Hardware/ dram · cxmt · memory-shortage · china

CXMT Could Nearly Double DRAM Production by 2028

Analysts expect CXMT to expand PC-grade memory production sharply, but the chips will likely stay in China while Dell's CEO warns of shortages to 2030.

Chinese chipmaker CXMT is forecast to nearly double its DRAM production by 2028, but the extra memory is unlikely to ease prices for buyers outside China.

The forecasts come from Counterpoint Research and TrendForce, as flagged by P Equity Research. They put CXMT's DRAM wafer production at around 320,000 units this year and 420,000 in 2027, growth of just over 30%. For 2028, the analysts see 500,000 to 600,000 units. At the top of that range, production would nearly double from this year's level. The source does not say what the unit is, such as wafers per month or per year, so the figures are best read as relative growth rather than absolute volume.

The detail that drew attention is the type of memory. DRAM is the working memory in PCs and servers. Much of the industry's recent focus has been on HBM (high-bandwidth memory, the stacked chips that feed AI accelerators). CXMT's expansion is in conventional DRAM, the kind that goes into a PC rather than a data center server.

The catch is where it ends up. China's own demand is substantial, and the government is likely to favor keeping local buyers supplied over exports. Some Western manufacturers already use CXMT memory, but only to a limited degree, and reportedly at roughly the same price as Samsung's. Without a discount, buyers have little reason to switch.

Meanwhile, Michael Dell told the Wall Street Journal that memory supply problems could last up to four years, until 2030, and that the shortage will worsen in 2027.

The timing gap is worth noting. CXMT's forecast expansion would be complete by 2028, two years before the end of Dell's horizon. If that capacity mostly serves China, the pressure on everyone else would remain, and the open question is whether any of it reaches export markets.

TR

The Revision

Written by an AI system from the public sources credited above. How we write →