SMIC does not design chips — it manufactures other companies' designs, like TSMC, but from inside a US export-control perimeter that denies it the most advanced tools. In FY2025 it earned a record US$9.33bn of revenue (+16.2%) and RMB 5.04bn of net profit (+36%), held an estimated ~5.3% of the global foundry market — #1 in mainland China — and ran its fabs at 93.5% utilization.[1]·[2]·[3]·[9] The question is not whether SMIC matters to China; it is whether it can keep advancing against an embargo, a margin squeeze, and its own concentration all at once.

Figures from SMIC's FY2025 results, Chinese financial press and industry trackers.[1]·[1]·[3]·[31] Revenue in US$; profit in RMB as reported.

The three questions this case study weighs

SMIC sits at the intersection of a normal cyclical foundry business and an abnormal geopolitical one. Read the company only through Chinese policy coverage and it looks like an unstoppable national champion; read it only through Western export-control coverage and it looks structurally capped. Both framings are sourced below — the honest answer lives in the tension between them. Three questions decide which way it breaks.

SMIC was built fast and contentiously — hiring away TSMC talent, losing a major IP suit, then being recruited as the vehicle for Chinese chip self-sufficiency. That dual inheritance, poached ex-TSMC and ex-Samsung process expertise plus deep state entanglement, is exactly what makes it both formidable and sanctioned today.

What SMIC does

SMIC is a pure-play foundry: it fabricates integrated circuits to other companies' designs and does not sell its own branded chips, the same model as TSMC and UMC. It spans mature nodes (28nm and above, the bulk of its output) through to a DUV-built advanced node branded internally as N+2/N+3. Its customers are concentrated in China — domestic fabless designers, power and display chips, and, most visibly, Huawei's HiSilicon for smartphone and AI silicon.[31]·[3] In 2025 it ranked first among mainland-China foundries and, by Chinese industry accounting, second among the world's pure-play foundries by sales — though trackers that count Samsung's foundry place it third.[17]·[9]

A contentious founding

SMIC's speed out of the gate — installing equipment to qualified production within months — was aided by hiring more than a hundred staff from TSMC, and TSMC alleged systematic trade-secret theft. After a 2005 settlement, a 2006 second suit, and a 2009 California jury verdict against SMIC, the company settled for US$200m and a ~10% equity stake to TSMC, and founder Richard Chang resigned.[6]·[7] It is an uncomfortable origin story that SMIC's boosters rarely foreground and its critics never forget.

Timeline

A foundry that has fought with itself

SMIC's advanced-node leap is inseparable from one polarising figure: co-CEO Liang Mong Song. Chinese tech press credits him with driving five process generations from 28nm to 7nm in roughly three years — and chronicles repeated boardroom friction, including a December 2020 resignation letter in which he said he no longer felt "trusted" after ex-TSMC veteran Jiang Shangyi was appointed vice-chairman without his knowledge. The episode halted SMIC's shares and rattled investors before he ultimately stayed.[9]·[8]·[10]

SMIC competes in two markets at once: a commoditised mature-node business where overcapacity and price pressure are real, and a politically priceless advanced-nodebusiness where it is China's only credible option. The first pays the bills; the second sets the stock's narrative.

Where the money actually comes from

The global foundry market is roughly a US$170bn industry in 2025, and it is extraordinarily concentrated: TSMC alone holds close to 70%.[9]SMIC's ~5% slice is built overwhelmingly on mature nodes (28nm and above) serving Chinese designers — power-management chips, display drivers, image sensors, microcontrollers, and consumer SoCs. In Q4 2025 consumer electronics was its single largest application at roughly 47% of wafer revenue, with smartphones around 20%, and China supplied close to 90% of sales.[1]·[31]

Approximate shares from results coverage; SMIC reports application mix without exact sub-segment splits.[1]·[23]

The domestic-substitution tailwind

SMIC's real market advantage is structural: it is the designated beneficiary of China's drive to replace imported chips. A widely-circulated broker deep-dive frames it as the "main force of domestic substitution, the light of advanced process" — capturing how Chinese capital markets see it less as a normal foundry than as a policy instrument.[16] That status brings a captive customer base (domestic fabless firms increasingly required, or incentivised, to source locally) and access to state capital through the Big Fund.[16]·[17]

The flip side: a market that can flood

That same domestic focus is a vulnerability. The flood of Chinese mature-node capacity — SMIC's own plus Hua Hong's and others' — has at times produced overcapacity and price competition in exactly the nodes SMIC depends on. Chinese industry commentary has described mature process nodes as "in uproar" over oversupply, even as SMIC's management insists mainland foundries are not slashing prices to grab orders.[18]·[7] The result is a business with strong unit-volume growth but persistent pressure on price and margin (see Business Model).

SMIC's revenue grows mainly by shipping more wafers, not charging more for them. With average selling prices flat-to-down and depreciation set to rise ~30% in 2026, the model only works if utilization stays very high and the advanced-node mix slowly enriches.[6]·[13]

How SMIC makes money

SMIC sells wafer fabrication capacity, priced per wafer and varying sharply by node. The business has three economic levers: utilization (idle fabs still depreciate), average selling price (a function of node mix and pricing power), and capex/depreciation (each new fab adds revenue capacity but also a heavy depreciation charge for years). In 2025 SMIC ran at 93.5% utilization, shipped ~9.7 million 8-inch-equivalent wafers, and spent US$8.1bn of capex — a sum larger than its entire 2025 revenue-derived net profit many times over.[1]·[12]

The "volume-up, price-down" trap

The core tension is visible in the 2024 numbers Chinese analysts dwell on: wafer shipments rose sharply, but the average selling price fell about 5%, and full-year gross margin sank to 18.59% — a nine-year low — as depreciation climbed.[6]·[5]SMIC's answer has been "以量补价" — making up for lower prices with higher volume. That works while demand absorbs new capacity, but it is a treadmill: each new fab must be filled at acceptable prices or it becomes a depreciation anchor. Margin recovered toward ~22% across full-year 2025 before Q4 slipped to 19.2% as fresh capacity began depreciating.[3]·[1]

Why expansion is the whole strategy — and the whole risk

SMIC is mid-way through one of the industry's most aggressive build-outs: multiple new 12-inch fabs in Shanghai, Beijing, Tianjin and Shenzhen, mostly targeting 28nm-and-above mature nodes for domestic demand.[14] Management skipped its 2025 dividend to fund it.[12]The bet is that China's self-sufficiency demand will fill these fabs faster than depreciation and price erosion can hurt. If demand disappoints — or if rival Chinese capacity floods the same nodes — the build-out converts directly into margin pain.[18]·[13]

SMIC is not really competing with TSMC for Apple or Nvidia — it can't. It competes to be China's default foundry: dominant domestically on advanced nodes, but in a knife-fight with Hua Hong and others on the mature nodes that actually pay its bills.[3]·[18]

The five forces, under embargo

Porter's framework looks unusual for SMIC because the most dangerous force is not a competitor but a supplier it is forbidden to buy from. Click each force for the evidence.

Where SMIC sits versus peers

On a map of process leadership against tool/market access, SMIC lands in a lonely middle: more advanced than the mature-only foundries (UMC, GlobalFoundries, Hua Hong) but well behind TSMC and Samsung, and uniquely constrained on access. Hover a point for the basis.

SMIC's moat is scarcity, not superiority: it is years behind TSMC, but it is the only company that can make advanced chips inside China, without EUV — and for Huawei and Beijing, that scarcity is worth more than raw competitiveness.[3]·[4]

The 2023 surprise: 7nm without EUV

In September 2023, teardown firm TechInsights confirmed that Huawei's Mate 60 Pro contained a Kirin 9000s built on SMIC's 7nm (N+2) process — the first commercial 7nm-class chip from a Chinese foundry, achieved without EUV lithography, using DUV multi-patterning on existing ASML DUV tools.[3] The political resonance was enormous: it appeared to show US export controls had not stopped China from reaching a node Washington intended to deny it.

SMIC has since pushed further, to an N+3node described as "5nm-class" — used in Huawei's Kirin 9030 and Ascend AI accelerators — which Tom's Hardware calls China's most advanced process to date, while noting it still "can't compete with 5nm" leading-edge parts on density and efficiency.[4] Much of this is credited to co-CEO Liang Mong Song, who Chinese press say drove five process generations from 28nm to 7nm in about three years.[9]·[18]

The export-control perimeter

SMIC was added to the US Entity List on 18 December 2020, restricting its access to US-origin technology and explicitly targeting sub-10nm production.[32] In December 2024, BIS tightened this to a presumption of denial for essentially all items subject to the Export Administration Regulations, with only narrow exceptions for older 200mm production.[8] The practical effect: SMIC can keep buying and running the DUV tools it already has, but its path to EUV — and thus to true leading-edge nodes by conventional means — is closed for now.

SWOT

SMIC out-earns the other mature-node specialists (UMC, GlobalFoundries, Hua Hong) and is the clear Chinese leader — but it is an order of magnitude smaller than TSMCand structurally walled off from the leading-edge customers that drive TSMC's growth.[9]·[29]

Foundry revenue, latest full year

TSMC and SMIC are reported FY2025. Samsung foundry, UMC, GlobalFoundries and Hua Hong are approximate, annualised from quarterly disclosures — treat as indicative, not exact.[9]·[29]·[30]

Benchmark table

Shares are 2025 estimates from foundry trackers; revenue as noted above. "Leading edge?" describes the most advanced node in volume.[9]·[27]·[29]·[30]

The metric that flatters SMIC — and the one that doesn't

SMIC's growth and domestic share flatter it: 16% revenue growth in a flat-to-soft year, clear #1 in mainland China. Its global share and margindo not: 5.3% and slipping as TSMC pulls away, with gross margin well below TSMC's high-50s%. Hua Hong, its closest mainland peer, is a fraction of its size and has at times posted revenue declines — underscoring how far ahead SMIC is within China even as it trails globally.[27]·[30]·[5]

In 2025 SMIC earned a record US$9.33bn of revenue and grew net profit 36% to RMB 5.04bn — yet its A-shares trade as a multi-hundred-billion-dollar strategic story, not on these earnings. The bull case is the narrative; the bear case is the margin.[1]·[3]·[4]

FY2025: the headline numbers

Revenue per company/industry reporting (US$); profit and market value as reported in RMB.[1]·[1]·[3]·[4]

The margin rollercoaster

Gross margin is where SMIC's story gets honest. It peaked near 38% in the 2022 boom, collapsed to ~22% in 2023 and a nine-year low of 18.59% in 2024 as the cycle turned and depreciation rose, then recovered toward ~22% across full-year 2025 — before Q4 2025 slid back to 19.2% as new fabs began depreciating.[5]·[1]·[3] Management has guided that 2026 depreciation will climb roughly 30%, a direct headwind it is trying to offset with mature-node price increases.[13]·[13]

The same quarter, side by side: who actually earns a margin

SMIC's margin only makes sense against its peers. In the same quarter (Q4 2025), TSMC earned a 62.3%gross margin; even mature-node specialists UMC and GlobalFoundries cleared the high-20s to ~30%. SMIC's 19.2%sits near the bottom of the global foundry table — above only its fellow sanctioned mainland peer Hua Hong. This is "scarcity, not superiority" made literal: SMIC can build China's most advanced non-EUV logic, yet the DUV-multipatterning penalty and depreciation leave it one of the thinnest-margin foundries in the world.[1]·[33]·[34]·[35]·[36]

Q1 2026: a wobble

The most recent quarter showed the tension live. Q1 2026 revenue rose to about US$2.51bn (RMB 17.6bn, +8.1% YoY) with gross margin recovering to 20.1% — but net profit of US$197m rose only 5% and missed the ~US$224m consensus, while smartphone-related revenue fell to 18.9%, its lowest share since at least 2021.[2]·[12] An earlier Q1 2025 miss had sent the shares down nearly 7% in a day, a reminder that the strategic narrative does not insulate the stock from execution wobbles.[11]

The investable question is not whether SMIC has risks — it is whether its state protection outweighs its state-linked exposure. The same government ties that guarantee its market are the reason it is sanctioned.[16]·[21]·[8]

1 · Export controls (the defining risk)

SMIC has been on the US Entity List since December 2020, and since December 2024 faces a presumption of denial for nearly all US-origin items.[32]·[8]US policy is explicitly designed to degrade China's ability to make advanced chips indigenously — a Congressional Research Service report frames the controls in exactly those terms.[26] The risk is asymmetric and ongoing: controls can tighten (extending to more nodes or better enforcement) far more easily than they can loosen.

2 · Mature-node overcapacity and price war

SMIC's bread-and-butter mature nodes face a domestic glut as Chinese foundries all expand at once. UBS and other analysts have flagged overcapacity and margin pressure from large-scale expansion as principal risks; Chinese commentary describes the mature-process segment as "in uproar."[19]·[18] Management disputes that mainland foundries are slashing prices, but the "volume-up, price-down" dynamic is visible in the margin trough.[7]·[6]

3 · Concentration risk

With ~85–90% of revenue from China and heavy exposure to consumer and smartphone end-markets, SMIC rises and falls with one economy and a handful of demand cycles.[31]·[2] A weak Chinese consumer or a Huawei-specific setback would hit it disproportionately.

4 · Governance, ownership and the security case

SMIC's major shareholders — Datang Telecom and the National IC "Big Fund" via Xinxin (HK) — tie it closely to the Chinese state, and the US Defense Department has designated it as linked to the military, the basis for much of the sanctions case.[21] Internally, the company has lived through repeated boardroom conflict around its advanced-node leadership — a real key-person and governance risk for a firm whose edge depends on a few veteran engineers.[8]·[10] Some former Taiwanese executives have even publicly questioned the decision to join.[15]

The single question that decides SMIC's decade is lithography: if China builds a viable domestic EUV (or a DUV workaround good enough at scale), SMIC re-rates; if not, it stays a vital but capped mature-node champion. Watch the tools, not the wafers.[19]·[28]

What to watch

• Domestic EUV progress. A Shenzhen facility reportedly validated an EUV prototype generating a stable 13.5nm beam in December 2025 — but a lab prototype is many years and many yield points from a production tool. Treat as speculative.[19]
• Huawei Ascend ramp. SMIC's advanced capacity increasingly serves Huawei's AI accelerators; reported Ascend die yields rose from ~20% to ~40% over a year, with HBM supply a separate bottleneck.[25]
• Mature-node pricing. Early-2026 reporting shows prices turning up as supply tightens — the swing factor for near-term margins.[13]
• Export-control moves. Any extension of controls (or a thaw) reshapes the whole thesis.[8]

Three scenarios

The narrative shift to watch

Part of SMIC's 2025–26 re-rating rode a change in how China frames the race itself. Chinese coverage ties the stock's record highs to Huawei's "Tao's Law" framing — the idea that value can come from system-level co-optimization and advanced packagingrather than node-shrink alone — which, if it holds, would make SMIC's EUV gap matter somewhat less.[20] Whether that is genuine engineering logic or narrative convenience is itself one of the things reasonable people disagree about.

Method

Research proceeded by fan-out web search and direct fetching of primary and reputable secondary sources, working per section from SMIC's own results coverage outward to trade analysts, press, teardown firms and tertiary context. Because SMIC is a Chinese company whose richest and most candid coverage is in Mandarin, a substantial share of the work was done natively in Chinese (新浪财经, 证券时报, 财联社, 澎湃新闻, 品玩, 钛媒体, 自由亚洲电台, 东方财富证券), including deliberate disconfirming searches (产能过剩 / 价格战 / 内卷 / 内斗 / 量增价跌 / 毛利率 创新低), with Chinese quotes shown in the original alongside an English translation. Every URL cited was opened during the research run; a link-checker validated each one, and quotes were spot-checked against the source. Each claim was transcribed into a structured manifest tagging it with a tier, a confidence level and a stance. The load-bearing figures are SMIC's FY2025 and Q1 2026 financials and operating metrics — revenue, net profit, gross margin, utilization, capex — together with third-party teardown findings (TechInsights) on its process node and (estimated) ~5.3% foundry share.[1]·[2]·[3]·[9]

Frameworks used

• Pyramid Principle — answer-first executive summary built around the balance of evidence on three open questions.
• Porter's Five Forces — for the competitive landscape, with the unusual feature that the binding force is a supplier SMIC may not buy from.
• Peer comparables — SMIC vs TSMC, Samsung, UMC, GlobalFoundries and Hua Hong on revenue, share, node and tool access.
• SWOT and a 2×2 positioning map — applied even-handedly, giving weaknesses and threats equal depth to strengths.
• Scenario analysis — bull/base/bear in the forward view, framed as possibilities to weigh, not a forecast.

Disclosed vs. estimated

SMIC discloses, in its results, revenue, net profit, gross margin, capacity, utilization and capex. It does not disclose node-level yields, customer-level revenue, or precise application splits. Consequently, several items here are third-party estimates or teardown findings and are labelled as such: process-node identity and yields (TechInsights and analyst estimates), foundry market shares (TrendForce/ Counterpoint trackers), application mix (approximate), peer revenue (annualised from quarterly disclosures), and the domestic-EUV prototype report (single-sourced, speculative).[3]·[9]·[19]·[29] RMB and US$ figures are reported as each source gives them; conversions are approximate.

Neutrality commitment

This study is a compilation meant to let you reach your own conclusion. Each section presents both supporting and critical evidence; Chinese domestic framing (national champion, domestic substitution) and Western framing (sanctioned, sub-scale, state-linked) are both shown rather than treating either as the neutral default. The source manifest tags every source's stance so the balance is visible on the Sources page.

Tier 1 = primary / authoritative (regulatory filings, TechInsights teardowns, US government / CRS). Tier 2 = reputable secondary (Reuters/CNBC, TrendForce, DIGITIMES, Caixin, 证券时报, 新浪财经, 财联社, 澎湃, EE Times). Tier 3 = tertiary / context (encyclopedias, market trackers, analyst blogs, broker notes) — used for colour, not load-bearing facts.