TOKYO, 17 AUG 2026 — Kioxia has confirmed that its largest shareholder is now an investment vehicle backed by SK hynix, after Toshiba sold enough stock to fall into second place. The Korean company holds convertible bonds in that vehicle rather than shares in Kioxia, and has agreed not to take more than 15 per cent of the voting rights before 2028.
A NAND flash manufacturer's largest shareholder is a fund financed by a rival. That sentence is only possible because of a structure that also explains why nothing has changed at the board table.
What changed
The vehicle is BCPE Pangea Cayman2, a special-purpose company under Bain Capital, which held 77.4 million shares as at 3 August. Toshiba sold 5.436 million shares between 22 July and 3 August, taking its holding from 82.47 million shares to 77.04 million and its stake from 15.10 per cent to just over 14 per cent.
Toshiba's exit has been methodical rather than sudden. It held roughly 40 per cent at Kioxia's December 2024 listing, 18.52 per cent by March 2025, 16.10 per cent in May this year and 15.10 per cent in mid-July. The August sale simply removed the shareholder that sat above the Bain vehicle in the register.
The structure, which is the whole story
SK hynix does not own Kioxia shares. It joined the Bain-led consortium that bought the business from Toshiba in 2018, and its exposure since has run through two instruments: convertible bonds issued by the Bain vehicle, and a limited partner position in a Bain-managed fund. Reported figures put the combined commitment at about 395 billion yen, roughly US$2.48 billion, split between 129 billion yen of convertible bonds and 266 billion yen as a fund investor.
Convertible bonds in a holding vehicle are not voting shares in the underlying company. SK hynix therefore has an economic interest in Kioxia's performance and no direct vote on how Kioxia is run. The separation is deliberate; it is how a competitor was able to finance the deal at all.
Converting those bonds into voting control is not a decision SK hynix can take alone. It agreed to stay at or below 15 per cent of Kioxia's voting rights through 2028 unless Kioxia consents, and conversion would trigger competition reviews in several jurisdictions. Japanese approval is reported to be the highest hurdle, which is unsurprising given that Kioxia is the successor to Toshiba Memory and is treated domestically as a strategic asset.
Why a competitor holding the position matters anyway
An investor with no vote still has information rights, board access negotiated at the fund level, and an obvious interest in what the company does next. Kioxia has said as much itself: its own disclosures identify SK hynix's position as a risk factor and point to the potential conflict of interest that comes with a rival's growing influence.
A company naming its largest shareholder as a risk in its own filings is a useful signal about how the relationship is understood from the inside.
The market arithmetic explains the sensitivity. In the first quarter of this year SK hynix and its group held 17.6 per cent of NAND flash revenue and Kioxia 13.9 per cent. Combined, that is 31.5 per cent, against Samsung at 31.6 per cent. The two companies are separated from market leadership by a rounding error, and only when they are counted together.
No competition authority would treat that combination as routine. The voting cap and review requirement are not incidental legal furniture; they exist for this reason.
What this means for the memory market
NAND pricing has been the quiet story underneath the AI build-out. Storage demand rises with model training and inference, and the industry spent years in oversupply learning not to add capacity quickly.
A consolidation of the second and third largest suppliers would change the discipline of that market considerably. It would also change it for buyers in this region, where contract manufacturers, handset assemblers and the data centre operators covered in these pages purchase memory in volume and rely on there being more than two credible suppliers to negotiate against.
This week's disclosure does not move that outcome closer, but it makes the possibility legible. The largest shareholder is now the vehicle for a rival's investment, and the only barrier is a contractual cap with an expiry date and a regulatory process.
Kioxia's own position
The company itself has been visible lately for reasons unrelated to its share register. We reported last week that it is taking a PCIe Gen6 optical solid-state drive to customer trials, having said openly that cost and reliability still stand between the prototype and mass production.
This is the context for Kioxia's work on a technically ambitious product line. The ownership question is not one its management controls, and research programmes of that kind need several years of stable funding, which makes the identity and intentions of the largest shareholder a practical matter rather than a governance abstraction.
What we could not establish
Toshiba's exact residual stake. Reporting of the same filing gives 14.12 per cent and 14.06 per cent, a difference that probably reflects different share counts or dates rather than a contradiction, but we could not confirm which figure is current. The same applies to SK hynix's investment total, which appears as 395 billion yen in one account and as approximately 4 trillion won in another, measuring different things across different dates.
Also unestablished: whether SK hynix holds board seats or information rights through the fund, whether it has stated any intention to convert, what Toshiba intends for its remaining holding, whether the 2028 cap can be extended, and which specific competition authorities would need to clear a conversion.
What to watch
First, watch Toshiba. It has sold steadily for eighteen months and now holds just over 14 per cent. If disposals continue at this pace, the control question will arrive well before 2028, settled by who buys what Toshiba sells.
Second, listen for any statement from SK hynix about conversion. It has had no reason to make one, and a change in that silence would be the clearest possible signal of intent.
Finally, track Kioxia's own risk-factor language in successive filings. A company that has already named its largest shareholder as a risk will either soften that wording or sharpen it, and either direction tells you more than a share-register change does.