Chinese memory chip makers expand capacity amid AI boom and global shortage
The technology race has never been as intense as in the era of artificial intelligence. Memory chips, once just a component of computers, have become a strategic resource comparable to oil or rare earth metals. Artificial intelligence consumes data on an unprecedented scale, and every new model demands more memory and speed. This demand is reshaping the global market, forcing manufacturers to retool factories, investors to reassess valuations, and analysts to guess when this cycle will end and what will follow.
Chinese memory chip makers are expanding their production capacity as the AI boom triggers supply shortages on the global market, fuelling consumer demand. Industry experts say the memory chip sector is poised for a period of high prosperity, with the global market experiencing an unprecedented surge. However, prices have risen not so much due to a sharp increase in end‑user demand but because of a "reallocation effect" of capacity towards more profitable AI chips, explained Xin Yi, research manager at IDC China. "Production lines are being retooled for more profitable AI chips, and only a few manufacturers still produce traditional memory chips. Cloud service providers are stockpiling out of fear of supply disruptions, leading to panic buying and further exacerbating the shortage," he added.
Many Chinese companies are accelerating capacity expansion. Biwin Storage Technology from Shenzhen has been ramping up production of enterprise‑grade memory modules since late last year. CEO He Han said: "Our production lines are currently running at nearly full capacity. We are expanding production in high‑demand areas, especially in the enterprise storage segment for data centres, including enterprise‑grade SSDs and memory modules." The shortage has also driven up stock prices of leading manufacturers — Samsung Electronics, SK Hynix and Micron — whose market capitalisation has surpassed $1 trillion.
Experts note that the current price surge is not just a cyclical upturn but a long‑term structural shift driven by AI, and the upward cycle could continue until 2028. Xin Yi of IDC stressed: "According to our baseline estimate, the memory chip supply shortage will not ease between 2027 and 2028." However, a recent JPMorgan report cautioned that the rapidly growing share of AI memory in cloud service providers' capital expenditure — projected to reach 52% this year and potentially exceed 70% next year — is not sustainable in the long term. "In the short term, high valuations are supported by the supply‑demand gap in HBM, so downside risk is limited. But after 2027, if AI commercialisation does not keep pace with capacity expansion, the storage sector will face a real test," Xin warned.
High‑Bandwidth Memory (HBM) has become a critical component for AI accelerators, and its production requires advanced technologies available to only a few players. Chinese manufacturers such as YMTC and CXMT are actively expanding capacity to reduce import dependence and meet growing domestic demand. As CCTV+ reports. Analysts agree that the current shortage cycle will last at least until the end of 2027, but long‑term prospects depend on the pace of AI application commercialisation and the emergence of new memory technologies.
