Morgan Stanley Initiates ACM Research: China's Advanced Packaging to Reach ¥100B by 2029, Equipment Makers Favored Over OSAT Players
nashnova research
Morgan Stanley initiates coverage of ACM Research with an Overweight rating and a $130 price target — implying 94% upside; its core call is that China's advanced packaging market will reach ¥100 billion by 2029, and in this capex cycle, equipment vendors will outperform the packaging houses.
Where does the ¥100 billion market come from?
Morgan Stanley sizes China's advanced packaging market at roughly ¥100 billion by 2029, a ~13% CAGR from 2025.
The fastest segment is Chiplet/2.5D — assembling multiple small dies into one package — reaching ¥17.7 billion at a 40% CAGR by 2029.
This means → the broad market grows steadily, but the real explosion is in 2.5D, where capital and capacity are concentrating.
Why bet on equipment makers over OSATs?
The logic is straightforward: OSATs profit when capacity is full; equipment vendors profit when capacity is being built. In an expansion cycle, equipment vendors get paid first.
The numbers confirm it: global OSAT capex rises 65% YoY in 2025 and an estimated 67% in 2026, pushing capex-to-revenue ratios to historic highs — money is flooding into equipment purchases.
In plain terms = in a gold rush, the shovel sellers collect first; whether OSATs strike gold depends on whether orders eventually fill the lines they are building.
Why is ACM Research the top pick?
ACM Research occupies the electroplating and cleaning segments — wet-process steps where liquid chemicals treat wafer surfaces — exactly the nodes seeing the steepest rise in equipment value per unit of 2.5D/3D and panel-level capacity.
In Q2 2026, its advanced packaging revenue (excluding electrochemical copper plating) grew 153% YoY, and it secured the first 510 × 515 mm horizontal panel-level plating tool production order from a mainland Chinese customer.
This means → panel-level packaging has crossed from "near production" to "real production," and ACM sits right at that inflection. Morgan Stanley's $130 target carries a risk-reward ratio of 6.3, the highest across its coverage.
How big is the overcapacity risk?
China's 2.5D annual capacity will surge from 120,000 wafers in 2025 to 436,000 in 2027 — roughly 16% of overseas capacity led by TSMC.
But demand faces two bottlenecks: advanced-node yields are still low, and high-bandwidth memory (HBM — high-speed memory paired with AI chips) is scarce. Domestically produced HBM3E totals only 3–4 million units, enough for just 0.75–1 million GPUs; the rest still relies on imports.
Combined, Morgan Stanley estimates 2027 utilization at just ~63%. In plain terms = a lot of lines are being built, but orders may fill only six out of ten — squeezing OSAT margins hard.
What does "S surplus, L shortage" mean?
Most Chinese capacity starts with CoWoS-S — TSMC's 2.5D packaging technology using a silicon interposer to link multiple dies — but next-generation products are shifting to CoWoS-L, which swaps the silicon for a larger organic substrate with a different cost structure.
This reflects a structural mismatch: the S-type capacity already built may sit idle while the L-type capacity the market actually needs remains tight.
This means → not every wafer start that gets built will find a customer; the "build it and orders will come" thesis does not hold across the board.
3DIC, HBM packaging, and co-packaged optics — too early to price in?
Morgan Stanley sees 3DIC constrained by cost and compound yield; the China market will remain in the tens-of-millions-of-dollars range for the next two to three years.
HBM packaging value stays largely inside the memory ecosystem (Samsung, SK Hynix), making it hard for independent OSATs to capture directly.
Co-packaged optics (CPO) — integrating optical communication modules directly into chip packaging to boost data throughput — is more likely to deploy at scale in 2027–2028.
In plain terms = all three technologies carry real strategic weight, but their near-term earnings contribution may fall short of market expectations. Investors need to manage the timeline.
市场有风险，内容仅供研究参考，不构成投资建议。