CONNECT WITH US
Jul 27
CXMT tops A-share market cap on IPO debut
Chinese DRAM maker ChangXin Memory Technologies (CXMT) listed on the STAR Market of the Shanghai Stock Exchange on July 27, with its market capitalization briefly topping CNY3.6 trillion (approx. US$533.1 billion) on the first day of trading, surpassing Industrial and Commercial Bank of China and becoming the largest listed company in the A-share market.
Zeiss Semiconductor Manufacturing Technology is expanding production capacity in Germany to relieve a critical constraint on ASML's lithography output, with AI data center investment accelerating demand for advanced logic and memory chips.
Apple and Micron are fighting over whether the smartphone maker can buy memory chips from Chinese manufacturer ChangXin Memory Technologies (CXMT) for non-US devices, a move opposed by Micron. Both companies are lobbying the administration of US President Donald Trump, which could force him to make an uncomfortable choice between his goals of cheaper prices or domestic chip production.
Nvidia is pushing into the server CPU market with its Vera chip as AI infrastructure competition enters a new phase, seeking to evolve from a chip supplier into a full AI computing systems provider. But the company still needs cloud service providers (CSPs) to embrace the platform at scale if it wants to challenge Intel and AMD in servers.
The push toward deeper sovereign control over the critical hardware stack underpinning AI models is gaining traction following news of Anthropic's outreach to SK hynix for supplies related to producing its own chips, spanning early design planning and key component procurement.

South Korean power semiconductor maker EYEQ Lab has completed mass-production validation of its dedicated 8-inch, or 200mm, silicon carbide (SiC) line in Busan, achieving an average yield above 90%, ET News reported, citing industry sources.

Memory price increases are expected to moderate in the second half of 2026, but shortages will remain severe, particularly for DRAM, according to Apacer CEO C.K. Chang.
Yunnan Germanium has secured a major supply deal for indium phosphide (InP) wafers, underscoring how AI data center buildouts are reshaping global demand for high-speed optical parts. The contract could bolster the company's revenue base, but execution risks, capacity expansion, and market swings will determine how much benefit reaches investors and customers worldwide.
China's AI computing infrastructure is developing along two sharply divergent paths. As large language models and AI agents gain traction across industries, rapidly rising token consumption is driving shortages of advanced computing capacity. At the same time, portions of the country's installed lower- and mid-tier capacity remain underutilized because of software compatibility issues, weak coordination between supply and demand, and inadequate resource scheduling.
South Korea and AMD have agreed to build and demonstrate heterogeneous AI computing infrastructure integrating AMD CPUs and GPUs with homegrown neural processing units, as Seoul seeks to give its AI chip industry practical reference models and a path into the global AI computing supply chain.

Samsung Electronics plans to manufacture the base die for its next-generation HBM5 memory using a 2nm gate-all-around foundry process, as faster high-bandwidth memory raises pressure on logic performance, power efficiency and heat management.

South Korean AI chip designer Rebellions plans to begin shipping its next-generation Rebel100 accelerator in the second half of 2026, betting that wider use of AI agents and commercial AI services will shift more infrastructure spending from model training toward inference.