Kioxia Holdings

Company history

Financial history 2022–2026 — revenue, cost structure, balance sheet, cash flow and key ratios, year by year →

Founded
2019
Head office
Minato-ku, Tokyo, Japan
Listed
2024
Founder
Toshiba Corporation (carve-out)
Revenue · FYE Mar 2026
$14.8B (¥2.34tn)
Net profit · FYE Mar 2026
$3.5B (¥555bn)
Kioxia Holdings: long-term performance & turning pointsSales (revenue) and profit-margin ratio
Sales (¥ bn)Net margin (%)

1951Toshiba’s memory business, and the invention of NAND

Revenue (¥ bn, bars) · net margin (%, line)
Source: securities reports
  1. 1951Toshiba prototypes a transistor at its Komukai works
  2. 1986World leader in 1-megabit DRAM; US–Japan Semiconductor Agreement
  3. 1987Masuoka Fujio presents the first NAND flash memory at IEDM
  4. 1991World’s first 4-megabit NAND product
  5. 1992Yokkaichi plant opens; NAND licensed to Samsung

Kioxia’s line runs back into the semiconductor business of Tokyo Shibaura Electric — Toshiba. The company prototyped a transistor at its Komukai works in Kawasaki in January 1951, early even by Japanese standards, and by 1956 was one of the first Japanese firms licensed to use Bell Labs and Western Electric transistor technology, alongside Tokyo Tsushin Kogyo (later Sony) and Hitachi. In the DRAM race of the 1980s Toshiba chose the planar cell and low-power CMOS over the harder-to-build three-dimensional structures its rivals pursued, and won on sheer manufacturability: at the 1-megabit generation it more than halved power consumption, lifted yields, and by 1986 was building a million parts a month — the world leader in that product category. The same year Japan overtook the United States in overall semiconductor share, which brought the retaliation of the US–Japan Semiconductor Agreement of September 1986.

The product Kioxia lives on today was invented in those labs. From around 1980 Masuoka Fujio led non-volatile memory work at the Kawasaki research centre, presenting a single-transistor flash cell — later called NOR — at the IEDM conference in 1984, and at IEDM 1987 the world’s first NAND flash memory, whose cells were wired together to raise storage density per unit of area. A colleague, Ariizumi Shoji, is credited with naming it “flash,” because erasing a whole block at once was as quick as a camera flash. Both the technology and the name came out of Toshiba.

Toshiba shipped the world’s first 4-megabit NAND product in 1991 and built the Yokkaichi plant in Mie prefecture in 1992 as its manufacturing base. In that same year it licensed NAND to Samsung Electronics, which used the licence as a foothold, invested at a scale Toshiba would not match, and became the largest memory maker in the world. The inventor left too: dissatisfied with how he had been treated, Masuoka moved to Tohoku University in 1994 and later sued Toshiba over compensation for the invention, settling in 2006. Invention and commercialisation parted company early here — a pattern that would repeat.

Read the full history in Japanese →


2000Betting the business on NAND alone

Revenue (¥ bn, bars) · net margin (%, line)
Source: securities reports
  1. 2000Flash joint venture with SanDisk
  2. 2001Exits commodity DRAM; Virginia fab sold to Micron
  3. 20073D flash technology developed — later BiCS FLASH
  4. 2011Memory and SSDs put into a dedicated in-house company
  5. 201548-layer BiCS FLASH samples ship
  6. 2016Western Digital acquires SanDisk

In 2000 Toshiba began working with SanDisk of the United States on flash memory and set up a 50/50 joint venture aimed at production in Yokkaichi. In December 2001 it walked out of commodity DRAM altogether, selling its Virginia fab to Micron Technology. Giving up the business that was fought over on price and concentrating resources on the memory it had itself invented left Toshiba’s memory operation built around NAND and its applications, with Yokkaichi as the core site — a division inside a diversified electrical conglomerate, but one running on an investment cycle of its own.

Yokkaichi expanded through the SanDisk partnership. Toshiba owned and operated the plant while the two firms split the capital spending and the wafer output down the middle, adding fab buildings through a succession of joint ventures — Flash Partners, Flash Alliance, Flash Forward. New 300mm buildings came up one after another, and each generation of shrink lowered the cost per bit. Against Samsung and SK Hynix in a permanent race of miniaturisation and capacity, the Toshiba–SanDisk pairing held its place among the leading suppliers of NAND in the world.

When shrinking approached its physical limit, Toshiba went vertical. It developed three-dimensional flash in 2007, later naming it BiCS FLASH, and began sampling a 48-layer part in 2015. In 2011 memory and SSDs were gathered into a dedicated in-house company, and as smartphones and data centres spread, memory grew into the single largest source of profit inside Toshiba — a business standing beside nuclear power and infrastructure, but with capital requirements and cyclical swings unlike anything else in the group. In May 2016 SanDisk was bought by Western Digital, so the Yokkaichi partner became a hard-disk company. The ever-heavier cost of each new NAND generation is what would soon decide the fate of the business, once Toshiba itself ran into trouble.

Read the full history in Japanese →


2017Carved out, sold, and renamed

Revenue (¥ bn, bars) · net margin (%, line)
Source: securities reports
  1. 2015Accounting scandal: $1.3B (¥152bn) of overstated operating profit
  2. 2017Westinghouse files Chapter 11; memory business split off
  3. 2018Sold to the Bain-led consortium for about ¥2tn (1 June)
  4. 2019Holding company formed; the group is renamed Kioxia

What pulled the memory business out of Toshiba’s hands was a crisis at the parent. In 2015 an independent committee found that Toshiba had overstated operating profit by roughly $1.3B (¥152bn) since fiscal 2008, and president Tanaka Hisao and other directors resigned. Then came the American nuclear subsidiary Westinghouse: construction cost overruns forced enormous goodwill impairments, and on 29 March 2017 Westinghouse filed for Chapter 11 bankruptcy. Toshiba faced negative net worth and, if that ran two years in a row, delisting. The most valuable thing it could sell was the memory business that was still growing. In February 2017 it set up Toshiba Memory Corporation as the receiving vehicle and on 1 April transferred the memory and SSD business into it by company split — a fast route out of negative equity, and a way to detach a capital-hungry business from the parent’s credit.

The auction drew Broadcom’s consortium, Hon Hai, SK Hynix and Western Digital, and bids climbed past ¥2tn. In June 2017 Toshiba chose a Japan–US–Korea consortium — the INCJ, the Development Bank of Japan, Bain Capital and SK Hynix — and in September signed with Bain’s acquisition vehicle, Pangea. Western Digital, the joint-venture partner, went to international arbitration arguing that its consent was required before Toshiba could move its interest to a third party, and sought to block the sale; that dispute was settled in December on terms extending and strengthening the joint venture. After Chinese antitrust clearance, the sale closed on 1 June 2018 at about ¥2tn — among the largest business sales Japan had seen. Pangea’s voting rights were split Bain 49.9%, a reinvesting Toshiba 40.2% and the optics maker HOYA 9.9%, so Japanese holders together held a majority; SK Hynix came in through convertible bonds, and Apple and Dell put money in without votes. Technology and production stayed in Japan; the logic of the capital was a foreign fund’s.

Under Bain, the company built itself a holding structure and a new name. In March 2019 a share transfer created Toshiba Memory Holdings above the operating company, and that October the whole group was renamed Kioxia — a coinage from the Japanese 記憶 (kioku, memory) and the Greek axia, value, picked from an employee competition. Dropping the Toshiba name declared independence from the old parent and an intent to compete worldwide as a pure-play memory maker. But independence meant funding the continuing multi-hundred-billion-yen investment at Yokkaichi and Kitakami on its own credit rather than Toshiba’s: in June 2019 it raised preferred shares taken up by the Development Bank of Japan and a syndicated bank loan. To carry generational investment on that scale through the swings of the memory cycle, it would need a thicker capital base — which made a stock-market listing the obvious next objective.

Read the full history in Japanese →


2020A failed merger, a delayed listing, and record profits

Revenue (¥ bn, bars) · net margin (%, line)
Source: securities reports
FY2022 · consolidated
Revenue$11.6B
Net income$806M
Net margin6.9%
FY2026 · consolidated
Revenue$14.8B
Net income$3.5B
Net margin23.7%
  1. 2020Planned TSE listing postponed in September
  2. 2023Western Digital merger talks collapse
  3. 2024Lists on the TSE Prime Market on 18 December at ¥1,455
  4. 2026Record year; Bain exits, Toshiba left the largest holder at 17.59%

Kioxia aimed at a listing almost as soon as it was independent, and twice stood down. In September 2020 it postponed a planned Tokyo listing, citing US–China tensions, the American export controls aimed at Huawei — a major customer — and a market softening under COVID. In parallel a combination with Western Digital was explored: a share-exchange proposal surfaced in 2021 and talks revived in 2023 around a new company headquartered in Japan. The two ran Yokkaichi and Kitakami on a fifty-fifty basis already; merged, they would have been a NAND maker on a scale approaching Samsung’s.

The talks collapsed in October 2023, and the reason was that a shareholder, SK Hynix, would not agree. Having joined the 2018 consortium through convertible bonds, SK Hynix held a large interest in Kioxia’s value and objected that a merger with a competitor would damage its own investment; terms with Bain, which had assembled Pangea, could not be bridged either. The plan to unite the two firms that had shared the lineage of the invention was defeated by a complicated coalition of capital. Western Digital afterwards turned instead to separating its own flash business as SanDisk.

Kioxia listed on the Tokyo Stock Exchange Prime Market on 18 December 2024 at an offer price of ¥1,455, closing its first day at ¥1,601 — four years after the first attempt, and directly after the worst of the cycle: in the year to March 2024 revenue was $7.1B (¥1.08tn) with an operating loss of $1.7B (¥253bn) and a net loss of $1.6B (¥244bn). Then demand from generative AI and data-centre SSDs turned it around — $11.4B (¥1.71tn) of revenue and $1.8B (¥272bn) of net profit in FY2025, and records across the board in FY2026 at $14.8B (¥2.34tn) of revenue, $5.5B (¥870bn) of operating profit and $3.5B (¥555bn) of net profit. That a loss and a record can trade places within two years is the other face of a business that makes one product.

The listing also rewrote the register. Bain sold down in stages and let go of its last holding in July 2026, exiting completely after eight years; Toshiba cut its stake too, to 17.59% at March 2026, but in a now-dispersed register that left the old parent as the effective largest shareholder — a fitting shape for this company’s history. Through the whole fund era, management stayed in the hands of Toshiba-trained engineers: the first president Naruke Yasuo, Hayasaka Nobuo who ran the company from 2020 to 2026 and took it public, and Ota Hiroo, who succeeded him in 2026. Ownership moved from a foreign fund to the market, but the technology Toshiba created, and the people carrying it, held the business together.

Read the full history in Japanese →


Key decisions — the author’s view

Revenue (¥ bn) · net margin % · around FY2017

Carving out Toshiba Memory: the Bain consortium, and the birth of Kioxia (2017)

A company born by leaving the hands that invented it

At the centre of this birth is a paradox: a business that was earning healthily was cut out of its parent for reasons that had nothing to do with its own performance. The NAND flash memory Toshiba was first in the world to create had, right up to the moment of separation, been producing most of the parent’s profit. It was sold not because it had failed, but because a parent overwhelmed by the failure of its nuclear business had no other way of turning an asset into cash at that price. A sale forced by crisis is what delivered an independent company into the world.

The same transaction looks different depending on where you stand — with Toshiba as seller, with the Bain consortium as buyer, or with the Kioxia that was bought into existence. For Toshiba it was the pain of giving up the family treasure; for Bain, an investment made in expectation of future appreciation; for Kioxia, the starting point of standing alone with the Toshiba name removed. Whether a business that became a separate company more than thirty years after the invention would grow — from a foreign fund’s capital as its point of entry — into a company that raises money from the market on its own account depended on the memory cycle and on what it made of that independence. Look at one deal from both sides at once, seller and buyer, and the history of Japanese semiconductors rises up as a single continuous story.

Revenue (¥ bn) · net margin % · around FY2021

Merger talks with Western Digital — and their collapse (2021)

A coalition of capital that blocked a union of the inventors

At the centre of this collapse is a paradox: the two parties that had shared the ground where flash memory was actually made — Kioxia, heir to the Toshiba line that gave the technology to the world, and Western Digital, its partner in joint production — could not be joined, defeated by a complicated coalition of capital. In the fabs the two could split output and investment down the middle; the roster of shareholders and their interests could not be halved the same way. The ownership structure itself, with a competitor, SK Hynix, sitting on the share register, is what blocked a merger that looked entirely natural in business terms.

That said, the merger falling through also meant that Kioxia chose to stand on its own. The year after the collapse it endured the trough of the memory cycle, listed on the TSE Prime Market in December 2024, and headed for record profits as demand recovered. Had the merger gone through, those fruits would have been shared between the new company and the former shareholders. In choosing to face the swings of the market with its own capital and its own technology rather than by fusing with a partner, the collapse became the point at which Kioxia settled its footing as a company standing alone.

Revenue (¥ bn) · net margin % · around FY2024

The TSE Prime listing, four years and two postponements in the making (2024)

Priced in the trough, earning in the peak

The character of this listing shows in how the four years and the two postponements were used. Both in 2020 and in 2024, what Kioxia stepped back from was a trough in the memory cycle, and the valuation investors put on it had fallen from roughly $14.0B (¥1.5tn) to about $5.3B (¥800bn). Since a pure-play maker’s price moves with the market it sells into, a company that avoided the trough before going to the capital market can be read as a private-equity exit opened by waiting for the terms to fall into place. Whether that waiting was right all the way to the end was, however, partly decided by the wave of demand that arrived after the listing.

The irony is that a listing twice withdrawn in a trough bore its fruit on the peak of AI demand. Little more than a year after listing, Kioxia posted a record operating profit of $5.5B (¥870bn) and its share price stood at several times the offer price. That harvest became the money with which the Bain that had waited eight years recovered its investment and withdrew — and what remained, once the owner of the capital had gone, was Toshiba, the company that invented this business and let it go. The very swing of pricing in the trough and earning in the peak traces the outline of a business that does nothing but NAND.

Each heading links to the full Japanese analysis — background, decision and outcome, with sources.


References & sources

This is a condensed English edition. The full, source-by-source history — with the detailed narrative, financial tables, shareholders and executives — is maintained in Japanese: 日本語版(詳細)— Kioxia Holdings full history in Japanese →

  1. Kioxia Holdings Corporation — 有価証券報告書 (annual securities reports) and investor materials.
  2. Kioxia — official account of the invention and naming of flash memory. History of Our Memory.
  3. Kioxia — chronology of the Yokkaichi plant, 四日市工場の歴史. Kioxia.
  4. F. Masuoka, M. Momodomi, Y. Iwata, R. Shirota — New ultra high density EPROM and flash EEPROM with NAND structure cell, IEDM 1987. IEEE Xplore.
  5. IEEE Spectrum — Chip Hall of Fame: Toshiba NAND Flash Memory. IEEE Spectrum.
  6. Computer History Museum — pioneers of semiconductor non-volatile memory, the first four decades. CHM.
  7. Semiconductor History Museum of Japan — 半導体歴史館 (1-megabit DRAM leadership; the US–Japan Semiconductor Agreement). SHMJ.
  8. Toshiba Corporation — IR releases: notice on the Westinghouse Chapter 11 filing, 29 March 2017; notice regarding closing of the memory business transfer, 1 June 2018. Toshiba.
  9. EE Times — Toshiba exits the DRAM market and sells its Virginia fab to Micron, 2001. EE Times.
  10. CNBC, Bloomberg, Fortune — reporting on the 2015 accounting investigation and the 2017 sale process. CNBC.
  11. Nikkei — 日本経済新聞 and Nikkei Asia: the Western Digital merger talks, the postponed offerings and the December 2024 listing.

Yen amounts are converted at the average rate of each figure’s own year — not today’s rate; revenue charts are shown in yen. Exchange rates & sources — the full ¥/US$ table →


Disclaimer


Data API

Kioxia Holdings’s history, financials, executives and shareholders are published as static JSON — no key, plain GET.

Method Endpoint Returns
GET /api/companies.json All companies
GET /api/285A/manifest.json Resource index
GET /api/285A/history.json History overview
GET /api/285A/timeline.json Chronology
GET /api/decisions.json All management decisions (index)
GET /api/285A/decisions.json Management decisions (index)
GET /api/285A/decisions/{slug}.json One decision (full dossier)
GET /api/285A/executives.json Executives
GET /api/285A/shareholders.json Major shareholders
GET /api/285A/financials.json Financial statements
GET /api/285A/financials-longterm.json Long-term results
GET /api/285A/segments.json Business segments
GET /api/285A/regions.json Sales by region
GET /api/285A/workforce.json Workforce