ReportGem ReportGem 中文

GLOBAL RESEARCH ARCHIVE

China Semiconductors (H/A): Takeaways from China Conf & AI/Tech Tour: continued strength in AI-led demand

Published: 2026-05-19Institution: BofA Global ResearchPages: 11Original language: 英语Evidence page: 2

Research evidence excerpt

China Semiconductors (H/A): Takeaways from China Conf & AI/Tech Tour: continued strength in AI-led demand

Earnings revision table

Exhibit 2: We revise estimates for Montage and GigaDevice

China semiconductors: estimates changes

FY26E FY27E FY28E

Company Rmb mn New Old Diff (%) New Old Diff (%) New Old Diff (%) Reason for estimates revision

Revenue 7,609 7,570 0.5 11,443 10,481 9.2 14,982 13,567 10.4 Potential faster MRCD/MDB ramp-up

Montage

Net profit 3,961 3,945 0.4 5,846 5,335 9.6 7,676 6,955 10.4

Revenue 21,502 20,403 5.4 27,193 25,177 8.0 32,079 29,220 9.8 Better DRAM price outlook and faster ramp-up by cooperating with

GigaDevice

Net profit 8,254 7,136 15.7 10,387 8,964 15.9 11,967 10,131 18.1 Changxin Memory Technology

Source: BofA Global Research estimates

BofA GLOBAL RESEARCH

InnoScience Technology (2577 HK) – Buy

Key takeaways from China Conference meeting

We hosted InnoScience management during our China Conference in Shenzhen.

Takeaways include: (1) AI infrastructure shifting toward energy efficiency drives step-up

in GaN content, with megawatt-level racks implying meaningful value uplift for GaN

power devices; (2) management sees clear margin improvement outlook, reiterating

FY26E targeted GPM to 15%+ with longer-term potential of c.45-50%; (3)

manufacturing capability (particularly yield and scale) remains a key barrier for the

industry, reinforcing competitive advantages for leading companies, such as

InnoScience. We maintain our Buy rating.

GaN content value uplift with roll-out of next-gen AI infra

Management highlighted that competition in AI infrastructure is increasingly shifting

from computing performance to energy efficiency, driving demand for higher power

density solutions. As data center architectures evolve from kilowatt- to megawatt-level

The English excerpt is extracted automatically from the cited source page and may contain layout or recognition errors. It is never batch translated.

Open report viewer