On 26 August, while preparing a ₩6.8tn (~$4.9bn) listing on the Shanghai Stock Exchange, China's Yangtze Memory Technologies said it intends to be the world's largest NAND flash maker by the end of 2027. Korean outlets carried it as a declaration of war on Samsung Electronics Co., Ltd. (KRX:005930) and SK hynix Inc. (KRX:000660).
The date is the weakest part of the claim and the pricing implication is the strongest, and most of the coverage got those the wrong way round.
Start with where the market actually sits. In the first quarter of 2026, on Counterpoint's tracking, Samsung had 29% of NAND, SK hynix 18%, Kioxia 14%, and Micron, SanDisk and YMTC about 13% each. YMTC was at 8% in the same quarter of 2025. So it added roughly five points of share in a year, which is genuinely fast for this industry, and it did it in a market where everyone was capacity-constrained and pricing was going up.
A note on the ranking, because the sources don't agree. Several Korean reports describe YMTC as third in the world. Counterpoint has it in a three-way tie at 13%, behind Kioxia at 14%. Whether it is third or fifth depends on which tracker you use and how you treat SanDisk, and nobody should build an argument on the difference.
Run YMTC's own recent pace forward. Five points a year, from 13% at the end of the first quarter of 2026, gets it to roughly 22% by the end of 2027. That is seven quarters of compounding at a rate it has hit exactly once.
Twenty-two percent is not first place. It's first place only if Samsung is below 22% by then, which means Samsung shedding about seven points of share in under two years.
That is the part worth sitting with. Overtaking doesn't require YMTC to close a sixteen-point gap on its own. It requires both sides to move, and the Samsung half of the move is as large as anything YMTC has ever done. Korean memory makers have given up seven points of NAND share before — the business has had genuinely bad stretches — but doing it during a shortage, with prices rising, would be a peculiar way to lose.
The plainest reading is that the 2027 date is an IPO document talking. A company raising ₩6.8tn from Chinese public markets has an obvious reason to state an aggressive terminal goal, and stating it in August, two months before a listing, is not a coincidence. That doesn't make the underlying capacity build fictional. It makes the date marketing.
For scale: YMTC's entire raise is about 22% of what Samsung spent on property and equipment in six months (₩31.2tn), and roughly a tenth of the cash Samsung parked in financial instruments over the same half. The IPO is not where YMTC's money comes from. It is the visible sliver of a state programme, and that is precisely why the raise size tells you very little about the capacity that's coming.
Rank is a vanity measure. What a producer with non-commercial funding does to a commodity market is set the floor price, and it can do that from 13% share as effectively as from 30%.
NAND is more exposed to this than DRAM, and much more exposed than HBM. HBM is qualified into a specific accelerator, the customer list is short, and switching costs are real. NAND at the volume end is close to a pure commodity: a controller, a stack of dies, a capacity number on a box. Samsung showed new P9 and P7 portable drives in late August with roughly doubled transfer speeds, which is a fine product and also a reminder of what that end of the business is. It competes on price per gigabyte.
So the mechanism to worry about isn't YMTC taking the crown in 2027. It's YMTC adding wafer capacity that doesn't need to earn a commercial return, in the segment of Samsung's memory business with the least defensible pricing, at a moment when Samsung's reported profitability is being driven almost entirely by price rather than volume.
That mechanism doesn't wait for 2027. It shows up whenever the shortage eases.
Here is the disclosure problem, and it's the reason this is hard to size.
Samsung's half-year report shows ₩305.4tn of revenue and ₩146.7tn of operating income for the six months to June. Memory sits inside Device Solutions. NAND sits inside memory. At no point does a published Samsung figure separate NAND revenue, NAND operating income or NAND average selling price from DRAM and HBM.
So an investor trying to answer "how much of this earnings power is exposed to a Chinese price war" cannot answer it from the filings. You can reason around the edges — HBM is the smaller volume and the larger margin, NAND is the larger volume and the thinner margin, historically NAND has been the first of the two to break in a downturn — but that's reasoning, not data.
I'd rather say that clearly than dress up an estimate. The number doesn't exist publicly. The absence is the finding.
The obvious counter is that YMTC's five-point gain was a shortage artefact. In a market where every fab is full, a marginal supplier picks up share simply by existing, because customers take whatever they can get. If that's what happened, the run-rate extrapolation above is too generous to YMTC and the 2027 date is even further off. This argument cuts in Samsung's favour and I think it's probably partly right.
The counter in the other direction is more serious. Export controls have limited YMTC's access to advanced tooling, and the general assumption has been that this caps its layer count and yields. If that constraint has loosened — through domestic equipment, or through stockpiled tools, or because 3D NAND scaling is more forgiving of older lithography than logic is — then the capacity ramp could be faster than the share data implies, and the last year is a leading indicator rather than a fluke. I don't have visibility into YMTC's toolset, and nobody outside the company really does.
There's also a version where none of this matters for two years because AI storage demand absorbs everything. That's the current consensus and it's been right so far.
The IPO prospectus, when it publishes. Chinese listing documents carry capacity and utilisation disclosure that YMTC has never had to make public before. Wafer starts per month and the layer roadmap are the two numbers that would turn this from an argument into a measurement.
NAND contract pricing relative to DRAM. If NAND decouples downward while DRAM and HBM keep rising, that's the entrant showing up in the price series, and it would show up there well before it shows up in a share table.
And Samsung's own segment disclosure. If the Q3 report starts breaking NAND out separately, it will be because someone decided investors need to see it. That decision is usually made for a reason.