 7nm-class runs on DUV multi-patterning with reportedly poor yields (weighing on 2025 margins per TrendForce) and a hard wall near 5nm-class economics
- Structural margin gap — 21% gross / ~7% net vs TSMC's ~59% gross - depreciation-heavy, price-taker on mature nodes
- Capex ~87% of revenue means persistently negative FCF and a net-debt balance sheet (~$2.2B net debt at Q1 2026); FY2025 dividend skipped - shareholder economics subordinated to the national-strategy buildout
- US Entity List (Dec 2020) restricts tool and spare-part purchases; equipment lead times stretched and costs inflated
- Mature-node overcapacity risk — China-wide 28nm+ buildout (SMIC, Hua Hong, Nexchip) could crush mature-node pricing
Opportunities
- Domestic AI accelerator ecosystem (Huawei Ascend class and startups) has few alternatives for local fabrication - strategic volume at premium pricing
- Localization of the toolchain (NAURA, AMEC, SMEE litho) would slowly relax the binding equipment constraint
- Price increases plus mix shift (Q2 2026 guide +14-16% QoQ) as domestic demand outstrips domestic advanced capacity
- Automotive/industrial silicon localization in China — a large mature-node TAM playing directly to SMIC's capacity
- Advanced packaging / chiplet approaches to partially offset the node ceiling
Threats
- Export-control escalation — tighter US/NL/JP rules on DUV tools, parts, or servicing could stall existing lines, not just future ones
- Geopolitical tail risk — further US investment bans, sanctions on customers (Huawei), or Taiwan-strait conflict scenarios
- Mature-node price war as global and Chinese capacity additions land into soft demand pockets
- Yield/cost wall at 7nm-class and below without EUV - competitors extend the gap each node cycle
- Valuation risk — ~123x TTM earnings prices in national-champion scarcity, not cash flow; policy or sentiment shifts hit hard (0.37 beta belies event risk)
Moats, dependencies & bottlenecks
Moats
durable while US-China tech decoupling persists Big Fund + municipal JVs underwrite capex; Beijing procurement steers demand to it
durable domestically; nonexistent globally Near-monopoly inside China at 14nm-and-below capability; irrelevant outside China vs TSMC/Samsung
node-cycle length Standard foundry stickiness once designs tape out
eroding vs leading edge 2-3+ generations behind TSMC; no EUV path under current controls
Dependencies
No EUV allowed; even DUV sales/servicing to SMIC are progressively restricted by NL/US rules
Lam Research (LRCX), KLA (KLAC) Entity List licensing chokepoint for tools, spares, service on installed base
TOELY US OTC) and Japanese materials/equipment Japan aligned with US controls since 2023
Co-funds the ~$8B/yr capex; also subordinates shareholder returns to policy goals
Concentrated, sanction-exposed demand base; also its greatest growth engine
The long-term substitution path; not yet capable at advanced nodes
Advantages
- Sole scaled domestic supplier of near-advanced logic in the world's largest chip-consuming market
- State-underwritten capex lets it expand counter-cyclically at ~87% of revenue
- 93.5% utilization + emerging pricing power (price increases into 2026)
- Broad mature-node portfolio (CIS, PMIC, RF, power, MCU) matching China's electronics manufacturing base
- Deep bench of returnee process engineers — co-CEO Liang Mong-song (ex-TSMC/Samsung) drove the 14nm-to-7nm-class jump
Weaknesses
- Gross margin ~21% and net margin ~7% vs TSMC ~59%/40%+ — a structurally inferior business at roughly one-tenth the revenue
- Persistently negative FCF through the buildout (net debt ~$2.2B, dividend skipped); returns subordinate to policy
- No EUV path: each leading-edge cycle widens the gap
- US investor access severed (ADR program terminated 2021) — governance and disclosure thinner than US-listed peers
- JV/minority structures complicate the parent P&L and cash entitlement
Bottlenecks
- EUV lithography access - categorically blocked; hard ceiling near 7nm-class economics
- Yields on DUV multi-patterned 7nm-class (reported to weigh on 2025 margins)
- Tool spares/servicing licenses for installed US/NL/JP equipment
- Equipment delivery lead times (cited by SMIC as a 2025 capex driver)
- Advanced-capacity shortfall vs fast-growing domestic AI demand - SMIC itself is China's bottleneck
Top signals & trends
Top signals
Improved pricing and product mix; the strongest sequential guide in years
positive-momentum, valuation-caution · ~123x TTM P/E; scarcity premium, not cash-flow support
Demand and mix improving off the 2023 trough
The DUV multi-patterning cost penalty is showing up in the P&L
negative for FCF and shareholder returns, positive for strategic capacity · FCF stays negative; watch depreciation drag on 2026-27 margins (Q4 2025 GM already dipped to 19.2% on depreciation)
negative tail risk · The single largest event risk on the name
Trends
double-edged: caps SMIC's technology while guaranteeing its demand · Entity List since Dec 2020; ADR terminated 2021; controls have tightened repeatedly since 2022
strongly positive for demand · Huawei Ascend-class and domestic GPU startups must fab locally; AI-adjacent demand supports the 2026 guide
positive long-term, slow · NAURA/AMEC/SMEE progress determines whether SMIC's ceiling ever lifts
negative for pricing on 28nm+ (the bulk of revenue) · China-wide buildout plus soft consumer demand pockets
TSMC captures the global AI pool; SMIC captures China's walled-garden share
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
DUV immersion litho only - EUV blocked by export controls
Deposition/etch/implant; Entity List license-gated
Etch/deposition; license-gated
Metrology/inspection; license-gated
Coat/develop, etch; Japan-aligned controls since 2023
Leading domestic etch/deposition substitute
Domestic etch; a key localization bet
Domestic lithography hope; far behind ASML
Reported 7nm-class Kirin SoCs and Ascend AI silicon; sanction-exposed anchor customer (SMIC does not confirm)
Long-standing mature-node customer (PMIC/RF); US customers have trimmed exposure under controls
NOR flash / MCU; domestic substitution demand
CMOS image sensors on SMIC CIS flows
Domestic mobile SoC fabless
Global leader (~60%+ foundry share, ~59% gross margin, N2 in production); SMIC is 2-3+ generations behind and cannot compete for global leading-edge
No.2 in advanced logic with EUV; competes for the same second-source global demand SMIC can no longer serve
US/EU specialty foundry SMIC passed in annual revenue; competes on mature/specialty nodes for non-China demand
Taiwan mature-node foundry; direct competitor for 28nm+ logic/specialty outside China
China's No.2 foundry (specialty/power/embedded); domestic mature-node competitor
Fast-growing Chinese mature-node foundry (display drivers, CIS); adds to domestic 28nm+ price pressure
Aspiring Western foundry alternative at the leading edge; not a direct SMIC competitor but shapes the global capacity picture