Skip to content
Chips & PerformanceSep 5, 20265 min read

Xiaomi 18 Fold Gives CXMT's LPDDR6 a 300,000-Phone Trial

CXMT says it beat Samsung and SK hynix to LPDDR6 mass production. Xiaomi's limited-volume foldable will reveal whether that announcement can become durable supply.

ByMobileTech Desk

Xiaomi launches the 18 Fold on 7 September, and the component most worth watching is not the one on the spec sheet's first line. It is the DRAM. The foldable will be the first shipping phone with LPDDR6 memory, and the company supplying it is ChangXin Memory Technologies — a Chinese firm that, one memory generation ago, was several years behind Samsung, SK hynix and Micron.

CXMT announced on 29 August that its LPDDR6 had "officially entered mass production" and would debut in the 18 Fold. Xiaomi CEO Lei Jun separately confirmed that the company's in-house Xring O3 processor supports CXMT's LPDDR6 and that the foldable would be the first handset to use it. TechNode noted the timing: CXMT began mass-producing the previous LPDDR5 generation less than a year ago.

Why LPDDR6 is a harder jump than the version number suggests

Mobile DRAM upgrades are usually incremental — higher clocks, better power curves, same basic plumbing. LPDDR6 is not that. As Tom's Hardware laid out, it changes three things at once.

It moves from NRZ signalling to PAM3, encoding three bits of information across two symbols where LPDDR5 and LPDDR5X encoded one bit per symbol. It pushes transfer rates to 43.2 GT/s. And it narrows the I/O width from 32 bits to 24, which is the awkward one: a change to channel geometry forces rework on both the hardware and the software side, so the memory controller in the SoC has to be designed for it from the start. That is why the LPDDR6 debut is arriving attached to a specific processor rather than as a drop-in part.

For a foldable in particular, the power-efficiency and bandwidth headroom matter more than the raw number implies. A device driving a large inner panel plus a cover display, with on-device AI workloads pulling at memory bandwidth, spends a meaningful share of its battery budget moving data around. That is the case Xiaomi is making by putting the newest memory standard in its most thermally and spatially constrained flagship.

First to announce is not the same as first to lead

Samsung and SK hynix have both developed LPDDR6. Neither has publicly tied mass production to a named handset. SK hynix said in March it had completed development of a 16-gigabit LPDDR6 on its 1c DRAM process and would begin supplying customers in the second half of the year; Korean media reported in July that Xiaomi was likely to be among the first customers, and SK hynix declined to confirm. Samsung unveiled its LPDDR6 before CES 2026 and said it had been working with major SoC makers on early validation, without disclosing a mass-production timetable or a customer.

So CXMT's claim to a global first rests on announcement order and a commercial deployment, not on demonstrated capacity. Both Tom's Hardware and the Korea Herald flagged the same limitation from different angles: the announcement covers one phone maker and one model, and the part still has to clear validation with other handset makers before it counts as a supply-chain fact.

The volume question sharpens it further. Reuters has reported Xiaomi is targeting roughly 200,000 to 300,000 units of the 18 Fold — small for a flagship. A South Korean semiconductor industry official, speaking to the Korea Herald, was blunt about what that implies: "If CXMT is producing LPDDR6 mainly for that level of demand, its actual production capacity and yields could still be lower than the mass-production announcement might suggest."

The same official described the arrangement in terms that cut both ways: "Xiaomi is effectively serving as an early test bed for CXMT by putting its own processor together with LPDDR6 in a flagship device." A limited-volume, in-house-silicon foldable is exactly the kind of product you use to prove a new memory part in public without betting a hundred million units on it. Xiaomi has not said whether CXMT is the sole LPDDR6 supplier for the phone.

The money behind it

CXMT is not operating on thin margins while it catches up. The company raised 57.9 billion yuan — about $8.6 billion — in a Shanghai IPO roughly a month before the LPDDR6 announcement, earmarking part of it for capacity expansion and process upgrades. It reported first-half revenue of 150.3 billion yuan, up 874% year on year, with net profit of 77.6 billion yuan.

Those are numbers produced by a global memory market where AI datacentre demand has been consuming supply and pushing prices up across every DRAM category, mobile included. A Chinese producer entering a new standard at commercialisation rather than years afterwards, with IPO money and margins like that behind it, is a different competitive proposition than CXMT was two years ago.

One foldable does not make a fourth supplier

Not much, immediately. LPDDR6 in one limited-run foldable does not change what phones cost or how fast they feel this year. What it changes is the number of credible suppliers for the memory in every phone after it.

Mobile DRAM pricing has been one of the pressures pushing up handset prices through 2026, and pricing pressure is a function of how many companies can actually make the part at volume. If CXMT converts this announcement into qualified, multi-customer supply, that count goes from three to four at the leading edge of the standard. If it does not — if the yields behind 300,000 phones do not scale — then 7 September was a very good day for Xiaomi's marketing and not much else.

More from the desk