SMIC (0981.HK) Zen Horizon In-Depth Research

(This section gives no investment stance or rating. Those judgments are left to Section 10, where they follow naturally from the facts in the first nine sections.)

2. Longitudinal Analysis: A Company History of Catching Up Between Sanctions and Litigation

SMIC's 26 years are almost a national project history of "China wanting its own TSMC." Every turning point, founding, litigation, listing, sanctions, and breakthrough, has been both a corporate event and a geopolitical event.

2.1 Origins: A Founder Pushed Into Entrepreneurship by TSMC's Acquisition

SMIC was founded in Shanghai in April 2000 by Richard Chang. Chang had spent about 20 years at Texas Instruments building fabs. In 1997 he founded Worldwide Semiconductor Manufacturing Corp. (WSMC) in Taiwan to compete with TSMC. In 2000, TSMC acquired WSMC for about USD 5 billion. Chang reportedly made "building a fab in mainland China" a condition for staying on, and after learning that TSMC founder Morris Chang had no intention of doing so, he gave up TSMC shares and moved north to mainland China to found SMIC (Zhang Rujing - Wikipedia). This history between the "two Mr. Changs" set the stage for the decisive lawsuit that followed. SMIC's mission at birth was blunt: grow a domestic foundry industry for China from zero.

2.2 Listing and the Lawsuit That Defined the Company (2003-2009)

SMIC listed simultaneously on the Hong Kong Stock Exchange (0981) and the NYSE in March 2004, telling capital markets the story of "China's TSMC." But what truly shaped its early fate was TSMC's trade-secret litigation. TSMC sued in 2003, settled in 2005, sued again in 2006 alleging breach of settlement and further theft of secrets, and in the liability phase of the 2009 California jury trial TSMC prevailed on 61 of 65 claims (Taipei Times 2009-11-05).

The November 2009 settlement terms still matter today: SMIC had to pay TSMC USD 200 million in cash in installments through 2013, and grant TSMC about 8% equity, 1.789 billion shares, plus warrants to subscribe for 696 million shares, reaching about 10% if exercised (SMIC 6-K filed with the SEC). Days after the settlement, founder Richard Chang resigned, ending the founder era. The long-term impact was twofold: it branded SMIC as a "technology follower that must be especially cautious on IP," and it cleared the founder-led setup, making room for the later governance structure of state-capital leadership plus professional managers and heavyweight technologists. TSMC gradually reduced its holding in later years and is no longer an SMIC shareholder today.

2.3 Development Stages: Four Arcs

Compress SMIC's history into four stages and its fate becomes clearer:

1. Stumbling catch-up phase (2000-2011): fab construction, expansion, litigation, and years of losses. After Richard Chang left, David Wang departed after shareholders rejected his reappointment in 2011, and Tzu-Yin Chiu became CEO in 2011, bringing operational stability. But SMIC's process technology remained two to three generations behind global leaders for years.

2. Process breakthrough phase (2017-2020): In October 2017, Haijun Zhao and Liang Mong Song became co-CEOs (PRNewswire 2017). Liang Mong Song, a key figure in advanced process development at TSMC and Samsung, lifted 14nm FinFET yield from single digits to above 95%, brought it into mass production in 2019, and pushed toward 7nm (Liang Mong Song - Wikipedia). This was the first time SMIC truly approached leading-edge international nodes. Governance drama also appeared in this period: in December 2020 Liang threatened to resign after Chiang Shang-Yi was hired as vice chairman "without consulting him," and the stock fell about 5% in one day (CNN 2020-12-16). Liang ultimately stayed and Chiang left in 2021, showing that this company's advanced-node progress is highly dependent on individual technical leaders.

3. Return to A shares and sanctions (2019-2020): In June 2019, SMIC delisted its ADR from the NYSE. The official explanation was cost and liquidity, while the market read it through the lens of U.S.-China frictions. On July 16, 2020, it listed on the STAR Market (688981), raising up to about RMB 53.2 billion, or about USD 7.5 billion, and rose 202% on the first day. It was the largest IPO in STAR Market history and one of China's largest equity offerings in a decade (Caixin 2020-07-06). The National Integrated Circuit Industry Investment Fund, or "Big Fund" Phase I/II, became an important shareholder through this process. Half a year later, on December 18, 2020, the U.S. Commerce Department added SMIC to the Entity List, with a presumption of denial for items "uniquely required" for 10nm and below, especially EUV (U.S. Commerce 2020-12). That cut determined all of SMIC's technology ceilings thereafter.

4. Breakthrough and positioning under sanctions (2021 to today): Without access to EUV, SMIC used DUV multi-patterning to force a 7nm outcome. In July 2022, TechInsights first found SMIC 7nm, N+1, in a bitcoin mining ASIC. In August-September 2023, the Kirin 9000s inside Huawei's Mate 60 Pro was confirmed to be SMIC 7nm, N+2, in volume production. This became the symbolic moment showing that "sanctions had not stopped China from mass-producing 7nm," with strong political meaning because it coincided with the U.S. Commerce Secretary's visit to China (TechInsights 2023-09). The Kirin 9010 in 2024 and Huawei Ascend AI chips continued this "SMIC-Huawei" axis. At the same time, SMIC pushed capex to the limit and built out four new 12-inch fabs, Beijing Jingcheng, Shanghai Oriental, Shenzhen, and Tianjin Xiqing, betting on domestic substitution in mature-node capacity.

2.4 Hindsight on Key Events

• 2009 TSMC settlement: In hindsight, it did not crush SMIC, but it established the posture and caution boundary of a follower.

• 2020 STAR Market listing + Big Fund investment: This turned SMIC from an ordinary listed company into a national-team vehicle, the key to understanding later capital allocation: heavy on capacity, light on dividends.

• 2020 Entity List: What was underestimated was not the short-term hit, but the permanent structural constraint. It locked in SMIC's process ceiling.

• 2023 Kirin 9000s: Its economic meaning may have been overestimated. It proved that SMIC "can do it," but not that it "can do it profitably or continue moving downward" (see Sections 3 and 4).

3. Longitudinal Financial Review: Revenue at a Record High, Returns Paper-Thin

SMIC's financial statements are the part of this report that most needs to be "taken apart," because headline growth and real returns are separated by three walls: subsidies, depreciation, and minority interests.

3.1 Revenue: Through a Full Cycle

In USD terms, SMIC's revenue shows a clear cycle (stockanalysis):

The business logic behind the numbers: 2021-2022 was the chip-shortage period of rising volume and price, pushing gross margin to a historical peak of 38%; in 2023, consumer electronics weakened, revenue fell 13%, and gross margin was cut to below 20%; the 2024-2025 recovery was driven by shipment volume, not pricing. FY2025 wafer shipments rose 20.9% to 9.697 million 8-inch equivalent wafers, but blended ASP actually fell from about USD 933 per wafer to about USD 907 per wafer (TrendForce 2026-02-11). In other words, SMIC's "growth" is more about building capacity and filling lines through domestic substitution than about pricing power.

3.2 Gross Margin: A Ceiling Pinned Down by Depreciation

SMIC's gross margin has repeatedly stayed in a narrow 18-22% range over the past three years (FY25 by quarter: Q1 22.5% -> Q3 22.0% -> Q4 19.2% -> 2026Q1 20.1%). The reason is structural: more than USD 8 billion of annual capex creates massive depreciation. FY2025 depreciation and amortization were about USD 3.81 billion, more than 40% of revenue and close to twice gross profit (stockanalysis cash flow). Every small lift in utilization and ASP is partly pulled back down by depreciation from new fabs. This is the financial-statement shadow of the "DUV 7nm ceiling": even with utilization above 95% and a favorable product mix, Q3'25 gross margin only touched 22% before guidance pulled it back to 18-20% (Kristal.ai Lens 2025-11).

3.3 Profit Quality: Strip Three Layers and the Core Business Is Thin

This is the most important "microscope slide" in the report. SMIC's FY2025 profit attributable to owners was RMB 5.041 billion, about USD 700 million, up 36.3%. That looks decent, but:

• Layer 1, government subsidies: FY2024 government subsidies were about USD 411 million, equivalent to about 87% of operating profit. Strip out subsidies and operating profit was only about USD 62 million, for an operating margin of about 3.6% (HDIN Research 2026-03).

• Layer 2, non-recurring items: FY2024 recurring profit attributable to owners was only RMB 2.646 billion, versus reported profit of RMB 3.699 billion, meaning about 28% was non-recurring. FY2025 recurring profit was RMB 4.124 billion versus reported RMB 5.041 billion, with a similar gap (Yicai, C114 2026-03). Roughly USD 10 billion of cash on the balance sheet also generated meaningful interest income, further lifting pre-tax profit.

• Layer 3, minority interests: Many of SMIC's new fabs are joint ventures with state-capital funds, and about 30% of net profit leaks to minority shareholders. FY25 minority interests were about USD 304 million, leaving only about 70% for parent shareholders.

After peeling all three layers, the "real foundry profit" left to H-share shareholders is extremely thin. A revealing comparison: in FY2024, SMIC's R&D spending was RMB 5.45 billion, about USD 760 million, the highest among Chinese chip companies (TrendForce 2025-05). R&D expense was larger than profit attributable to owners itself.

3.4 Capex and Free Cash Flow: Deep Bleeding Year After Year

SMIC is one of the most capital-intensive foundries in the world. Capex over the past three years was about USD 7.6 billion in 2023, USD 7.7 billion in 2024, and about USD 8.1-8.4 billion in 2025, about 90% of revenue. Corresponding free cash flow, operating cash flow minus capex, has been deeply negative year after year: about USD -4.3 billion in 2023, USD -4.5 billion in 2024, and about USD -5.0 billion to USD -5.2 billion in 2025 (stockanalysis cash flow, HDIN).

This bleeding is structural, not cyclical. As long as domestic substitution requires SMIC to keep building capacity, capex will continue to equal or exceed operating cash flow. The clearest "funding stress signal" is that SMIC canceled its 2025 annual dividend and instead sought expanded share-issuance authorization at the AGM (TipRanks AGM). Fortunately, the balance sheet remains sound: cash plus short-term investments are about USD 10 billion, total debt about USD 12.6 billion (total liabilities about USD 17.3 billion), net debt only about USD 2.6 billion, and debt/equity about 35%. Expansion is supported by state-capital financing and equity funding from the STAR Market listing.

3.5 Returns: Low-Single-Digit ROE

Put the above together and the conclusion is harsh: parent ROE is only about 3.2% in FY25, 2.4% in FY24, and 4.5% in FY23 (profit attributable to owners divided by parent equity of USD 21.4 billion). ROIC is also low single digits. A huge capital base of about USD 35 billion in total equity and more than USD 8 billion of annual capex squeezes out only about USD 500-700 million of profit attributable to owners. This is a "capital-consuming" business, not a "capital-returning" business, at least at the current process and cycle position.

3.6 Latest Quarter (2026Q1) and Guidance

2026Q1 revenue was USD 2.505 billion, up 0.7% QoQ, gross margin was 20.1%, utilization was 93.1%, shipments fell 0.2% QoQ, but blended ASP rose 2.5% QoQ. Q2'26 guidance: revenue up 14-16% QoQ, gross margin 20-22% (Q1'26 earnings call highlights). The strong guidance reflects AI pushing mature-node demand and prices upward. But gross margin is still stuck below the old 22% ceiling.

4. Share Price and Valuation History: From "Technology Choke Point" Narrative to "AI Price-Up Cycle"

Since its H-share listing in 2004, SMIC has long been a stock that "swings violently with the big cycle and political sentiment." The market has given it three labels: follower discount when process technology lagged and losses persisted, strategic premium after the Entity List and Kirin 9000s, and cyclical growth during the 2025-2026 AI mature-node price-up cycle.

The migration in valuation center explains the issue well. According to GuruFocus, SMIC's H-share 10-year PE range is 7.5-166x, with a median of only 22x, while the current TTM multiple is about 110-120x, about 300% above the median and in the top decile (GuruFocus). P/B has also moved from below 1x in FY2023 to a high level today (see Section 7). The upward move in valuation center has mainly come from "market preference changing" (domestic substitution + AI theme premium), rather than "business quality changing". ROE has remained low single digits in recent years. That is a dangerous combination: when a stock's high valuation is built on narrative preference rather than improving returns, the valuation drawdown can be severe when the narrative fades.

5. Business Model and Moat: Strong in Positioning, Weak in Economics

5.1 Revenue Mix and Operating Leverage

About 94% of SMIC's revenue is foundry service fees. By product, 12-inch wafers are about 77% and 8-inch wafers about 23%; by application, consumer electronics are about 43%, smartphones about 23%, computers/tablets about 15%, industrial and automotive about 11%, the fastest-growing category, and connectivity and wearables about 8% (FY25 results summary). On the cost side, fixed costs, especially depreciation, are extremely high, so in theory there is operating leverage: when utilization is full, profit sensitivity should be large. In reality, this leverage is continuously offset by depreciation from new fabs. Whenever utilization or ASP improves, another newly commissioned line adds a fresh round of depreciation. When revenue falls, as in 2023, gross margin can be cut from near 40% to below 20%. Downside leverage is larger than upside leverage, a typical feature of heavy-asset cyclicals.

5.2 Moat: Three Real Ones, One Is a "Policy License"

The few that truly hold up:

• 1. Scale and local capacity positioning (strong and strengthening): SMIC is mainland China's largest foundry by capacity, with year-end monthly capacity of about 1.059 million 8-inch equivalent wafers and FY25 utilization of 93.5%. Under the policy force of domestic substitution, being "in China, for China" is itself a moat. Many local fabless companies, including HiSilicon, Cambricon, UNISOC, and Will Semiconductor, must use domestic capacity for supply-chain security reasons.

• 2. Domestic monopoly in advanced process (strong, but capped): SMIC is the only mainland company capable of volume production at the 7nm class. Huawei's Kirin and Ascend have no alternative. This moat is "one of one" inside China, but globally it is "an island capped by the EUV ban." It can be exclusive, but it cannot easily move lower, and it is not cheap.

• 3. Customer switching costs (medium): Foundry tape-out and qualification cycles are long, and switching costs are high, especially for advanced-node customers that are tightly locked in.

• 4. Regulatory-license-style "national team" status (strong, but double-edged): Big Fund backing, policy orders, and almost bottomless capital support are advantages private competitors cannot replicate. But this also means capital allocation serves national strategy, not shareholder returns (see 5.3 and Section 8).

The key distinction: SMIC's moat is "positioning-based," not "profitability-based." It guarantees "irreplaceability," but not "profitability." TSMC's moat, EUV advanced nodes, high yield, and pricing power, turns directly into a 59% gross margin and 36% ROE. SMIC's moat turns into "strategic security," not cash returns.

5.3 Management and Governance: Technical Leaders + State-Capital Leadership

Co-CEOs Haijun Zhao (operations) and Liang Mong Song (technology) are central. Liang's retention directly affects the pace of advanced-node progress, as the 2020 resignation episode already proved this "person dependency" risk. In ownership, SMIC has no single controlling shareholder. Big Fund Phase I/II, Datang Holdings (CICT system), and other state-capital funds are major shareholders (2024 annual report (HKEX)). In capital allocation, management's recent choices have consistently pointed to national strategy first: canceling the dividend, expanding share-issuance authorization, and buying out subsidiary minority stakes at a share premium. Morningstar has argued that such actions are not friendly to minority shareholders (Morningstar). This is not a governance "fraud" issue; it is a governance "objective function" issue. For H-share minority shareholders, the company's first goal is not to make you money. It is to ensure China has chips.

6. Industry and Cycle: Domestic Substitution Is a Tailwind, Mature-Node Overcapacity Is a Headwind

6.1 Industry Structure and SMIC's Niche

Global foundry is a highly concentrated market: FY2025 top-10 foundry revenue was about USD 169.5 billion, with TSMC alone at 69.9%, Samsung 7.2%, SMIC 5.32%, third overall and second among pure-play foundries, UMC 4.35%, GlobalFoundries 3.87%, and Hua Hong about 2.6% (TrendForce 2026-03). Most of the profit pool is captured by TSMC's advanced nodes. This is a structure where "the winner takes advanced nodes, and the rest fight in mature nodes." SMIC's niche is unusual: China's No. 1 by scale + domestic monopoly in advanced process + global cost/performance laggard. All three identities are true at the same time.

6.2 Cyclicality: Semiconductor Cycle + Capex Cycle + Policy Cycle

SMIC is driven by three cycles at once: the semiconductor demand cycle, currently in an AI-driven recovery upswing; its own capex cycle, currently in expansion with heavy depreciation pressure; and the policy cycle, with accelerating domestic substitution and tightening export controls. The current 2025-2026 phase is a divergent recovery: advanced nodes for AI GPU/TPU are tight, while mature nodes had been weak but began tightening again from the second half of 2025 due to demand for AI power management ICs (PMICs) and other products. SMIC and Hua Hong are fully loaded, and some processes have raised prices by about 10% (TrendForce 2026-01-27). This is the factual basis for the current bull narrative.

6.3 Mature-Node Overcapacity: SMIC Is Both Beneficiary and Creator

But mature nodes carry a long-term concern: China is aggressively expanding mature-node capacity. By end-2025, China is expected to account for about 28% of global mature-chip capacity, and by 2029 new fabs in China are expected to account for nearly half of global new wafer fabs (Tom's Hardware). In early 2025, SMIC itself reportedly cut 28nm quotes from about USD 2,500 per wafer to about USD 1,500, down 40%, based on a single source and indicative. This double-edged sword means SMIC is enjoying domestic-substitution volume while helping create the future price war with its own hands. On 2024-12-23, the U.S. USTR launched a Section 301 investigation into China's "legacy chip" industrial policies (Federal Register 2024-12-30). No final tariff has been issued.

6.4 Export Controls: The Structural Constraint That Determines Fate (With Dates)

This is the core for understanding SMIC's long-term ceiling:

• EUV has never been approved for sale to China. Since around 2019, under U.S. pressure, the Netherlands has not issued licenses. SMIC is structurally capped around the 10nm class.

• 2020-12-18 Entity List, presumption of denial for <=10nm; 2022-10-07 broad BIS rules on advanced computing, HBM, equipment, and FDPR; 2023-10-17 tightening update.

• DUV immersion lithography: Since 2024, Dutch/ASML exports and service for NXT:2000i and above have been restricted.

• The 2026 MATCH Act, proposed but not enacted, would legally prohibit sales of DUV immersion tools to China and prohibit service for existing China-based machines, explicitly naming SMIC (Asia Times 2026-04). If a "service ban" is implemented, it threatens maintenance of SMIC's existing DUV fleet, which is much more serious than banning new machine sales.

In technology economics, using DUV multi-patterning to run 7nm requires about 34 process steps, versus about 9 with EUV. Yield is low, around 40%, "just profitable," with high cost and constrained capacity (EDN 2025). This is an economic ceiling, not an absolute wall: SMIC can do it, but it is hard to do cheaply, at scale, and with a clear path below 5nm. One telling signal: on May 11, 2026, founder Richard Chang publicly said advanced processes account for less than 20% of demand, while 80% of demand lies in mature and specialty processes (TrendForce 2026-05-11). From the company's spiritual founder, this effectively corrects the "advanced-node champion" narrative toward "pragmatically doing mature and specialty processes."

7. Horizontal Peer Comparison: Compare Returns With TSMC, Positioning With Hua Hong

7.1 Group Portrait: What Each Foundry Has Become

Data sources: TSMC, Hua Hong, UMC/GFS SEC 6-K filings, TrendForce.

The business differences behind the numbers:

• Versus TSMC, the return gap: TSMC's 33x PE and 10.6x P/B are premiums earned with 36% ROE. SMIC's roughly 120x PE and 3.8x parent P/B correspond to only about 3% ROE. On PE, SMIC is about 4 times more expensive than TSMC, while its ROE is only about 1/12 of TSMC's. SMIC is not a "cheap TSMC." It is a very different asset, more expensive and much less profitable. TSMC's capex is high at USD 40.9 billion, but it can generate more than TWD 1 trillion of positive free cash flow. SMIC's capex consumes 90% of revenue and free cash flow is deeply negative.

• Versus Hua Hong, same theme and neither is cheap: One might have assumed Hua Hong was the "cheap China foundry comparison," but that premise has failed. Hua Hong's H shares have rerated sharply in 2026 to about 4.4x P/B. PE is distorted above 500x because earnings are close to zero and has limited value. Gross margin is only 11.8%, ROE is near zero, and Q2'25 even saw a net loss. The two Chinese foundries today are both priced as strategic/domestic-substitution options, not on cash returns, and Hua Hong is even more expensive than SMIC on P/B.

• Versus UMC/GlobalFoundries, mature nodes can also have higher gross margins: UMC at 29% and GlobalFoundries at 24.9% both have gross margins above SMIC's 21%, while capex intensity is much lower, about 22% for UMC. This shows SMIC's 21% gross margin is not high among mature-node peers. Its "advanced-node halo" has not translated into better overall gross margin.

7.2 Niche Conclusion

SMIC is a three-in-one asset: China's strategic foundry champion / the largest beneficiary of domestic substitution / a global cost-performance laggard. The market gap it fills is "local Chinese advanced + mature capacity." It most directly takes share from TSMC/UMC among Chinese customers. The most likely threat to its profit pool, however, is China's own rising number of new mature-node fabs such as Hua Hong, Nexchip, and Huali. If process substitution occurs, and SMIC is already blocked from EUV; if price war intensifies, and SMIC is already in it; or if demand weakens, its financial position will weaken. Its strategic position, supported by policy, is harder to shake.

8. Current Fundamentals and Bull-Bear Divide

8.1 Last Four Quarters: Volume Is There, Price Is Stable, Profit Is Thin

Revenue has risen steadily over the last four quarters, from USD 2.25 billion in Q1'25 to USD 2.49 billion in Q4'25 and USD 2.51 billion in 2026Q1. Gross margin has fluctuated narrowly in the 19-22.5% range, and utilization has stayed above 93% for a long period. The market reacted positively to the strong Q2'26 guidance of "revenue up 14-16% QoQ." Analysts have generally raised near-term revenue expectations, while staying cautious on gross margin and returns.

8.2 What the Market Is Trading Today

The current share price mainly trades three things: 1. domestic-substitution volume, high certainty; 2. AI pushing up mature-node pricing, being realized since the second half of 2025; 3. the strategic option value of advanced process / Ascend AI chips, high optionality and low visibility. It is necessary to separate "real fundamentals" from "market narrative." The first two have concrete support from utilization and pricing. The third is more option value, and it has been fully, perhaps excessively, priced into the high valuation.

8.3 Bull-Bear Divide (Each Point With Evidence)

Bulls argue: (a) SMIC is China's only answer for advanced process and is irreplaceable; (b) AI is making mature nodes "rise more as they are restricted," with full utilization and 10% price increases; (c) 7nm capacity is set to double in 2026, creating operating leverage; (d) the domestic-substitution runway is huge, with China's self-sufficiency rate around 30%, far below target.

Bears argue: (a) the EUV ceiling makes advanced process "possible but hard to profit from," with gross margin capped below 22% by depreciation; (b) capex consumes 90% of revenue, free cash flow has been USD -4.0 billion to USD -5.2 billion for years, and the company survives on state financing and share issuance; (c) profit quality is poor, subsidies are about 87% of operating profit, recurring profit is thinner, and another 30% leaks to minority shareholders; (d) valuation is completely disconnected from returns, sell-side consensus targets are roughly flat to slightly below the current price, about HKD 77.6-80, leaving limited upside; (e) geopolitical risk is not fully priced, especially the MATCH Act DUV service ban.

This is a contest between "strategic value" and "business returns." Both sides are right; they are pricing different things.

9. Valuation Analysis: Use P/B for Returns, Scenarios for Margin of Safety

9.1 Cash-Flow Look-Through (Required Before Any Valuation)

SMIC's operating cash flow / net profit ratio has been above 1 for the past five years because massive depreciation makes operating cash flow much higher than net profit. But capex is almost entirely expansionary. The huge expansion capex beyond maintenance capex pulls owner earnings deeply negative. On an owner-earnings basis, SMIC's free cash flow is negative, and headline PE cannot reflect its true cash-generation ability, which is really cash consumption. For SMIC, net-profit PE has limited meaning. P/B x ROE is the more honest yardstick.

9.2 Historical and Peer Valuation

• Historical: H-share PE is in the top decile of its 10-year history, with a median of 22x versus the current about 120x. P/B has risen from below 1x in FY23 to about 3.8x today on a parent-equity basis, near a multi-year high.

• Peers: TSMC trades at 10.6x P/B with 36% ROE, GlobalFoundries at 3.97x with 6.8% ROE, and UMC at about 3.9x with about 12% ROE. Behind these companies' 3-4x or even 10x P/B are meaningful ROE levels: TSMC 36%, UMC about 12%, GlobalFoundries about 7%. SMIC's about 3.8x parent P/B corresponds to only about 3% ROE, making it the worst mismatch between P/B and returns among peers. Warning: do not think SMIC is cheap just because Hua Hong is more expensive. Hua Hong's 4.4x P/B is also a theme-premium bubble. Both being expensive does not make either reasonable.

9.3 Absolute Valuation: Three Scenarios (All Endpoints Feed Into Section 10 Price Signals)

For a foundry, the best intrinsic-value anchor is P/B corresponding to achievable normalized ROE, supplemented by normalized earnings. Current book value attributable to owners is about USD 2.68 per share, or about HKD 20.9, and the current price of HKD 79.55 implies about 3.8x parent P/B. Three scenarios, intrinsic value per share in HKD:

Note: upside and downside are severely asymmetric. Even in the bull case, upside is only about +5% to +20%; the base case is already -15% to -30%; the bear case is a halving. The current price of HKD 79.55 sits between the upper end of the base case and the lower end of the bull case, effectively trading the bull case as the baseline. This is consistent with the observation that sell-side consensus target prices are roughly flat to slightly below the current price, with limited upside.

9.4 Expectations Gap

The market's current implied expectation is that "SMIC can significantly lift ROE from about 3%, and the strategic premium can persist for a long time." The indicators most likely to create a downside expectations gap are gross-margin guidance, because if it falls below 20%, the volume-growth logic will be questioned, and mature-node pricing, because if price increases fail, the narrative fades. In the next earnings report, the market will focus on whether gross margin can hold 20-22%, the progress of doubling 7nm capacity, and the depreciation pace.

9.5 Margin of Safety Review (Independent Discipline, Choose One of Four)

• The current price implies a large premium to the bear scenario of HKD 35-45, so the margin of safety is zero.

• The most fragile assumption in the three scenarios is "no mature-node price war + sustained strategic premium." If that assumption is cut by 30%, the base valuation falls toward the bear range, about HKD 40-50.

• If earnings are flat over the next three years, the current price at about 120x PE implies annualized returns far below the risk-free rate. There is no margin of safety at this buy price.

• This is a classic "good strategic company, bad price": worth waiting for a better price, not buying now.

• Margin-of-safety conclusion: none.

10. Risk Analysis

The key source of "permanent capital loss" to watch is not business disappearance, since strategic support makes bankruptcy unlikely, but a return of valuation from the top decile combined with a cyclical downturn. Heavy-asset cyclicals often fall sharply in this type of "double hit" from earnings and multiple compression.

11. Catalysts and Tracking Dashboard

Positive catalysts: gross margin stays above 22%+ for two consecutive quarters; mature-node price increases continue to be realized; 7nm capacity doubling lands and yield improves; Ascend AI chips scale; U.S.-China tensions ease temporarily, such as a one-year pause of the "50% affiliates rule" after the 2025-10 leaders' meeting.

Negative catalysts: gross-margin guidance is cut below 20%; mature-node price war becomes public; MATCH Act advances; another round of BIS controls; share issuance dilution; Big Fund and others reduce holdings.

Tracking dashboard (metric / why it matters / normal range):

• Quarterly gross margin (core profitability / 20-22% is normal, below 20% turns weaker, above 23% turns stronger)

• Capacity utilization (demand temperature / 90%+ healthy, below 85% warning)

• Capex and free cash flow (capital discipline / capex about USD 8 billion, FCF still negative, FCF turning positive would be a qualitative change)

• Government subsidies and recurring-profit share (profit quality / the lower subsidies are as a share of operating profit, the healthier)

• A/H premium (sentiment/valuation / currently about 80-95%, narrowing signals valuation cooling)

• Consensus target price vs current price (sell-side sentiment / currently roughly flat to slightly below the current price; clearly positive would mean analysts see upside)

• Export-control developments (structural constraint / MATCH Act, BIS updates, ASML service in China)

12. Zen Horizon Synthesis

12.1 What This Company Has Truly Proven

Longitudinally, SMIC has proved something remarkable and very "Chinese": with access to the most advanced equipment, EUV, cut off, it used national capital, technical leaders, and almost cost-insensitive capex to catch up to 7nm mass production and build the largest capacity base in China. Its success mainly comes from era tailwinds, namely domestic substitution as national policy, capital leverage from national-team money, and individual technical leaders such as Liang Mong Song, rather than sustainable business-model advantages or pricing power. Most of these factors remain today. Policy is stronger, money is still being invested, and Liang is still there. But they create "positioning," not "returns."

Horizontally, SMIC's true advantage versus peers is the unique intersection of "in China, advanced, and large-scale." Its weaknesses, low gross margin, negative free cash flow, and an EUV-capped roadmap, are mostly structural rather than temporary. TSMC converts its moat into 36% ROE. SMIC converts its moat into 3% ROE and strategic security. The gap cannot be closed by a cycle alone.

12.2 What Current Valuation Rewards, and What the Market May Be Misjudging

At about 120x PE and 3.8x parent P/B, current valuation rewards the identity of being "China's only answer for advanced process," rather than any cleanly calculable return. The market's most likely mistake is equating "irreplaceable strategic status" directly with "attractive shareholder returns." In SMIC, these two are separate. The more it is positioned as a national strategic tool, heavy capacity, light dividends, obedient to national policy, the more H-share financial returns may be suppressed for a long time. Even the founder has started correcting the narrative toward "advanced processes are only 20%; be pragmatic in mature processes," while the market is still paying about 120x PE for the advanced-process option.

12.3 Key Variables Over the Next 1/3/5 Years

• 1 year: whether mature-node price increases continue and gross margin can hold 22%, which decides whether the AI-cycle narrative is real.

• 3 years: the severity of price war after China's concentrated mature-node capacity additions, and whether valuation returns toward its historical center.

• 5 years: whether export controls tighten or loosen (EUV/DUV), and whether SMIC can find a profit model outside advanced process, such as specialty processes and advanced packaging, to lift ROE from about 3%.

12.4 Bull and Bear Cases (Traceable to Earlier Sections)

Bull case (3 points): 1. domestic-substitution volume is real and policy-driven, with FY25 shipments up 20.9%, utilization 93.5%, and China revenue share moving from 85% to 89% (§3.1, §1); 2. AI is making mature nodes "rise more as they are restricted," with SMIC fully loaded and some processes up about 10% (§6.2); 3. China's domestic monopoly in advanced process is irreplaceable, and Huawei Ascend has no alternative (§5.2).

Bear case (4 points): 1. the EUV ceiling is structural. DUV 7nm yield is about 40%, gross margin is capped below 22% by depreciation, there is no clear path below 5nm, and even the founder has shifted his rhetoric (§6.4); 2. returns and price are disconnected, with about 3% ROE against about 120x PE / 3.8x P/B, in the historical top decile, and even consensus target prices below the current price (§9); 3. profit quality is poor and cash burn persists, with subsidies about 87% of operating profit, free cash flow about USD -5.2 billion, and the dividend canceled (§3.3, §3.4); 4. geopolitical escalation risk, because if the MATCH Act DUV service ban lands, it threatens maintenance of the existing fleet (§6.4).

12.5 Pre-Mortem: If the Stock Is Down 50% in Three Years, What Is the Script?

Script A (no geopolitical escalation needed: price war + valuation derating): From the second half of 2026, 25+ new Chinese mature-node fabs ramp together and hit a demand air pocket. The AI mature-node price-up story proves one-off. SMIC's gross margin slides back to about 18%, and 2027 earnings miss consensus expectations built around "25% gross margin." The market stops paying a strategic premium, and PE compresses from about 120x to about 45x. Even if EPS is flat, about 60% valuation compression is enough to drive H shares from HKD 79.55 to HKD 35-40, about -50% to -55%. Early signals: gross-margin guidance below 20%, A/H premium narrowing, and Hua Hong losses widening.

Script B (geopolitical escalation: DUV service ban): The Senate companion to the MATCH Act advances, a DUV maintenance service ban is implemented, or FDPR expands after 150 days. SMIC's installed DUV fleet faces a maintenance cutoff, the 7nm doubling plan is shelved, utilization falls, depreciation idles, and Ascend foundry work is restricted again. Fundamentals and valuation both fall, and the narrative flips from "sanctions champion" to "sanctions target," taking the stock to the HKD 30s, down 50%+.

12.6 Final Research Conclusion

【Company Profile Score】

• Fundamental quality: Medium (record revenue and full utilization, but poor profit quality)

• Growth: Medium (volume growth is clear, price growth uncertain, returns weak)

• Moat: Strong (positioning-based) / weak (profitability-based)

• Financial soundness: Medium (balance sheet stable, but free cash flow deeply negative year after year)

• Management credibility: Medium (strong execution, but capital allocation follows national policy and is unfriendly to minority shareholders)

• Valuation attractiveness: Low

• Risk level: High

• Suitable investor type: theme/event-driven investors who understand geopolitics and cyclicality and can tolerate high volatility; not suitable for value investors seeking margin of safety and cash returns, and not suitable for ordinary investors.

【Investment Rating】

• Rating: Watch

• One-sentence investment thesis: A strategically irreplaceable domestic foundry leader, but the current price has capitalized that strategic status into about 120x PE against about 3% ROE, with no margin of safety and asymmetric downside.

• Three price signals (endpoints from §9.3 scenarios):

Ideal buy price: about HKD 35-45 (below the bear-scenario implied value, with margin of safety)

• Acceptable hold price: about HKD 55-68 (near the base-scenario implied value)

• Clearly overvalued price: above about HKD 88 (above the upper end of the bull scenario)

• Current price classification: clearly overvalued, with a downside bias. The current price of HKD 79.55 is above the upper end of the "acceptable hold" range and trades the bull case as the baseline. Combined with "consensus target price < current price" and "margin of safety = none," the risk-reward is heavily skewed downward.

• Worth waiting for a better price: yes. Buy triggers: the share price returns below about HKD 45, or about 1.7-2x parent book, or a real ROE inflection appears, with two consecutive quarters of gross margin above 23%+ and a trend toward positive free cash flow. Opportunity cost of waiting: possibly missing part of an AI mature-node supercycle upswing. But taking halving risk at about 120x PE to chase about +20% upside is not attractive.

• Target holding period: 3-5 years if entered at a reasonable price.

• Expected annualized return (based on §9.3): bear about -15% to -20%/year; base about -5% to -10%/year; bull about +5% to +7%/year. The weighted expected value across the three scenarios is negative.

• Maximum downside risk: based on the pre-mortem, the worst case is about -50% to -55% from price war + valuation derating, or a DUV service ban.

• Signals that trigger reassessment: 1. gross margin above 23% for two consecutive quarters and guidance revised up; 2. free cash flow shows a turning-positive trend, with capex peaking and operating cash flow rising; 3. export controls materially loosen (DUV service/EUV path) or materially tighten (MATCH Act enacted); 4. A/H premium narrows to within 30%, indicating valuation cooling; 5. dividend is restored after cancellation, showing capital allocation becoming more shareholder-friendly.

Again: this report is based on public information and is not investment advice. The rating refers to the risk-reward at the current price, not a denial of the company's strategic value. SMIC is an important company, just not a good investment at the current price.

13. Key Data Table

14. Research Uncertainties (Blind Spots)

• Exact FY2025 government-subsidy amount has not been separately disclosed: "subsidies ~= 87% of operating profit" uses the FY2024 HDIN basis. For FY2025, the recurring-profit gap of about RMB 900 million suggests the pattern continued, but the precise ratio awaits the original FY2025 annual report.

• Minority interests and JV fab structure are complex: the split between parent and total equity/profit relies on third-party data sources, stockanalysis + Yicai + Kristal. The direction is robust, but absolute values contain noise of plus/minus several percentage points.

• The sustainability of mature-node price increases is uncertain in both directions: whether AI-driven mature-node price increases, the core bull argument for 2026, or oversupply from 25+ new fabs, the core bear argument, dominates is the biggest variable from the second half of 2026 to 2027, and is difficult to determine today.

• The MATCH Act is a proposal, not law: the DUV service ban has not been enacted. After the 2025-10 leaders' meeting, the "50% affiliates rule" was paused for one year. Geopolitical direction is highly uncertain, and this report does not use it as the base case.

• The node and volume of Ascend foundry work are third-party estimates from SemiAnalysis/DigiTimes, not disclosures by SMIC or Huawei, and may be biased.

• A/H valuation methodology: third-party data vendors often calculate PE/P/B using total equity or lagging prices, such as stockanalysis's "2.86x P/B" on a total-equity basis. This report recalculates using parent-equity basis + base-date price, about 3.8x, to make it comparable with peer parent P/B.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.