How CXMT Became the Most Profitable Memory Maker in Q2 2026

CXMT’s extraordinary profitability in the second quarter of 2026 looks chiefly like the result of a favorable memory-market cycle: a shortage of standard DRAM, rising DDR5 prices, rapid sales growth and a product mix that benefited while major Korean suppliers prioritized high-bandwidth memory (HBM). The 82% EBIT margin is a meaningful milestone, but it should not be confused with leadership in the more technically demanding HBM market. 311

The margin lead came from a rare pricing-and-cost gap

CXMT recorded an 82% EBIT margin for April through June, compared with 76% for SK Hynix and 70% for Samsung Electronics’ semiconductor division, according to reporting based on QUICK FactSet data. 311

The key is operating leverage. CXMT’s first-half revenue rose 873.64% year on year to 150.31 billion yuan, while attributable net profit reached 77.6 billion yuan, reversing a 2.3 billion yuan loss in the comparable prior-year period. 1 One analysis of the interim figures reported that cost of sales increased only 70.76%, while first-half gross margin rose from 13.00% to 84.74%. 8

That combination—selling prices and shipment volumes rising far faster than costs—can produce exceptional margins in memory, a business where manufacturing costs are comparatively fixed over short periods. CXMT’s reported Q2 gross margin was 87.59%, reinforcing how unusually favorable the quarter was. 12

Why DDR5 became so profitable

The AI infrastructure boom has pushed leading memory manufacturers to direct attention and capacity toward HBM, the stacked, high-bandwidth memory used alongside AI accelerators. Reporting on CXMT’s margin surge attributed the tight supply of commodity memory in part to South Korean producers’ HBM focus, which lifted DDR5 pricing.

That created an opening for CXMT’s mainstream DRAM portfolio. A higher margin on DDR5 in one quarter does not mean DDR5 has become strategically more important than HBM, or that CXMT has surpassed Samsung and SK Hynix in leading-edge memory design, packaging or customer qualification. Rather, it shows that scarce standard DRAM can be exceptionally profitable when demand is strong and supply is constrained.

The distinction matters because HBM remains central to the AI accelerator supply chain. CXMT’s Q2 result demonstrates commercial momentum in DRAM; it does not, by itself, establish equivalent HBM capability or ecosystem position.

CXMT’s financial turnaround was also a scale story

The first-half figures show more than a margin improvement. CXMT reported 150.31 billion yuan in revenue, nearly 10 times its year-earlier level, and 77.6 billion yuan in attributable profit. 113 Revenue for the first half alone exceeded the company’s revenue for all of 2025, according to reporting on its earnings. 15

CXMT’s global DRAM position also expanded. Counterpoint Research data cited in reporting put its Q2 share of global DRAM revenue at 10%, the first time it reached double digits. The company was described as the world’s fourth-largest DRAM maker. 1718

A 10% revenue share is not market leadership—Samsung, SK Hynix and Micron remain the three largest suppliers—but it changes CXMT’s relevance to global DRAM supply and to customers seeking a Chinese source of memory.

Domestic customers support demand, especially Tencent

Reuters reported that CXMT signed a long-term memory supply agreement with Tencent worth more than 20 billion yuan, citing three people with knowledge of the deal. 17 CXMT’s IPO prospectus also identified Tencent, Alibaba Cloud, ByteDance, Lenovo and Xiaomi among its major customers. 17

That customer list is important because Chinese cloud and device companies can provide a large domestic demand base. However, public reporting in the supplied evidence substantiates the Tencent agreement most specifically; it does not establish the value or volume of purchases by Alibaba Cloud or ByteDance. Those companies should therefore be viewed as important named customers rather than as confirmed drivers of a quantified share of CXMT’s revenue.

Capacity expansion could reshape commodity DRAM

Reported estimates put CXMT’s monthly wafer capacity at around 300,000 in 2026, with capacity projected to reach roughly 500,000 to 600,000 wafers per month by 2028. 19

If achieved, that scale would give CXMT much greater influence in conventional DRAM markets. But wafer starts are not the same as durable technology leadership. The value of new capacity depends on process capability, yields, product mix, advanced packaging, supply-chain access and customer qualification. Rapid expansion can also add supply to a cyclical market, eventually pressuring the prices that made the current quarter so profitable.

Korea’s AI-memory boom is not contradicted by CXMT’s rise

CXMT’s success does not indicate a collapse in South Korean memory competitiveness. South Korea’s exports rose 53.2% year on year in May 2026 to a record $87.75 billion, with semiconductor exports up 169% as AI investment lifted chip demand, Reuters reported.

The two developments can coexist. Samsung and SK Hynix can benefit from high-end HBM and server-memory demand while CXMT captures a profitable opening in DDR5 and other standard DRAM. CXMT’s margin lead reflects where supply was tightest during the quarter; it is not a complete measure of long-term technology leadership.

What would make the shift structural?

The strategic issue is whether repeated profitable cycles give CXMT enough cash flow to fund additional fabs, process development, packaging and eventual high-end-memory programs. Success in those areas could narrow the gap with established suppliers, especially in conventional DRAM.

For now, the evidence supports a more limited conclusion: CXMT has become a significant fourth force in DRAM and captured an exceptional cyclical profit opportunity. Its 82% Q2 EBIT margin is real and consequential, but it is not yet evidence of a permanent profitability regime—or of parity with Samsung and SK Hynix in HBM.

The next test is what happens when standard-DRAM supply recovers. If CXMT can sustain quality, customer adoption and returns through a less favorable pricing environment, its rise will look more structural. If prices normalize sharply, Q2 2026 may be remembered as an unusually lucrative point in the memory cycle rather than a lasting reordering of the industry.