CXMT surges 500% on debut, tops Intel's market cap - Chinese memory reprices at cycle peak

In plain terms - CXMT (ChangXin Memory Technologies), China's leading DRAM maker, opened +471% (Nikkei) to +500% (Tech in Asia) on its Shanghai debut. At intraday highs its market cap briefly crossed Intel's. This is what a memory bull cycle plus a sovereign compute premium looks like priced in one candle.

Contexte

CXMT priced its Shanghai IPO at 8.66 yuan per share and raised 57.9 billion yuan ($8.6 billion) - one of the largest tech IPOs of the cycle (an earlier filing size of ~$9.8B was flagged in publication #1194; the final raise landed at $8.6B). It's China's answer to the SK Hynix / Samsung / Micron oligopoly on DRAM and - increasingly - HBM. Its listing coincides with a global memory supercycle driven by AI accelerator demand (HBM bundled with every GPU) and a Chinese policy pull to build a sovereign memory stack that isn't gated by US export controls.

Les données

• Debut move: +471% (Nikkei Asia) / +500% (Tech in Asia) - 27 July 2026
• IPO raise: 57.9 billion yuan ($8.6B), priced at 8.66 yuan/share (source TIA)
• Intraday market cap: briefly above Intel's

Analyse

Two overlapping stories collided at listing: (1) memory is the tightest single input in AI capex, and every dollar spent on GPUs pulls dollars into HBM; (2) Chinese buyers face restricted access to Micron/SK/Samsung high-end memory, so a domestic winner gets bid as strategic infrastructure. The Intel comparison is symbolic more than fundamental - Intel's revenue base dwarfs CXMT's - but the market's willingness to say "yes, that comparison is on the table" is the signal.

Scénarios

• Base: material post-lockup pullback as float expands - see the MiniMax-IPO precedent (fil minimax-ipo-crash).
• Optimistic: HBM3E/4 qualification with Chinese hyperscalers validates the strategic premium.
• Adverse: export-control tightening on the equipment side (ASML, LAM) caps CXMT's process roadmap; premium collapses.

Under the hood

CXMT is in volume on DDR5; HBM is the strategic prize. Cost per bit still trails Korean incumbents but the gap narrows every node. Watch equipment orders - that's the leading indicator.

So what

For allocators: the market is now willing to price sovereign-compute premiums in the memory layer, not just the logic layer. For operators: any bill-of-materials that assumes stable memory pricing needs a China-supply-scenario branch - either a floor from CXMT competition, or a ceiling from a Korean cartel with less exposure to Chinese demand.