CXMT's DRAM Share Surge: Which Memory Incumbent Pays the Commodity Margin Bill?

The prevailing story is that ChangXin Memory Technologies, flush with IPO proceeds and domestic state backing, will flood the commodity DRAM market and crush incumbent margins — and that the incumbent least protected by HBM will pay the bill first. The logic is sound as far as it goes. It is also, in the current market, wrong about the direction of causation.

CXMT's share gains are real and fast. It took roughly 9.5% of global DRAM revenue in Q2 2026, nearly double the ~4% it held a year earlier, and its wafer capacity is tracking toward roughly 350k wafer-starts per month by end-2026 — within shouting distance of Micron's projected ~375-385k, which would make it the second- or third-largest supplier by capacity and about 17% of global DRAM capacity. The problem for the margin-compression thesis is that share is not the same as pricing power. CXMT's cost per DDR5 bit is more than 30% above the three incumbents, its ASP sits only 5-10% below them, and most of its output feeds anchor-locked domestic Chinese demand. A supplier with a cost disadvantage and no price discount does not crash a market; it fills the pieces the incumbents do not want.

That is why the feared pricing break has not arrived. The conventional reading predicts collapse in commodity DRAM contract prices; TrendForce instead forecast conventional DRAM contract prices to rise 13% to 18% quarter-on-quarter in Q3 2026. Prices are still climbing, not falling. This is the falsification condition, and for now it is met: contract prices have kept rising through CXMT's entire share ramp because the global market remains structurally undersupplied — HBM consumes roughly three times the wafer capacity per bit as DDR5, and incumbents' new commodity capacity does not land until 2028.

The incumbent you expect to pay first is the one printing the biggest margins.

Which incumbent is most exposed to commodity DRAM relative to HBM? Public quarterly share work puts each supplier's HBM share of its own DRAM revenue as follows in Q2 2026:

Supplier	DRAM revenue share	HBM mix of DRAM	Commodity exposure
SK Hynix	24.9%	~50%	lowest
Samsung	39.4%	~33%	moderate
Micron	23.3%	~18%	highest

Micron, not Samsung, is the leanest-HBM incumbent: HBM is only about 18% of its DRAM revenue, so roughly 82% of its DRAM book is conventional commodity parts. On the raw thesis, that should make it the first victim of CXMT. Instead, Micron is the incumbent that has benefited most. Its short-cycle commodity and server-DDR5 contracts repriced fastest into the shortage, driving the fastest Q2 DRAM revenue growth (65.5% quarter-on-quarter) and a record gross margin of 84.6% in its fiscal third quarter, with a fiscal fourth-quarter guide near 86%.

The inversion is most visible at the opposite extreme. SK Hynix, the HBM-heaviest of the three at roughly 50% HBM mix and the safest from commodity pressure on any forward view, was the relative revenue laggard in Q2 at 37.9% growth. Its HBM is sold through long-term contracts that capped its ASP capture while commodity prices surged. The company still posted a record 76% operating margin — but its bottleneck is not commodity competition; it is that it already contracted the upside away.

The ranking flips forward, and it is conditional.

The reason the current picture is so counterintuitive is that the cycle has not yet reached the point where CXMT's additions collide with incumbent commodity pricing. That collision is a 2027-2028 event: CXMT's modeled capacity rises to ~500 kwspm by end-2028 at roughly 17% global share, and its path to breaking commodity prices depends on resolving a domestic DUV lithography constraint that may not produce meaningful volume until the early 2030s. When that capacity does land, the ranking reverses to the intuitive one, by relative commodity exposure:

1. Micron — roughly 82% of DRAM in commodity, the largest single bet on DDR5 prices holding. Its take-or-pay HBM backlog (sold out through 2026 and into 2027 under multi-year agreements) hedges only the 18% HBM slice; the commodity 82% is unhedged. Most margin and multiple compression if pricing rolls.
2. Samsung — moderate commodity exposure at ~67%, but the largest absolute commodity book in the industry. Its HBM4 ramp and multiyear customer agreements soften, but do not cancel, the exposure.
3. SK Hynix — least exposed at ~50% commodity, and its HBM volumes are already sold out through the cycle under take-or-pay with a premium on HBM4. Most insulated.

Here is the tension the numbers leave unresolved. The assignment's second falsification test — that the leanest-HBM incumbent's take-or-pay HBM backlog offsets its commodity pressure — is only partly met. Micron's HBM is bound by binding contracts through 2026 and into 2027, a genuine floor. But that floor sits under 18% of its DRAM revenue, not the 82% that is exposed. A take-or-pay hedge over a fifth of the book does not neutralize a commodity downturn across the other four-fifths.

The margin bill from CXMT is real, but it has not been presented. Today the ranking runs inverted from intuition: commodity exposure is a tailwind, and the leanest-HBM incumbent carries the largest commodity bet at record margins. The investor question is not whether CXMT gains share — it already has. It is whether CXMT can convert that share into enough supply to break commodity pricing before the incumbents' own 2028 capacity arrives, and which book is unhedged when it does. By relative exposure, that book is Micron's.