Chinese memory manufacturer ChangXin Memory (CXMT) approaches parity with Western DRAM/HBM producers, threatening near-term pricing leverage for AI accelerators.

Acer's chairman Jason Chen recently accused the world's largest memory chip suppliers—Samsung, SK Hynix, and Micron—of using public shortage forecasts as a covert pricing signal. Speaking to reporters this month, Chen flatly stated: "How could it stay short forever? Chinese capacity keeps coming onto the market. The supply shortage problem has already completely disappeared."

His timing was deliberate. On September 20, China's CXMT confirmed that a new memory chip platform had entered mass production, a development Barron's flagged as a larger threat to SK Hynix than to Micron. Chen believes the pricing bubble in dynamic random access memory (DRAM) and NAND flash is about to deflate.

Chen argues that when SK Hynix or Adata publicly forecast shortages extending to 2030, they are signaling coordination rather than making genuine predictions. Since antitrust rules prevent direct price coordination, these companies use public statements to communicate with one another. The stakes are enormous: PC prices are expected to rise 5 to 20 percent toward the end of 2026, plateau in the first half of 2027, and begin declining thereafter, with component costs peaking around mid-2027 as new capacity reaches scale.

The structural evidence supports Chen's thesis. CXMT and YMTC memory components are already being designed into products by Acer, HP, Asus, and Lenovo—a significant shift from previous cycles, when Chinese memory was regarded as second-tier and reserved for domestic consumption. Once the major PC manufacturers qualify a supplier, incumbent pricing power erodes quickly as buyers can negotiate competitively on each purchase order.

The macroeconomic backdrop strengthens Chen's argument. The Federal Reserve raised the funds rate's upper bound to 4.00 percent on September 17, and the core Personal Consumption Expenditures index reached 130.66 in July, its highest on record. Chen contends that hardware price increases in memory, solid-state drives, printed circuit boards, and fiberglass cloth directly feed inflation, forcing the Fed to maintain elevated rates. He characterized this dynamic as "not a normal phenomenon."

Chen expects further price increases in the first quarter of 2027, though smaller than current levels, before the reversal begins. The critical signals to watch are narrow: gross margins reported by SK Hynix and Micron on their next two DRAM earnings reports, and the bit-shipment share disclosed by CXMT in Chinese trade filings. If Chinese output continues climbing while incumbent margins remain near cycle highs, Chen's cartel-signal theory holds credibility—and regulatory scrutiny will likely follow. If margins compress first, market competition will have solved the problem independently. Either way, the 2030 shortage narrative does not survive the mid-2027 capacity wave that Acer is betting its 2027 procurement strategy upon.