China advanced packaging: China advanced packaging growth favors equipment vendors and selective OSAT share gainers

Overview

The report examines China’s advanced-packaging value chain amid AI-driven demand growth. Morgan Stanley distinguishes a positive long-term market opportunity from a more difficult near- to medium-term supply-demand outlook for OSATs, favoring equipment providers and selected packaging companies with credible share-gain pathways.

Core views

Morgan Stanley expects China’s advanced-packaging market to reach approximately Rmb100bn by 2029, implying roughly 13% CAGR from 2025-29. The chiplet/2.5D segment is expected to grow faster, reaching Rmb17.7bn in 2029 at a 40% CAGR. The report argues that advanced packaging is strategically important because it raises compute density, memory bandwidth and system integration for AI accelerators; this role is especially significant in China, where constraints on leading-edge front-end equipment increase the value of packaging-led performance improvements. The central debate is whether rapid industry growth will translate into OSAT earnings. Morgan Stanley is more cautious than consensus because it expects China’s 2.5D annual capacity to expand from around 120k wafers in 2025 to 436k in 2027, while effective demand is constrained by qualified accelerator-die yields, HBM availability, customer qualification and other upstream inputs. Its bottom-up demand assumptions include Chinese-vendor AI GPU shipments rising from 1.1mn units in 2025 to 4.9mn in 2027; assuming roughly 21 chips per wafer and an 85% overall yield, this implies around 277k wafers of 2027 demand. The resulting estimated utilization rate of about 63% points to moderate overcapacity, weaker pricing and limited margin expansion despite revenue growth. The report further argues that capacity mix matters as much as headline supply. Domestic capacity initially focuses on CoWoS-S-like solutions, while newer large AI packages increasingly favor CoWoS-L-like architectures. China could therefore have excess S-like capacity while L-like capacity remains relatively tight. Flexible lines can partly convert between architectures, but differences in process flow, equipment and customer qualification mean capacity is not immediately interchangeable. A shift toward vertically stacked 3D integration could also reduce 2.5D package footprint per accelerator die, weakening the link between AI GPU shipment growth and 2.5D demand. Morgan Stanley therefore prefers back-end equipment vendors to broad OSAT exposure. OSAT earnings depend on filling capacity, whereas equipment revenue depends chiefly on capacity being built. The transition to 2.5D/3D raises equipment intensity through fine-pitch RDL, TSV, wafer-level processing, temporary bonding/debonding, high-precision placement, thermo-compression or hybrid bonding, and advanced inspection. ACM Research is highlighted for plating, cleaning and wet-process exposure: advanced-packaging revenue excluding ECP grew 153% year-on-year in 2Q26, while ECP, furnace and other technologies grew 168%. ASMPT is favored for TCB, flip-chip and hybrid-bonding tools; its TCB revenue grew about 146% year-on-year in 2025, while advanced-packaging revenue reached a record US$339mn in 1H26, up 17% year-on-year and representing 30% of group revenue. Among OSATs, the report favors JCET as a share gainer. JCET is building 2.5D capacity from a low base and is expected to gain share as its Jiangyin facility ramps, customer visibility improves and it broadens into computing, automotive, CPO and next-generation packaging. Jiangyin revenue reached Rmb212mn in 1H26, with a Rmb89mn net loss that narrowed as utilization improved; management expects profitability in 2027, and Morgan Stanley estimates end-2027 capacity of about 5kwpm. JCET’s Rmb7.8bn Shanghai Lingang advanced-packaging project adds a second growth leg. In 1H26, computing revenue rose 40.4% year-on-year and reached 30% of revenue from 22% in 1H25; automotive revenue rose 25.0% to Rmb2bn and reached 11% of mix from 9%. For 2025, China revenue grew about 22% to Rmb8.3bn and domestic gross margin improved 6.6 percentage points to 20.4%, which the report interprets as higher-value mix-driven share gains rather than price-led competition. Morgan Stanley upgrades JCET to Overweight and raises its price target to Rmb90.2. It raises 2026/27/28 EPS estimates by 2%/3%/12%, reflecting stronger AI and AI-peripheral chip demand, higher sales forecasts and favorable OSAT pricing, partly offset by higher interest expense from capacity-expansion bonds. The target is based on a residual-income model with 8.6% cost of equity, 19.0% intermediate growth and 4.0% terminal growth. The report’s bull and bear values are Rmb147.7 and Rmb60.6, respectively, depending principally on the pace of 2.5D/3D business ramp. SJ Semiconductor is initiated at Equal-weight with a Rmb118.4 price target. Morgan Stanley recognizes SJ Semi as China’s leading 2.5D vendor, with about 77% market share in 2025, a roughly 30% gross margin in 1H26, and a first-mover position in domestic 3DIC commercialization. It forecasts 25% revenue CAGR and 34% earnings CAGR over 2025-28, with chiplet/2.5D packaging expected to grow at 36% CAGR. However, competition could reduce SJ Semi’s built-capacity share to 34% by 2027, Huawei represented about 74% of 1H25 revenue, and tighter HBM supply or slower leading-node yield improvement could limit accelerator shipments. Morgan Stanley’s 2027/28 earnings estimates are 6%/4% below consensus, and it considers the valuation demanding: SJ Semi trades at 31x/24x/19x 2026-28E P/S, above OSAT and equipment peers. Its Rmb118.4 target uses a residual-income model with 8.0% cost of equity, 50% payout, 26.0% medium-term growth and 5.0% terminal growth. For emerging technologies, the report sees 3DIC, HBM packaging and CPO as strategically important but not major near-term OSAT earnings drivers. China’s 3DIC market is estimated to remain only in the tens of millions of US dollars over the next two to three years because of high cost and compound-yield sensitivity. HBM is a substantial volume opportunity, but packaging value is largely retained within the memory ecosystem; CXMT is estimated to produce 3-4mn HBM3E stacks in 2027, which could support only 0.75-1mn GPUs. CPO has long-term potential, but bottlenecks in fiber alignment, thermal management, testing and supply-chain integration make large-scale deployment more likely in 2027-28, limiting its contribution to China OSAT earnings over the next one to two years.

Analysis framework

The report combines a bottom-up capacity-and-demand analysis with value-chain reasoning. It compares projected 2.5D capacity, effective demand constraints and utilization, separates equipment spending from OSAT earnings capture, evaluates technology transitions and then applies company-specific operating, competitive and residual-income valuation analysis to JCET and SJ Semi.

Methodology notes

• Industry AnalysisSupply-demand framework

  Bottom-up 2.5D packaging capacity, effective-demand and utilization analysis

  Morgan Stanley compares projected capacity additions with demand limited by accelerator-die yields, HBM availability and qualifications to assess overcapacity and pricing risk.

• Industry AnalysisUpstream-Midstream-Downstream Transmission

  Equipment vendors versus OSAT value capture

  The report distinguishes equipment revenue, which is linked to capacity construction, from OSAT profitability, which requires capacity utilization and pricing to hold.

• Valuation methodsRIM (Residual Income Model)

  Residual income valuation for JCET and SJ Semiconductor

  The price targets are derived from projected profits and equity returns relative to the cost of equity, with stated intermediate and terminal growth assumptions.

Asset mapping & comparison

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

• ACM Research (ACMR.O)

  Direct beneficiary of China advanced-packaging capacity expansion through plating, cleaning and wet-process tools.

  Strengths

  Advanced-packaging revenue excluding ECP grew 153% year-on-year in 2Q26; first production order for a 510×515mm horizontal panel-level electroplating system.

  Comparison

  Preferred as a capex beneficiary rather than broad OSAT exposure.

• ASMPT (0522.HK)

  Beneficiary of rising demand for TCB, flip-chip and hybrid-bonding equipment.

  Strengths

  TCB revenue grew about 146% year-on-year in 2025; advanced-packaging revenue was US$339mn in 1H26.

  Comparison

  Preferred equipment exposure to advanced-packaging investment.

• JCET Group Co Ltd (600584.SS)

  Preferred domestic OSAT and expected 2.5D share gainer.

  Strengths

  Low-base 2.5D ramp, improving customer visibility, computing and automotive mix expansion, and Lingang capacity investment.

  Weaknesses

  2.5D/3D ramp remains in development and requires sustained execution.

  Comparison

  Preferred over SJ Semi for expected market-share gains.

  Risks

  Economic weakness, slower technology progress, or price competition and local-customer share loss.

• SJ Semiconductor Corp (688820.SS)

  Leading domestic 2.5D and 3DIC player with balanced risk-reward.

  Strengths

  ~77% China 2.5D share in 2025, leading 3DIC commercialization and superior gross margin profile.

  Weaknesses

  Premium valuation, high customer concentration and potential competitive share loss.

  Comparison

  Technology leader but less preferred than JCET because much of the growth outlook is already priced in.

  Risks

  Capacity competition, HBM and yield constraints, Huawei outsourcing risk, and potential export-control disruption.

Key data

• China advanced packaging TAMApproximately Rmb100bn by 2029About 13% CAGR over 2025-29.
• Chiplet/2.5D marketRmb17.7bn by 2029Expected 40% CAGR over 2025-29.
• China 2.5D capacity120k wafers in 2025 to 436k in 2027Morgan Stanley expects moderate overcapacity.
• Estimated 2027 utilization~63%Based on projected supply and effective demand.
• JCET target priceRmb90.2Overweight; EPS estimates raised 2%/3%/12% for 2026/27/28.
• SJ Semi target priceRmb118.4Equal-weight initiation; projected 25% revenue CAGR over 2025-28.

Impact & implications

Morgan Stanley’s main implication is that AI-driven advanced-packaging growth should not be treated as uniform OSAT earnings upside. Capacity expansion and rising process complexity offer cleaner exposure for equipment vendors, while OSAT outcomes depend on utilization, pricing, technology mix and company-specific share gains. JCET is preferred among domestic OSATs, while SJ Semi’s leadership is viewed as largely reflected in its valuation.

Risks

• China AI compute demand, domestic advanced-node capacity or HBM availability could grow faster than Morgan Stanley expects, tightening utilization and improving OSAT margins.
• A slower shift from 2.5D toward 3D integration could preserve more 2.5D packaging demand than assumed.
• OSAT technological breakthroughs could lower 3DIC cost and improve margins.
• JCET faces risks from weaker communications, computing and consumer-electronics demand, delayed advanced-packaging ramp, and price competition.
• SJ Semi faces competitive capacity additions, Huawei customer concentration, HBM and leading-node bottlenecks, and possible expansion of U.S. export controls.

What to watch

• Domestic advanced-packaging capacity, especially separate CoWoS-S-like and CoWoS-L-like capacity, utilization and customer qualifications.
• Domestic advanced-node yields, HBM availability and Chinese AI GPU shipment growth.
• OSAT capex, margin trends, revenue and earnings growth.
• 3DIC, HBM and CPO customer qualification, dedicated capacity, utilization and disclosed revenue contribution.
• SJ Semi’s October 21, 2026 IPO lock-up expiration, late-October 3Q26 results, and 4Q26-1Q27 capacity and customer-qualification progress.