REAL-TIME GLOBAL RESEARCH
Jianghai Capacitor (002484 CH): Downgrade to Hold: Concerns amid heated sentiment
Research evidence excerpt
Jianghai Capacitor (002484 CH): Downgrade to Hold: Concerns amid heated sentiment
e expect lower demand for horn-type capacitors in Nvidia’s
800V DC solution” for details); the 800VDC solution is likely to use more film capacitors
which Jianghai also makes but enjoys much smaller market share now.
◆ Meanwhile, we think further growth upside for Jianghai depends on the application of LIC
supercapacitors (Lithium-Ion Capacitor) in AIDC. The market could be as large as USD1.6bn
in 2028e (see exhibit 3) and there are fewer competitors of LIC or HUC than of EDLC
supercapacitors globally, but the final design of next-generation AI servers is yet complete and
the timing of massive delivery is uncertain. We are not sure if it will be used in large scale yet.
◆ MLPC (multilayer polymer aluminium capacitors) is also a promising growth driver for
Jianghai and it did not draw much attention from investors, mainly because the addressable
market is relatively smaller (exhibit 3). We also find that many Japanese peers can make
similar products, in addition to Panasonic.
Overall, we think it is fortunate that Jianghai has exposure to almost all the capacitors to be
used AI servers (except MLCC and tantalum electrolytic capacitors, see exhibits 3-4). This is
thanks to Jianghai’s comprehensive product portfolio and leading market position. AI
infrastructure will be the major growth driver for Jianghai in 2026-28e (exhibit 8).
In 2026-28e, the major growth contributor will be horn-type aluminium capacitors used in AI
servers, and we highlight that our market sizing in exhibit 3 includes those for ASIC servers and
is not limited to those in Nvidia’s Vera Rubin racks (exhibit 1). Our Vera Rubin delivery forecast
is from HSBC global research report NVIDIA Corp (NVDA US), 19 May 2026.
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