SK hynix and Micron: The Rivals Who Can Afford a Bad YearThin moat

SK hynix and Micron are Kioxia's competitors in the sense that they sell the same product to the same buyers. They are not its competitors in the sense that matters most, which is what a downturn does to them.

Both run NAND alongside DRAM, and in both cases NAND is the worse half — fewer participants in DRAM, better discipline, better returns, a point this collection has already made in detail from Micron's own pages. In the AI cycle the gap has become extreme: DRAM and high-bandwidth memory have re-rated both companies while their flash businesses have grown ordinarily.

The consequence is asymmetric survival. When flash breaks, their DRAM earnings absorb the fixed costs and the flash business is a segment that had a bad year. Kioxia has no such absorber, which is why it lost money in the years to March 2023 and March 2024 while its diversified competitors lost money for one year or none¹.

The counter-argument is that focus is worth something: management attention, capital allocation and a roadmap that never competes with a DRAM programme for the same fab.

The falsifier is a full cycle. If Kioxia's returns through the next trough are no worse than the flash segments of its diversified rivals, focus was the better structure. The last cycle said otherwise.