AI, self-sufficiency, and localized supply chains jointly drive semiconductor capacity expansion and technological upgrades in China

Overview

This is the fourth report in Goldman Sachs’ “China Semis CHIPS” series. Its core focus is the progress of China’s semiconductor supply chain toward self-sufficiency in equipment, advanced process nodes, AI chips, advanced packaging, and memory, as well as the impact of continued capacity expansion on Chinese and overseas suppliers. The report believes supply-chain gaps are narrowing and that generative AI and customer supply-chain localization will support long-term capital expenditure, while noting that lithography, R&D investment, export controls, and advanced-product yields remain the main constraints.

Core views

The report first reviews the progress of China’s semiconductor supply chain over the past three years. Equipment supply has expanded from etching and deposition into high-end models such as high-aspect-ratio etching and atomic layer deposition, while gradually adding categories such as ion implantation, inspection, and metrology. Wafer manufacturing is progressing toward 7nm and below, while chiplets and advanced packaging are being used to improve production yields. Planned AI chip computing capability is evolving from 320 TFLOPS (FP16) in 2023 to 4 PFLOPS (FP8) by the fourth quarter of 2028, while SuperPods enhance system-level computing power by increasing the scale of chip interconnections. Memory products are being upgraded from LPDDR4(X) in 2024 to LPDDR6, HBM3, and HBM3E by the fourth quarter of 2026, while EDA is expanding from analog design into digital place-and-route, physical verification, and memory EDA. Three forces are driving this upgrade: generative AI, the “local serves local” supply-chain strategy, and the industry’s continued commitment to raising self-sufficiency. Generative AI requires more advanced process nodes, packaging, and high-end memory, increasing the capital intensity per IC. Global fabless companies are increasing local production to reduce exposure to foreign-exchange movements, pandemics, and geopolitical risks. Advanced capacity at leading international suppliers is also being absorbed by AI demand, creating substitution opportunities for Chinese suppliers in conventional ICs and memory. China accounts for approximately 30% of global semiconductor demand, and large domestic brands in smartphones, PCs, televisions, and automobiles further support the adoption of locally produced components. Significant structural gaps in industry self-sufficiency remain. China’s IC production self-sufficiency rate increased from 38% in January 2010 to 70% in June 2026, but the value-based supply-demand gap remains large due to insufficient lithography capabilities. Even though leading domestic companies are rapidly increasing R&D and capital expenditure, the gap versus global leaders remains wide. The report therefore believes substantially more capital remains necessary to catch up with global leaders and achieve global operations. Capital expenditure is the report’s second major theme. After posting consecutive double-digit growth in 2023—2024, Chinese semiconductor capital expenditure was flat in 2025. Goldman Sachs expects YoY growth to recover to 13%/15%/15%/10%/10% in 2026—2030, reaching US$82bn in 2030. Compared with the prior forecast from August 2025, capital expenditure estimates for 2026—2030 were raised by 15%/32%/51%/64%/79%, respectively. Foundries and memory manufacturers are expected to contribute 80%—85% of China’s capital expenditure over the next several years. The report’s charts indicate that 6m wpm of 8-inch-equivalent and 12-inch-equivalent capacity will be added in 2025—2030, increasing from 7m wpm to 14m wpm. Based on the main text’s estimates for 8-inch and 12-inch production lines, capacity is approximately 6m wpm in 2025 and reaches 14m wpm in 2030, corresponding to approximately 7m wpm of newly built capacity. Equipment required for advanced process nodes and high-end memory is more technologically complex and carries higher average selling prices, meaning capacity growth will generate higher capital expenditure. Wafer fabrication equipment spending will also accelerate. The report expects China’s WFE spending to grow 13%/20%/15% YoY in 2026—2028, with WFE revenue rising 20% to US$53bn in 2027. China’s share of global WFE spending is expected to reach 29%—31% in 2026—2028, above 22% in 2022. Revenue generated by Chinese equipment companies in the domestic WFE market is expected to increase from US$10.2bn in 2025 to US$13.5bn, US$17.8bn, and US$22.9bn in 2026—2028. The value-based share of domestic equipment is projected to rise from 26% in 2025 to 31%/34%/38% in 2026—2028. Localization benefits Chinese equipment vendors, but global suppliers remain key participants due to their leading technologies and intellectual property. For 7nm and below, Goldman Sachs separately estimates equipment shipments, advanced-node penetration, and die area across eight end markets. It expects related wafer demand in China to grow at a 17% CAGR in 2025—2035, reaching 619k wpm in 2035. AI servers are the main driver, with their demand expected to grow at a 42% CAGR over the same period. The supply-side model uses two assumptions: capacity expansion plans and yield ramp-up. Domestic supply at 7nm and below is expected to grow at a 46% CAGR to 410k wpm by 2035. SMIC is the main contributor, adding 30—50k wpm annually in 2026—2031 and 20k wpm annually in 2032—2035. Yields are expected to rise from 23% in 2026 to 50% in 2030 and reach 75% in 2035. Under these assumptions, the domestic supply-demand gap declines from 92% in 2025 to 34% in 2035. The report constructs estimates of China’s GPU and DRAM market opportunities for the first time. Under the bull, base, and bear cases, China’s AI chip TAM is expected to grow at CAGRs of 142%, 69%, and 6%, respectively, in 2025—2030, reaching US$4,123bn, US$678bn, and US$66bn in 2030. Corresponding shipments are 237m, 39m, and 4m chips, while IT electricity consumption is 196GW, 32GW, and 3GW, respectively. Under the AI chip base case, China’s DRAM TAM is expected to grow at a 50% CAGR in 2026—2028 to US$257bn, with HBM growing at a 188% CAGR to US$32bn. Expanding demand also requires AI chip, advanced packaging, and memory suppliers to increase capital investment and improve yields. For global equipment companies, Goldman Sachs expects global WFE to grow 36% and 45% in 2026 and 2027, respectively. GAA logic structures, stacked DRAM/NAND, and advanced packaging will drive faster growth in etching and deposition equipment. Goldman Sachs relatively prefers AMAT, LRCX, and ONTO, all rated Buy. US equipment companies are expected to maintain high but gradually declining market shares in China due to additional export controls affecting sales from the second half of 2025 and Chinese capital expenditure increasingly favoring domestic suppliers. Nevertheless, supported by stable Chinese capital expenditure, Goldman Sachs expects long-term shipments to China to remain at the upper end of the mid-20% revenue-share range indicated by company guidance. Japanese equipment companies are similarly highly exposed to Chinese demand. The combined China revenue of Tokyo Electron, SCREEN, Kokusai, and Ebara was ¥1.21tn in FY2025, representing 35% of their semiconductor equipment revenue, down from ¥1.44tn and 41% in FY2024. The decline may reflect advance purchases ahead of export restrictions and competition from domestic Chinese equipment vendors. All four companies expect the absolute value of China revenue to increase in FY2026 and beyond, while Goldman Sachs also forecasts revenue growth in FY2026—FY2028. However, because non-China operations are growing faster, the share of China revenue is expected to decline moderately. In back-end equipment, Disco, Advantest, and Tokyo Seimitsu benefit from Chinese OSAT manufacturers building AI ecosystems, provided there are no major additional export restrictions in the future. For European equipment companies, Goldman Sachs reiterates Buy ratings on ASML, ASMI, and BESI. China is expected to contribute approximately 20% of ASML’s group revenue in 2026, at the upper end of its historical 15%—20% range. New fab construction and Chinese memory expansion support lithography demand, and the company is also studying an increase in annual EUV capacity to 110 systems by 2028. ASML’s 12-month target price is €2,200, based on 32x 2HCY27+1HCY28E P/E. ASMI benefits from demand for deposition equipment from Chinese mature-logic and foundry customers, with strong orders in the first half of 2026 and solid demand visibility for 2027. Its target price is €955, based on 20x 2HCY27+1HCY28E EV/EBITDA. BESI benefits from advanced-packaging capacity expansion required for 2.5D CoWoS, photonics, and pluggable modules, and is seeing signs of recovery in Chinese industrial, automotive, and high-end smartphone module demand. Its target price is €325, based on 29x 2HCY27+1HCY28E EV/EBITDA. Finally, the report initiates coverage of leading Chinese DRAM producer CXMT (688825.SS) with a Buy rating and a 12-month target price of Rmb129. The company listed in China in July 2026 and raised Rmb58bn (US$8bn). Goldman Sachs expects its capacity to reach 665k wpm by 2030, more than doubling from 2026. CXMT’s conventional DRAM supply as a proportion of Samsung’s and SK Hynix’s supply is expected to rise from 28%/35% in 2025 to 41%/50% in 2028, while covering 50% of China’s DRAM demand in 2028. High conventional DRAM prices and global manufacturers shifting capacity toward HBM are prompting US smartphone and PC customers to seek additional suppliers. Goldman Sachs therefore believes some customers may qualify CXMT’s mobile and conventional DRAM products. However, CXMT’s process technology remains several generations behind global peers, and its initial HBM technology is also insufficiently mature, making it difficult to enter US customers through HBM in the short to medium term. CXMT’s earnings forecasts reflect high operating leverage from capacity expansion, demand growth, and product-mix upgrades. The report’s tables project revenue rising from Rmb61,799m in 2025 to Rmb320,756m in 2026, Rmb661,563m in 2027, Rmb878,714m in 2028, Rmb1,369,487m in 2030, and Rmb1,663,927m in 2031. Net profit is projected to increase correspondingly from Rmb1,875m to Rmb161,373m, Rmb361,256m, Rmb472,573m, Rmb744,511m, and Rmb863,644m. Diluted EPS is expected to rise from Rmb2.5 in 2026 to Rmb5.4 in 2027, Rmb7.1 in 2028, Rmb8.9 in 2029, Rmb11.1 in 2030, and Rmb12.91 in 2031. Average EPS growth in 2027—2028 is 77%. The target price corresponds to 24x 2027E P/E, with subsequent implied P/E multiples in the table declining to 18x, 14x, 12x, and 10x. Key uncertainties stem from geopolitics and import/export restrictions, the technological maturity of advanced products, and the impact of aggressive capacity expansion on the global DRAM supply-demand balance in 2028.

Analysis framework

Goldman Sachs first updates the coverage map of Chinese semiconductor equipment and products to identify the supply-chain segments that have been filled and those that remain missing. It then maps AI demand, localized supply chains, and self-sufficiency targets to capital expenditure and capacity forecasts. Advanced-node demand is estimated bottom-up based on shipments, penetration rates, and die area across eight end markets, while supply is derived from planned capacity and yield ramp-up. For GPUs, bull, base, and bear cases are used to estimate TAM, shipments, and electricity consumption, after which the base case is used to forecast DRAM and HBM demand. Finally, the industry forecasts are mapped to Chinese, US, Japanese, and European equipment companies, with target prices for key companies set using forward P/E or EV/EBITDA.

Methodology notes

• Industry/Sector Analysis FrameworkUpstream-Midstream-Downstream Supply-Chain Transmission

  Semiconductor ecosystem map and supply-chain gap analysis

  The report reviews domestic supply capabilities layer by layer across IC design, wafer foundries, equipment, packaging and testing, memory, and downstream customers, and analyzes how AI and localized procurement transmit through each segment.

• Industry/Sector Analysis FrameworkSupply-demand framework

  Supply-demand model for wafers at 7nm and below

  Demand is derived from shipments, advanced-node penetration, and die area across eight end markets, while supply is derived from new capacity and yield ramp-up. The difference between the two measures the domestic supply-demand gap.

• Industry/Sector Analysis FrameworkVolume-price decomposition

  Bottom-up estimation of AI chip and equipment markets

  The report separately estimates AI chip shipments, computing power, and electricity consumption, and combines equipment categories, capacity, and average selling prices to estimate WFE revenue and domestic suppliers’ shares.

• (Out-of-Vocabulary Method)

  Bull, base, and bear case analysis

  The report uses three demand scenarios to present ranges for China’s AI chip TAM, shipments, and IT electricity consumption in 2030, and uses the base case as the premise for its DRAM demand forecast.

• Valuation MethodP/E and PEG Valuation

  Forward P/E valuation

  CXMT’s target price is based on 24x 2027E P/E, while ASML’s target price is based on 32x 2HCY27+1HCY28E P/E.

• Valuation MethodEV/EBITDA valuation

  Enterprise-value multiple valuation

  The target prices for ASMI and BESI are based on 20x and 29x 2HCY27+1HCY28E EV/EBITDA, respectively.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• CXMT (688825.SS)

  A key beneficiary of Chinese DRAM capacity expansion, rising self-sufficiency, and AI memory demand; the report initiates coverage with a Buy rating.

  Strengths

  Capacity is expected to reach 665k wpm in 2030, covering 50% of China’s DRAM demand in 2028, while its conventional DRAM products may be qualified by overseas smartphone and PC customers.

  Weaknesses

  Its process technology is several generations behind global memory peers, and HBM remains at an early stage of technological maturity.

  Comparison

  CXMT’s conventional DRAM supply as a proportion of Samsung’s and SK Hynix’s supply is expected to rise from 28%/35% in 2025 to 41%/50% in 2028.

  Risks

  Geopolitics and import/export restrictions may affect overseas shipments; aggressive capacity expansion may alter the global DRAM supply-demand balance in 2028.

• Chinese Domestic Semiconductor Equipment Suppliers

  Beneficiaries of rising Chinese capital expenditure, equipment category expansion, and localized procurement by wafer fabs.

  Strengths

  Their share of China’s WFE market is expected to increase from 26% in 2025 to 38% in 2028, while product coverage is expanding from etching and deposition into ion implantation, inspection, and metrology.

  Weaknesses

  They remain dependent on or behind global suppliers in lithography, high-end technologies, and intellectual property.

  Comparison

  Domestic WFE revenue is expected to rise from US$10.2bn in 2025 to US$22.9bn in 2028, although global suppliers remain key market participants.

  Risks

  R&D and capital expenditure remain below those of global leaders, while qualification and yield improvement for advanced products require time.

• AMAT, LRCX, ONTO

  Beneficiaries of global WFE growth and equipment demand arising from GAA, stacked memory, and advanced packaging.

  Strengths

  They have leading capabilities in high-technology segments such as etching, deposition, and metrology. Goldman Sachs rates all three Buy and expresses a relative preference for them.

  Weaknesses

  Their market shares in China are expected to decline as export controls and purchases of domestic equipment increase.

  Comparison

  Global WFE is expected to grow 36% and 45% in 2026 and 2027, with etching and deposition equipment expected to deliver more pronounced growth.

  Risks

  Additional US export controls and a further shift by Chinese customers toward domestic WFE suppliers.

• Tokyo Electron, SCREEN, Kokusai, Ebara

  Capacity expansion in Chinese advanced process nodes and high-end memory supports orders and revenue for Japanese front-end equipment companies.

  Strengths

  Yield requirements at advanced nodes lead Chinese wafer fabs to continue purchasing non-Chinese equipment.

  Weaknesses

  China revenue declined from 41% of total revenue in FY2024 to 35% in FY2025 and is expected to continue declining moderately.

  Comparison

  The four companies generated combined China revenue of ¥1.21tn in FY2025, down from ¥1.44tn in FY2024, but the report expects absolute revenue to recover in FY2026—FY2028.

  Risks

  Further export restrictions and competition from domestic Chinese equipment vendors may pressure revenue and market share.

• ASML (ASML.AS)

  A beneficiary of Chinese fab construction and domestic lithography demand; Goldman Sachs reiterates its Buy rating.

  Strengths

  Leading lithography technology; China is expected to contribute approximately 20% of group revenue in 2026, while the company is studying an increase in annual EUV capacity to 110 systems in 2028.

  Weaknesses

  Chinese wafer fabs still need to absorb lithography equipment delivered in recent years, and the timing of future revenue depends on new fab construction.

  Comparison

  China revenue is expected to represent approximately 20% of total revenue in 2026, at the upper end of the historical 15%—20% range.

  Risks

  EUV delays, capital expenditure cyclicality, and adverse market-share changes.

• ASMI (ASMI.AS)

  A beneficiary of Chinese mature-logic and foundry demand for deposition equipment; Goldman Sachs reiterates its Buy rating.

  Strengths

  Chinese orders were strong in the first half of 2026, and the regional sales mix benefits revenue and gross margin.

  Weaknesses

  Demand visibility for 2027 still depends on customer behavior and the regulatory environment.

  Comparison

  12-month target price of €955, based on 20x 2HCY27+1HCY28E EV/EBITDA.

  Risks

  A deterioration in the semiconductor cycle, stronger-than-expected competition, and high customer concentration.

• BESI (BESI.AS)

  A beneficiary of advanced-packaging capacity expansion required for 2.5D CoWoS, photonics, pluggable modules, and high-end smartphones; Goldman Sachs reiterates its Buy rating.

  Strengths

  Strong advanced-packaging demand and initial signs of recovery in Chinese industrial and automotive demand.

  Weaknesses

  Recovery in the conventional packaging portfolio remains dependent on end-market demand.

  Comparison

  12-month target price of €325, based on 29x 2HCY27+1HCY28E EV/EBITDA.

  Risks

  Cyclicality in customer spending, delayed adoption of hybrid bonding, and intensifying competition.

Key data

• China’s IC production self-sufficiency rate70%June 2026, significantly higher than 38% in January 2010
• Chinese semiconductor capital expenditureUS$82bn2030E; 2026—2030E YoY growth is expected to be 13%/15%/15%/10%/10%
• Capital expenditure forecast upgrades15%/32%/51%/64%/79%Upgrades to 2026—2030E, respectively, versus the August 2025 forecast
• China WFE revenueUS$53bn2027E, up 20% YoY
• China’s share of global WFE spending29%—31%2026—2028E; 22% in 2022
• Chinese domestic WFE equipment share31%/34%/38%2026—2028E, calculated by value of China WFE revenue; 26% in 2025
• Wafer demand at 7nm and below619k wpm2035E; 2025—2035E CAGR of 17%
• Wafer supply at 7nm and below410k wpm2035E; 2025—2035E CAGR of 46%
• Domestic supply-demand gap at 7nm and below34%2035E, down from 92% in 2025
• Advanced-node yield assumptions23%/50%/75%Corresponding to 2026E, 2030E, and 2035E, respectively
• China AI chip TAMUS$4,123bn/US$678bn/US$66bn2030E bull/base/bear cases, corresponding to 2025—2030E CAGRs of 142%/69%/6%
• China AI chip shipments237m/39m/4m2030E bull/base/bear cases
• IT electricity consumption related to Chinese AI chips196GW/32GW/3GW2030E bull/base/bear cases
• China DRAM TAMUS$257bn2028E; 2026—2028E CAGR of 50%
• China HBM TAMUS$32bn2028E; 2026—2028E CAGR of 188%
• CXMT capacity665k wpm2030E, more than double the 2026E level
• CXMT coverage of China’s DRAM demand50%2028E
• CXMT target price and valuationRmb129; 24x 2027E P/E12-month target price; average 2027—2028E EPS growth of 77%
• CXMT diluted EPSRmb2.5/5.4/7.1/8.9/11.1/12.91Corresponding to 2026E—2031E, respectively
• Global WFE growth36%/45%Corresponding to 2026E and 2027E, respectively
• China revenue of four Japanese front-end equipment companies¥1.21tn, 35% of revenueFY2025; ¥1.44tn and 41% in FY2024

Impact & implications

The report believes China’s semiconductor capacity expansion will benefit both domestic and overseas equipment suppliers, but with different benefit profiles. Domestic vendors will increase their shares through category expansion and localized procurement, while global vendors will continue participating in the market through leading technologies required for advanced process nodes and yield improvements. AI demand for advanced logic, HBM, and advanced packaging will raise capital intensity per unit of capacity and drive demand for etching, deposition, metrology, and back-end equipment. In the memory market, CXMT’s capacity expansion will help narrow China’s DRAM supply-demand gap, but will also become an important variable for global investors assessing the industry’s supply-demand balance in 2028.

Risks

• China has insufficient lithography capabilities, while the R&D and capital expenditure of domestic leaders remain significantly below those of global leaders, potentially limiting improvements in value-based self-sufficiency.
• China-US semiconductor import/export restrictions and other geopolitical measures may affect equipment, GPUs, EDA, memory technology, and CXMT’s overseas shipments.
• US equipment companies may continue losing Chinese market share due to export controls and Chinese wafer fabs shifting toward domestic suppliers.
• CXMT’s process technology is several generations behind global peers, and its HBM technology has low maturity, making it difficult to enter US customers through HBM in the short to medium term.
• CXMT’s aggressive capacity expansion may materially affect the global DRAM supply-demand balance in 2028.
• ASML faces risks from EUV delays, capital expenditure cyclicality, and adverse market-share changes.
• ASMI faces risks from a deterioration in the semiconductor cycle, stronger-than-expected competition, and customer concentration.
• BESI faces risks from customer spending cyclicality, delayed hybrid-bonding adoption, and intensifying competition.

What to watch

• Monitor the upcoming China-US political summit and other trade discussions to assess whether additional restrictions may emerge in semiconductors, memory, GPUs, equipment, and EDA.