CXMT's 600% Surge Is Changing DRAM Power
China's CXMT is forecast to triple its revenue growth with a 600% Q3 surge, claim 10% DRAM market share and flip to profit. Why Western semiconductor watchers are underestimating what's happening.
The Number Nobody in the West Is Talking About
CXMT, China’s state-backed memory chip maker, is forecasting a 600 percent revenue surge in the third quarter. The company expects third-quarter sales of 114.7 billion to 119.7 billion yuan—roughly 22.9 trillion to 23.9 trillion won—up from a single-digit figure a year ago. Net profit, which was in the red last year, is now expected to hit roughly 534 billion to 549 billion yuan. This is not a modest growth story. This is a structural break.
What the Western wire services are largely treating as a regional curiosity is actually a reordering of the global DRAM market. CXMT now sits at an estimated 10 percent global market share after jumping from 4 percent in the same quarter last year. That makes it the fourth-largest DRAM supplier on Earth, trailing Samsung (38 percent), SK Hynix (25 percent), and Micron (24 percent). Six percentage points of market share shifted in twelve months. That kind of velocity in memory is extraordinary.
Why CXMT Is Moving This Fast
CXMT itself pointed to two drivers in its earnings preview: surging AI-related demand and tight DRAM supply pushing prices up globally, plus the company’s own ramp in wafer output and equipment utilization. Both factors are real, but they also obscure what matters most—CXMT is executing on a model that Washington’s export controls have failed to stall.
The chipmaker went public in July, raising up to 66.6 billion yuan through its IPO. Its stock surged on listing day to become the largest market-cap company on mainland China’s exchanges. Management is now pushing for a 1 billion to 2 billion yuan share buyback to support the stock, targeting roughly 1 percent of the currently free-floated share pool. With only about 6 percent of total shares in circulation, the impact of any buyback on trading dynamics is amplified.
More importantly, CXMT is reinvesting those gains. The company announced it will launch a server-grade DDR5 DRAM using a next-generation 4F² cell structure—designed to shrink cell area and raise integration density—before year’s end. Samsung and SK Hynix dominate HBM and high-bandwidth memory markets, but general-purpose server DRAM is a different, larger slice of the market, and CXMT is moving aggressively into it.
The 5-to-7-Year Gap Is the Real Story
Analysts at KB Securities cautioned against over-interpreting the gap closing. CXMT has not disclosed specific yield rates or full product specifications for the new 4F² device, and the product itself is not high-bandwidth memory—it is conventional DDR5 server DRAM. In raw process technology, KB Securities estimates Samsung remains roughly five to seven years ahead of CXMT.
But the gap may not close linearly. Every profitable quarter CXMT books is capital available for its own process advancement. Every percentage point of market share captured in the mid-range DDR market is revenue that does not depend on Samsung or SK Hynix licensing or supplying anything. The export controls designed to keep Chinese memory makers decades behind are producing the opposite effect in a narrower band: forcing them to build independence faster than they otherwise would have.
Who Wins, Who Loses
CXMT wins first. It converts AI-driven pricing power into hard profit, funds share buybacks, and accelerates process development without relying on foreign suppliers. Its IPO proceeds and domestic capital markets give it breathing room that earlier Chinese memory efforts never had.
Samsung and SK Hynix face a squeeze. For years they have competed primarily with Micron for mid- to high-end market share. Now a fifth player—a state-supported Chinese firm with aggressive pricing incentives and rapidly scaling capacity—is entering the same segments. Korean companies are responding. Both Samsung and SK Hynix have been ramping capacity and revising pricing strategies harder than usual, but CXMT’s cost structure is fundamentally different, subsidized by Chinese industrial policy rather than purely market forces.
Micron is in a weaker position than it realizes. With only 24 percent share, it sits between the Korean duopoly and the rising Chinese challenger. If CXMT continues gaining share in DDR5 server DRAM, Micron has the most to lose among the non-Korean players.
US export controls look less effective than advertised. The policy was designed to slow Chinese memory advancement. In practice, it has concentrated China’s industrial effort, pushed CXMT toward self-reliance, and given it a captive domestic market that insulates it from some of the demand cyclicality that hits Samsung and SK Hynix harder.
What Comes Next
Three things will determine whether CXMT’s surge is structural or cyclical. First, whether the company can translate its profit expansion into meaningful yield improvements on advanced nodes. Second, whether AI-driven DRAM demand holds or softens—if prices retreat, CXMT’s revenue growth will compress faster than Samsung or SK Hynix’s, simply because its cost base is less mature. Third, whether US policymakers decide to expand restrictions to cover the specific DDR5 manufacturing equipment CXMT is likely depending on.
The quarterly numbers are already eye-catching. But the market-share trajectory is what will matter in six to twelve months. If CXMT reaches 12 or 13 percent global DRAM share by end of next year, the conversation shifts from surprise to strategy. Korean chipmakers, US trade officials, and investors will need to decide whether this is a temporary AI blip or the beginning of a sustained erosion of the Pacific DRAM order.
The numbers from this week suggest it is not a blip.