#GoldmanSachsBullishOnCXMT

Goldman Sachs Turns Bullish on CXMT: A Structural Shift in the Global Memory Oligopoly

For years, the global DRAM market was treated as a closed three-player game. Samsung, SK Hynix, and Micron controlled over 94% of supply, and pricing power was theirs alone. That assumption is now being formally challenged on Wall Street.

Goldman Sachs has initiated a bullish stance on ChangXin Memory Technologies (CXMT), China's largest DRAM producer, in a report that is less about a single company and more about the fragmentation of the memory market itself. The call comes as CXMT quietly crosses a threshold that its Korean and American rivals can no longer ignore: scale at acceptable yields.

Why Goldman is Changing Its View Now

The report, led by Goldman's Asia tech hardware team, rests on three structural observations that have changed in the last 9 months.

1. Yield Breakthrough at 16nm and the Path to 12nm: CXMT has historically been stuck at 19nm and 17nm with low yields, producing primarily low-end DDR4 and LPDDR4X for domestic clients. Goldman notes that CXMT's 16nm process, using domestic and non-restricted DUV lithography, has now reached yields of 75-80% for DDR4, up from ∼55% a year ago. More critically, pilot production of DDR5 at a 12nm-class node is showing initial yields of 40-45%. That is not yet competitive with Samsung's 1b-nm (12nm) yields at 85%+, but it is commercially viable for China’s domestic server and smartphone ecosystem, which is under direct government mandate to de-risk from US suppliers.

2. Capex That Defies Sanctions: Despite US export controls on EUV and advanced HBM equipment, CXMT’s capex for 2025-2026 is estimated at $4.8 billion, funded by state-backed financing from Hefei and Beijing. Goldman estimates this will lift CXMT’s monthly wafer capacity from 170k wafers per month in late 2024 to 280k wafers per month by Q4 2026. At that level, CXMT would hold 11-12% of global DRAM bit supply, up from 4.5% in 2022. In LPDDR5 for Chinese smartphones, its share is already approaching 28%.

3. HBM is Not the Whole Story: The market is obsessed with HBM3 and HBM3E, where SK Hynix leads with 53% share and where CXMT is absent. Goldman argues this creates a blind spot. HBM accounts for only 18% of total DRAM bit demand but 42% of industry revenue due to premium pricing. The remaining 82% - standard DDR5, DDR4, LPDDR5 - is where CXMT is attacking. In a market where server DDR5 spot prices rose 19% YTD due to AI server builds, a low-cost, "good enough" second source is exactly what Chinese hyperscalers like Alibaba, Tencent, and Baidu want.

Implications Beyond CXMT

Goldman’s bullish call is not a buy recommendation on CXMT stock itself - the company is still private and aiming for a STAR Market IPO in late 2026 at a rumored $22 billion valuation. It is a de-rating signal for the incumbents.

The bank lowered its 12-month price targets for Micron by 6% and maintained a Neutral rating, citing the risk that CXMT will cap the upside in commodity DDR4/LPDDR4 pricing. Their model suggests every 1% of global share gained by CXMT depresses commodity DRAM ASP by 1.3-1.5%.

For the supply chain, it also means a bifurcation. Two memory markets are forming: a high-end, US-allied HBM market controlled by Korea and the US, and a mid-range, China-centric DDR5/LPDDR market where CXMT, backed by state subsidies and protected demand, becomes the price setter. Equipment makers like Naura and AMEC, not ASML or Lam Research, benefit in this second market.

Goldman's base case: CXMT will not kill Samsung or Hynix. But it will end the era of perfect oligopoly pricing. If CXMT hits its 300k WPM target by 2027 and successfully ramps DDR5, global DRAM will move from a 3-player to a 4-player market for the first time since Elpida’s bankruptcy in 2012.

That is why a single bullish note on a private Chinese company matters for every semiconductor investor.

This analysis is for informational purposes only and does not constitute investment advice.

#GoldmanSachsBullishOnCXMT