Stacking, Not ShrinkingNarrow moat

Kioxia Holdings (285A) — moat facet

Flash stopped getting cheaper sideways a decade ago; everything since has been a manufacturing problem rather than a lithography one, which is why experience compounds here.

For most of semiconductor history, cost came down by making the cell smaller. Flash gave that up earlier than anything else, because a charge-storing cell has a floor below which it stops holding a charge reliably.

Why flash went verticalThe old leverSmaller cells - hit a charge-holding floorThe new lever321 layers, etched through in one channelWhat it takesManufacturing precision, not lithographyWhy that mattersExperience compounds; equipment does notKioxia moved mass production 176 to 238 to 321 layers.
Cost per bit is now an engineering problem rather than a printing one.

The answer was to build vertically: lay the cells in a stack, then etch a hole straight down through the whole thing to connect them. Kioxia's implementation is called BiCS FLASH, and it has moved mass production from 176 layers to 238 and then 3211. Each step is a manufacturing problem rather than a lithography one — drilling a uniform channel through three hundred layers without a defect anywhere along it is closer to precision engineering than to chip design.

That distinction is why this is the sturdiest facet of the company. Lithography can be bought; a hole that works three hundred layers down is learned.

It also caps what the advantage is worth. Every competitor is stacking, the layer counts are announced, and a customer choosing between a 321-layer part and a 300-layer part is choosing on price per gigabyte, not on architecture.

The measure is cost per bit, which nobody discloses. The visible proxy is gross margin in a quarter when prices did not move — and there has not been one of those in two years.

Moat trajectory: Holding steady

176 to 238 to 321 layers is a steady cadence, and the whole industry is executing the same transition. Being first buys quarters, not generations.

The number that tests this moat
Third-party estimate
Revenue, June 2026 quarter
¥1.77tn, a record

Adding layers lowers the cost of each bit without shrinking cells. Revenue this high on low single-digit bit growth shows the gains coming from price, not from stacking more bits; the next downturn will show what the cost advantage is worth.

Source: BigGo Finance, Kioxia Q1 FY2026 earnings call ↗
⚠ Threats to the moat
References
  1. ReportedBiCS FLASH mass production moved from 176 layers to 238 and then 321.
    Kioxia Holdings Corporation, Annual Securities Report for the year from 1 April 2025 to 31 March 2026 (8th Period) — revenue ¥2,337,628M against ¥1,706,460M, gross profit ¥1,012,904M, operating profit ¥869,013M, profit for the year ¥554,490M; research and development cost ¥141,052M against ¥132,798M; purchases of property, plant and equipment ¥281,062M against ¥223,847M; operating cash flow ¥616,540M; proceeds from government grants ¥56,396M against ¥43,748M, from an approved ceiling of ¥150.0bn for flash production at the Yokkaichi and Kitakami plants with about ¥31.8bn not yet received. Revenue by application: SSD & Storage ¥1,362,638M, Smart Devices ¥759,978M, Other ¥215,012M — 'Other' including retail products and sales to the Sandisk group recorded through the three manufacturing joint ventures. Revenue by geography: Japan ¥263,252M, North America and Europe ¥1,217,643M, Asia ¥856,733M, with the United States ¥1,098,832M, China ¥381,857M and Taiwan ¥300,932M. Non-current assets: Japan ¥1,658,950M, North America and Europe ¥1,986M, Asia ¥6,298M. Major customers: Apple group ¥476,014M (20.4%), with the Sandisk and Dell groups omitted for the year as each fell below 10% of sales. Flash Partners Ltd., Flash Alliance Ltd. and Flash Forward LLC are accounted for as joint operations with 50.1% of the voting rights and equal decision-making rights shared with Sandisk. Net interest-bearing debt ¥552,085M against equity of ¥1,398,929M — a net debt-to-equity ratio of 0.39 times, from ¥931,035M against ¥737,565M and 1.26 times a year earlier; USD-denominated senior notes at 6.25% (2030) and 6.625% (2033); goodwill of ¥395,585M from the 1 June 2018 acquisition of the former Toshiba Memory Corporation. Bain Capital funds indirectly hold 21.87% and Toshiba Corporation 17.59% of outstanding common shares, both having sold substantial holdings during the year. Risk factors state that the Yokkaichi Plant is located in an area with a high risk of earthquakes and floods and the Kitakami Plant in an area severely damaged by the 2011 Tohoku Earthquake, and name US-China trade frictions and US tariff policy among factors that may materially affect the business. The company does not provide plans or progress reports for the overall fiscal year. — year to 31 March 2026 · publ. 2026-06 · source ↗
Sources
Generated September 23, 2026