CXMT’s stock jumped as much as 470 percent on its Shanghai debut last month, the biggest first-day pop attached to the largest mainland Chinese semiconductor listing on record, and the company is already spending that windfall on something concrete. Reuters reported Monday that ChangXin Memory Technologies is in financing talks for a second 12-inch wafer plant in Beijing’s Yizhuang district, about 12 miles southeast of the city center, right next to the DRAM fab it already runs there. The company is reportedly asking for at least 60 million yuan in local government support to get it moving, on top of new plants already under construction in Shanghai and Hefei that could more than double CXMT’s total capacity to over 600,000 wafers a month once everything is running.

That expansion push is happening at the exact moment CXMT is closing in on something more interesting than raw capacity: real parity with Samsung and SK Hynix on next-generation mobile memory. China Business News reported this week that CXMT has nearly finished R&D verification for LPDDR6 at 12.8 gigabits per second, one of the last technical gates before trial production, and the company could start mass production before the end of the year. I’ve been tracking the memory price crunch on this blog since the CMF Phone 3 Pro got killed by RAM inflation and Samsung tacked another hundred dollars onto its foldables, and every one of those stories assumed the supply side was a fixed problem: Samsung, SK Hynix, and Micron control something like 90 percent of global DRAM, so whatever they charge, phone makers eat it. A second serious LPDDR6 supplier changes that math even before it ships a single production unit, because it gives every OEM in the room a name to threaten the incumbents with at the negotiating table.

The context makes the timing almost too neat. LPDDR5X prices jumped 78 to 83 percent in the second quarter alone as AI datacenters soaked up production capacity that used to go to phones, according to TrendForce, and Samsung is reportedly reconsidering LPDDR6 and UFS 5.0 storage for the Galaxy S27 Ultra specifically because of cost and mass-production risk. LPDDR6 is expected to run about 20 percent more expensive than LPDDR5X at launch, which was already going to restrict it to the most expensive flagships for a year or two. A second supplier with competitive pricing doesn’t fix the underlying shortage, but it gives Xiaomi, which is reportedly lining up an early LPDDR6 flagship, and everyone else building phones in that price bracket a reason to believe the premium won’t stay locked at whatever Korea decides to charge.

What actually surprised me digging into this is that Apple is already testing CXMT chips for the devices it sells inside China, despite the political noise out of Washington about exactly that kind of qualification. That’s not a company hedging on principle. That’s the world’s most margin-obsessed hardware maker deciding a Chinese memory supplier is now good enough to test against its own bill of materials, and doing it quietly enough that it only shows up as a footnote in coverage of CXMT’s other announcements. It fits the same pattern I’ve been watching in China’s broader chip export numbers: the domestic industry stopped being a subsidized also-ran a while ago, and nobody outside the trade press seems to have fully priced that in. CXMT already holds something like 30 percent of China’s domestic LPDDR market and roughly 8 percent of global DRAM on commodity DDR5 alone. Adding a competitive LPDDR6 line moves it from the cheap domestic option to a company that can plausibly sit at the same table as SK Hynix on the memory China’s own phone industry actually wants to use.

I’ll say the obvious caveat, because it matters: CXMT hasn’t proven yield, reliability, or manufacturing scale at LPDDR6 speeds yet, and “nearly completed R&D verification” is not the same sentence as “shipping in volume.” Plenty of Chinese semiconductor milestones announced with this much fanfare have taken years longer than promised to turn into product on shelves, and a 60 million yuan ask for a second Beijing plant is a rounding error against what a genuinely competitive fab actually costs to build out. But the DRAM oligopoly hasn’t faced a credible new entrant at the leading edge in longer than I can remember, and CXMT is now doing three things at once that none of its predecessors managed: it has the IPO capital, it has Apple testing its parts, and it’s racing toward the same next-generation spec Samsung is hesitating on for cost reasons. That combination is worth watching even if the Beijing plant ends up being the least important part of the story.

Sources