CXMT surges 466% in Shanghai debut, but HBM tech gap looms

Shares of CXMT soared nearly 466% on their first trading day in Shanghai on July 27, 2026, and climbed more than 5% on July 31.

Investor enthusiasm stoked optimism about China’s semiconductor push, but analysts cautioned that closing the technology gap remains the real test.

In high-bandwidth memory (HBM), a critical AI technology, Samsung Electronics, SK Hynix and Micron hold commanding leads.

David Gibson, senior analyst at MST Financial, said the listing does not alter the industry outlook because demand still outstrips supply.

CXMT lacks access to extreme ultraviolet (EUV) lithography machines, restricted under U.S. export controls, forcing it to use about 30% more wafers than rivals to produce the same memory.

The company aims to produce HBM from the end of 2026 at a new Shanghai fabrication plant, said MS Hwang, research director at Counterpoint Research.

Initial products are expected to be HBM3E or HBM3, leaving the company one to two generations behind as competitors move to HBM4 and HBM4E.

Samsung said on July 30 it had scaled up HBM4 sales and shipped the industry’s first HBM4E samples to major customers.

Gibson warned that CXMT’s HBM yields will be low and volumes limited, keeping it behind rivals who demand high capacity and speed.

Ellie Wang, an analyst at TrendForce, said that even without EUV, a noticeable performance gap will likely persist.

Gibson also forecast that CXMT will gain global DRAM market share, with growth mainly from Chinese customers such as Tencent, ByteDance and Alibaba.

Market data from Counterpoint Research shows Samsung leads the DRAM market with a 38% share, followed by SK Hynix at 29%, Micron at 22%, and CXMT at 8%.

Wang added that CXMT supplies memory chips to many Chinese smartphone brands but remains focused on mainstream and mid-range segments.

Hwang described two scenarios: a bull case where CXMT catches up on technology, and a bear case where it fails to overcome equipment and HBM barriers.

Wang concluded it will be challenging to substantially narrow the gap before AI-driven memory demand normalizes, though CXMT continues developing new processes and products.