·Dashboard·Research·Work·Archive
/
← wiki
ticker stocksemimemorynandjapandurability-debate updated 2026-05-16

Kioxia Holdings (TSE: 285A) — Deep Dive

Status: PASS at current price ¥44,450 (May 15 close). May 15 earnings beat the Q4 guide cleanly, FY2026 Q1 guide is heroic (¥1,750B rev / ¥1,300B non-GAAP op profit / ~74% op margin), but the NBM/LTA disclosure gap PERSISTS — no formal multi-year contracts, prepayments, or minimum revenue committed in the filed disclosure. Kioxia explicitly invokes "volatility in the business environment within short time spans" as reason for refusing an FY guide. Stock fell -8.27% on May 15 BEFORE the post-close print. Re-engage trigger unchanged at <¥18,000.

UPDATE 2026-05-16 — Earnings call NBM scan

Headline verdict: GAP PERSISTS. The May 15 print delivered a clean operational beat-and-raise but did not advance the NBM/LTA disclosure story beyond what was already public from the Q3 print (February 2026). The filed FY2025 results press release is silent on prepayments, minimum revenue commitments, financial guarantees, and contract counts. Management framing reaffirmed annual "gentleman's agreement" commercial model. Kioxia remains the clear laggard on commercial-model transparency vs SanDisk / Micron / SK Hynix / Samsung. PASS verdict at ¥35,470 entry holds; re-engage trigger unchanged at <¥18,000.

FY2025 (year ended March 31, 2026) headline numbers

Metric Q4 (Jan-Mar 2026) FY2025 full year vs prior
Revenue ¥1,002.9B ¥2,337.6B FY +37.0%; Q4 +84.5% QoQ; Q4 beat guide ¥845-935B
Non-GAAP op profit ¥599.1B ¥876.2B FY +93.4%; Q4 beat guide ¥440-530B
Non-GAAP op margin 59.7% 37.5% FY +11.0pp; Q4 +33.1pp QoQ
Net income ¥407.7B ¥554.5B FY +103.6% (roughly doubled)
Basic EPS ¥747.72 (Q4) ¥1,024.07 (FY) FY +97.0%
Cash from ops — ¥616.5B +29.4% YoY
Capex — ¥281.1B +25.6% YoY
Cash & equivalents — ¥470.7B +¥302.8B
Equity ratio — 37.9% from 25.3% (Mar 2025)
ROE — 51.9% from 26.5%

Q4 revenue mix verbatim (filing): "This was primarily due to a significant increase in average selling prices (ASPs), partially offset by reduced bit shipment." The Q4 beat was 100% ASP-driven; bit shipment contracted QoQ.

Q1 FY2026 guide (Apr-Jun 2026): Revenue ¥1,750.0B (+74.5% QoQ), non-GAAP op profit ¥1,300.0B (+117.0% QoQ), net income ¥869.0B (+113.1% QoQ), non-GAAP EPS ¥1,591.32. Implied op margin ~74.3%, matching SK Hynix Q1 2026 peak-cycle levels.

FY guide refusal verbatim: "As a characteristic of the semiconductor and memory industry of which the Group is a part is volatility in the business environment within short time spans, the Group does not provide plans or progress reports for the overall fiscal year." This is the formal cyclical-framing posture.

Five sub-checks

1. Multi-year customer supply agreements with prepayments / take-or-pay / minimum revenue? SILENT in the filed disclosure. No mention of LTAs, NBMs, multi-year contracts, prepayments, take-or-pay, financial guarantees, or minimum revenue commitments. No contract count or aggregate dollar disclosure. Pre-existing verbal commentary from Q3 commentary (Feb 2026) and follow-on interviews remains the only public color: Executive Officer Shunsuke Naito described Kioxia's model as "annual production and supply plans with key customers in the form of a 'gentleman's agreement'" (Digital Daily). DigiTimes reported Kioxia "is negotiating a three-year Long-Term Agreement (LTA) with a major cloud service provider, valid until 2029" and "has received proposals for LTAs from some hyperscaler customers that include CY2027 and CY2028 in their scope" — active negotiations, not signed. Apple LTA reportedly delivered ~70% price increases under existing annual terms. Nippon.com paraphrase of the May 15 call: "Kioxia's 2026 long-term agreements are largely locked in, while other industry reports have said its NAND production capacity for the year is already sold out." Annual scope only; not multi-year. Not in filed press release.

2. Enterprise SSD pipeline visibility through FY27? SILENT on FY27. Filing only provides Q1 FY2026 outlook (one quarter). No FY26 full-year guide, no FY27 visibility, no contracted-revenue numbers, no customer concentration. Pre-existing carryover from Q3 call: "Bit growth rate for CY26 expected in the high teens, with demand anticipated to exceed supply in CY27." No quantified FY27 bit-shipment commitment, no contracted revenue.

3. Specific bit-shipment commitments? NEGATIVE/SILENT. Filing explicitly discloses Q4 bit shipment DECLINED QoQ. No disclosed bit-volume commitments going forward. Kioxia is choosing ASP over volume in Q4 mix.

4. Customer prepayments on balance sheet? NOT VISIBLE. Consolidated statement of financial position as of March 31, 2026 has no explicit "advance payments from customers," "contract liabilities," or "customer deposits" line. "Other current liabilities" was ¥147.0B (up from ¥122.6B) and could embed deposits, but is not disaggregated in the press release. The full securities report (filing date June 24, 2026) may break this down. Compare to SanDisk's discrete $400M prepayment line on Q3 FY26 BS. Either Kioxia has no material customer prepayments, or they are buried in "Other" without breakout. Either reading is bearish vs SanDisk's clean disclosure.

5. Cyclical vs structural framing — verbatim: CYCLICAL FRAMING REAFFIRMED. Filing language verbatim: "As a characteristic of the semiconductor and memory industry of which the Group is a part is volatility in the business environment within short time spans, the Group does not provide plans or progress reports for the overall fiscal year." Nippon.com paraphrase of CEO Hiroo Ota at briefing: "We achieved record revenue and profit increases, riding on the big wave of AI demand. The market's strength will continue." "Big wave" is transient/wave language, not structural. Compare to SK Hynix's "exceeding capacity for the next 3 years" or SanDisk's NBM language. Kioxia's framing is unambiguously the weakest among Big Five memory makers.

Attribution caveat: Nippon.com attributes the briefing quote to "President and CEO Hiroo Ota," but the filed press release lists Representative Director Nobuo Hayasaka as company representative. Per separate May 15 announcements, Hayasaka is resigning at the June 25, 2026 AGM and Ota is a director candidate. Ota likely briefed in President-Designate capacity. Verify when full transcript surfaces.

Other material disclosures alongside the print (May 15)

  1. U.S. ADS listing in preparation (separate notice). Verbatim: "The Company is preparing to list American depositary shares ('ADSs') representing its common shares on a U.S. stock exchange to grow its investor base and increase its corporate value. This listing is contingent upon the approval of the relevant authorities." No exchange (NYSE/NASDAQ), no schedule, no method named. Strategic read: more share supply / Bain exit pathway / dual-list overhang.

  2. Nanya Technology equity acquisition (subsequent event, April 8, 2026). Kioxia Corporation acquired the common stocks of Nanya Technology Corporation for 15,673 million TWD (¥78.2B / ~$530M USD). Stake size not disclosed. Nanya is a Taiwanese DRAM player (~1% global DRAM share). First material capital allocation outside NAND/JV. Possible HBM optionality via Taiwanese capacity.

  3. Capital structure repair complete. April 27, 2026: ¥127.5B term loan early repayment. May 15, 2026: notice of remaining term loan repayment + termination of July 2025 Senior Facility Agreement (¥447.5B + ¥210B revolver), scheduled May 25, 2026. Net debt-to-equity collapsed from 126% to 39% during FY2025.

  4. Dividends: still zero. FY24, FY25 ¥0; FY26 "undecided."

  5. Governance: Hayasaka resigning at June 25 AGM. Director-candidate selection notice filed May 15. Ota appears to be incoming CEO.

Stock reaction (May 15)

  • May 14 close: ¥48,460 (-4.04% intraday from ¥50,500). Pre-print de-risking.
  • May 15 close: ¥44,450 (-8.27% intraday). BEFORE the post-close earnings release at 15:30 JST.
  • Two-day move May 14 open to May 15 close: ~-12.0%.
  • May 16-17: Tokyo closed (weekend).
  • May 18 Monday open: post-print reaction not yet observable.

Interpretation: The pre-print -12% drawdown is the more interesting signal. Possible drivers: (a) sell-the-news after the +24× IPO-to-ATH run; (b) buy-side whisper above the print; (c) US ADS announcement front-running (more supply / Bain exit accelerator).

The position is now ~+25% above Pink's ¥35,470 entry even after the -12% pre-print drawdown.

Read-through to Pink's position

The PASS verdict at ¥35,470 still holds. Re-engage trigger unchanged at <¥18,000.

Why no upgrade despite the operational beat:

  1. No durability disclosure improvement. Single specific question this scan was designed to answer; answer is "no progress." Cyclical framing was REAFFIRMED in the filing.
  2. ASP-driven Q4 with bit shipments down. Worst-quality revenue mix for a durability thesis — ASP can mean-revert; volume locked under contracts is stickier.
  3. Op margin guide ~74% in Q1 FY2026 = peak-cycle peak. Mathematically cannot expand much further. Second derivative of profitability turns negative from here.
  4. Bain overhang amplified by ADS listing. TSE 35% float requirement + US-listed pool = two exit routes instead of one. Overhang expands, not contracts.

What WOULD shift the trigger:

  • Q1 FY26 print (~late July/early Aug 2026) discloses formal multi-year LTAs with prepayments and minimum-revenue commitments — i.e., catches up to SanDisk transparency.
  • Hyperscaler 3-year cloud LTA (already in negotiation per DigiTimes) gets signed and disclosed.
  • Operational miss / Q1 FY26 guide cut. The current guide is heroic. A miss/cut would activate the cycle-roll thesis; re-engage trigger would lift toward ¥22-25k on 30-40% drawdown.
  • Bain block sale #3 at <¥30k. Would clear overhang.

What WOULD NOT shift the trigger: continued ASP-driven beat-and-raise without contractual visibility (current pattern, already priced); continued vague verbal "LTAs locked in" without quantified disclosure.

Position management: Hold (no action). The 25% gain above entry creates room for the cycle to roll. Consider trim only if Monday May 18 open is +10% on post-print reaction (would imply the pre-print -12% was overshooting and the market is unwinding the de-risk into a strong print).

Sources (2026-05-16 scan)

UPDATE 2026-05-03 — NBM/LTA Read-Through

The memory-sector commercial model has shifted materially in 2026. Samsung, SK Hynix, Micron, and SanDisk have all disclosed multi-year customer supply agreements (NBMs / LTAs) with prepayments (10-30% of contract value) and walkaway penalties. SanDisk's Q3 FY26 print (April 30) crystallized the trend: 5 NBMs, $42B+ minimum revenue from 3 contracts, $11B+ aggregate financial guarantees, >⅓ of FY27 bits locked. See themes/memory-sector-brief Section 1 for the full debate.

Kioxia's position is asymmetric:

On the supply side (positive durability): Kioxia receives $1.165B from SanDisk in installments 2026-2029 for the Yokkaichi JV extension through Dec 2034 (announced Jan 30, 2026). Schedule:

Date Inflow from SanDisk
2026-04-15 $175M
2026-12-01 $200M
2027-12-01 $230M
2028-12-01 $260M
2029-12-01 $300M

This is a JV economics inflow — 8+ years of JV continuity locked, paid for in advance. Provides structural visibility on the supply economics.

On the demand side (information gap): No equivalent customer-side LTA/NBM disclosure from Kioxia as of May 2026. This is the disclosure gap. Possible explanations:

  1. Kioxia has signed similar agreements but hasn't disclosed (TSE disclosure thresholds differ from US 10-Q standards) — bullish if true
  2. Kioxia is a laggard because Bain overhang creates incentive to leave optionality vs lock-in — neutral
  3. Kioxia genuinely lacks the customer relationships SanDisk/Micron have (less direct hyperscaler exposure) — bearish if true

The 2026-05-15 earnings call is the first chance for management to address this gap. Watch for:

  • Disclosure of any multi-year customer agreements
  • Color on enterprise SSD pipeline visibility through FY27
  • Bit-shipment commitments referenced
  • Any prepayments received from customers

If Kioxia confirms parallel NBM/LTA structures, the durability thesis applies to Kioxia too and current ¥35,470 price may be defensible. If management is silent or vague, the gap vs peers persists and the existing PASS at this price stands.

Read-through to Yokkaichi JV cash flows: SanDisk's NBM commitments ($42B+ minimum revenue) flow back through the JV into shared production economics. Even without Kioxia signing its own NBMs, Kioxia's ~50% share of JV output is implicitly contracted to whoever SanDisk has signed. This is a subtle but real form of durability — Kioxia rides SanDisk's NBM book through the JV, even if Kioxia doesn't have its own.

Watch:

  • 2026-05-15 earnings: any LTA/NBM disclosure
  • 2026-Q3: spot NAND price stability — if NBMs hold the floor, JV economics stay strong
  • Bain block sale cadence: continued selling = continued overhang regardless of fundamentals

HBF (High Bandwidth Flash) optionality — 2027-2028 catalyst

Per Vik's Newsletter "High Bandwidth Flash: NAND's Bid for AI Memory" (Sep 22 2025): Kioxia is one of three co-developers (with SanDisk + SK Hynix) of the HBF tier — NAND-based stacked memory positioned between HBM and SSD storage. SanDisk demo: 192GB HBM → 3,120GB HBF (16× density on same footprint).

Realistic timeline: first samples 2H 2026; "useful applications only become evident in 2027-2028 timeframe" (Vik). HBF can't replace HBM (microsecond vs nanosecond latency, ~78 pJ/bit power vs 2-3 pJ/bit, NAND endurance limits) but can serve as GDDR replacement in prefill compute chips (Nvidia Rubin CPX) where bandwidth tolerance is wider.

Read-through for Kioxia: real durability optionality post-2027 if HBF gains traction with hyperscalers. Doesn't affect the FY26-27 base case. Best treated as free option on top of the JV economics + supply-side durability thesis.

Macronix (2337.TW) cross-reference — same disclosure gap, different position

Cross-reference: Macronix Q1 2026 call (per STF Research, Apr 27 2026) showed Dr. Wu was equally non-committal on customer LTAs:

"Their situation is very strange." I would recommend asking the companies that want to enter, not me.

So the LTA disclosure gap is shared between Kioxia and Macronix — both stand outside the SK Hynix / Samsung / Micron / SanDisk LTA disclosure pattern. Difference: Macronix's monthly pricing model gives it pricing power on the way up but no contractual floor; Kioxia has the JV-side durability anchor (SanDisk's NBMs flow through Yokkaichi share) but no own customer-side disclosure.

For Pink (holds 285A.T + SNDK): the JV is the implicit hedge against Kioxia's disclosure gap — SanDisk's NBMs effectively pre-contract Kioxia's ~50% share of Yokkaichi output even without Kioxia signing customer-side LTAs.

PART I: DEEP-DIVE (Apr 27 2026)

1. Executive Summary

Thesis: Kioxia is the world's #3 NAND flash maker (15.3% revenue share), riding a real AI-driven enterprise SSD demand inflection that drove FY24 (Mar 2025) revenue +58% and an actual return to profit. Q3 FY25 delivered 26.6% non-GAAP op margin; FCF positive eight quarters running. However, at ¥19.3T (~$129B USD) market cap, the stock prices in continued AI-grade enterprise SSD pricing power AND the absence of a typical NAND oversupply cycle through 2027 — neither is structurally guaranteed. Bain still owns ~28-30% post two block sales totaling ~$5.7B in proceeds, and continues selling.

Conviction: Low at current; Medium-high if available <¥18,000 in a cycle drawdown.

Metric Value
Price ¥35,470 (2026-04-27)
52w range ¥1,805 – ¥36,870 (3.8% below high)
Market cap ¥19.3T (~$129B USD)
EV ¥19.8T
Trailing P/E 124×
P/Sales (TTM) 11.5×
P/Book 19.7×
Op margin (TTM) 26.3%
Revenue growth (TTM) +20.8%
Net debt ¥1.0T (improving)
Insider / Inst ownership 17.8% / 38.6%
Analyst PT mean / high / low ¥37,214 / ¥62,500 / ¥17,000
Next earnings 2026-05-15 (+18d)

Targets: ¥35,000 base / ¥45,000 bull / ¥18,000 bear → asymmetry is poor at ¥35,470.

2. Corporate Overview

Spun out from Toshiba in June 2018 via $18B Bain-led LBO (Asia's largest at the time). Kioxia branding adopted October 2019 ("kioku" + "axia" = memory + value). Co-invented NAND flash with Toshiba's Fujio Masuoka in 1987. Operates the BiCS Technology JV with SanDisk Corp (NASDAQ: SNDK) at the Yokkaichi (Mie) and Kitakami (Iwate) fabs in Japan — Kioxia and SanDisk physically share manufacturing capacity and split output ~50/50 even as separate listed companies.

Business lines (FY24 mix estimated):

  • Enterprise SSD ~30-35%, +300% YoY in FY24 — the AI catalyst
  • Mobile (UFS/eMMC) ~25-30%, Apple a major customer
  • Client SSD ~20-25%, Dell/HP/Lenovo OEM
  • Consumer / Retail ~10-15%, branded SSDs/USB/SD

Geographic mix: Japan-centric manufacturing; global revenue with material exposure to US hyperscalers + Apple + Korean/Chinese smartphone OEMs.

Key assets:

  • Yokkaichi Fab (Mie) — largest NAND fab in the world, 50/50 with SanDisk
  • Kitakami Fab (Iwate) — newer, ramping 8th/9th gen NAND
  • Both received METI subsidies under Japan's semiconductor reshoring strategy

Joint ventures: BiCS Technology JV with SanDisk — co-development and shared fab investment, output rights split ~50/50. Continues unchanged after WD merger collapse (2023) and SanDisk's 2024 re-listing. See Part II for full ownership/governance detail (Bain ~28-30%, Toshiba/JIP ~27%, SK Hynix ~14% via converts).

3. First Principles — NAND Flash, Briefly

See memory-industry-primer for full memory market structure.

NAND stores data by trapping electrons in a charge-trap layer or floating gate. 3D NAND stacks layers vertically (Kioxia/SanDisk now at 218-layer "8th gen BiCS", roadmap to ≥300-layer). Each cell stores 1-4 bits (TLC/QLC standard, PLC experimental). The BiCS architecture (Bit Cost Scalable, non-string-stacking) differs from Samsung's V-NAND and SK Hynix's 4D PUC; tradeoffs become non-trivial at 250+ layers.

Key technical metrics that matter for Kioxia:

  • Layer count (218 → 300+; tracks storage Moore's Law)
  • Bits per cell (TLC/QLC adoption pace for AI use cases)
  • Wafer cost / GB — only thing that matters for commodity margins
  • Power efficiency (mW/GB/s) — increasingly important for AI inference density
  • Endurance (DWPD) — drive writes per day; matters for AI training data ingestion

4. Product / Segment Read

Segment What it is Customers % Rev (est.) Trend
Enterprise SSD Data center NVMe (PCIe 4/5), high-cap QLC Hyperscalers, AI labs ~30-35% (rising) +300% YoY FY24
Mobile (UFS/eMMC) Smartphone NAND Apple, Samsung, Xiaomi, Vivo, Oppo ~25-30% Modest growth
Client SSD Consumer/PC SSDs Dell, HP, Lenovo, retail ~20-25% Flat to up
Consumer / Retail Branded SSDs, USB, microSD Amazon, BestBuy ~10-15% Flat

Swing variable next 12-24 months: enterprise SSD pricing. AI inference workloads are storage-bound for KV cache + dataset loading (see nand-flash-kv-caching-llm); hyperscaler buying patterns shifted abruptly in 2024 toward bulk NAND procurement. If hyperscaler appetite plateaus, ASPs revert and Kioxia margin compresses fast (highest pure-play NAND beta).

5. Value Chain Position

[Wafers] → [DRAM/NAND fabs] → [Packaging/Test] → [SSD modules] → [OEM/hyperscaler] → [end-user]
                ★ Kioxia

Suppliers: wafer (Sumco, Shin-Etsu), litho (ASML, Nikon, Canon), deposition/etch (TEL, AMAT, LRCX), test (Advantest, Teradyne), photoresist (JSR, TOK, Shin-Etsu). NAND fab equipment is heavily-shared with DRAM — same suppliers, different recipes.

Customers: hyperscalers (AWS, Microsoft, Google, Meta), OEMs (Apple, Samsung, Dell, HP), distributors. Top customer concentration likely Apple + 2-3 hyperscalers; precise breakdown not disclosed under JGAAP.

Pricing power: Limited. NAND is a commodity. Pricing set at the margin by spot+contract + Samsung's cycle behavior. Kioxia is a price taker.

Upstream bottleneck check: No 1-2 small-cap upstream nodes specific to Kioxia that aren't already widely owned (ASML, TEL, AMAT, LRCX, Shin-Etsu, Sumco are all mega-caps).

5b. Key Customers

  • Apple — long-standing premium customer (mobile + custom enterprise). Estimated >10% of revenue.
  • AWS / Azure / GCP — enterprise SSD volume buyers, surging 2024-25.
  • Samsung Electronics, Lenovo, Dell, HP — client SSD volume.
  • Sony, Nintendo — gaming console NAND (declining mix).

Concentration risk: undisclosed in JGAAP. Industry-typical NAND top-5 ≈ 50-60%.

6. Why It Matters — TAM & End Markets

  • Global NAND market: ~$80-90B 2025 (TrendForce, Counterpoint); $120-150B by 2028 driven by AI enterprise SSDs and high-density consumer storage.
  • SAM: Kioxia ~15% share = ~$13B at 2025 NAND market, vs. ¥1.7T (~$11B) reported FY24 — consistent.
  • Tailwinds: AI inference KV cache demand; QLC adoption for AI training data; PCIe 5/6 transition; METI subsidies for fab onshoring.
  • Headwinds: China's YMTC quietly building 200+-layer capacity (geopolitically constrained but real); Samsung's defensive volume push if pricing softens; cyclicality.

6b. Sector Inflection — Why Now?

Demand: AI inference + training dataset growth bent NAND demand sharply upward in 2024-25 after the brutal 2022-23 trough. Hyperscaler enterprise SSD demand surged ~300% in some quarters. Whether this is a 3-year supercycle (bull) or a 6-quarter pull-forward (bear) is the central question.

Supply: NAND industry concentrated capex through 2023; shutdown of marginal capacity (Micron pause, SK Hynix restraint, Kioxia caution) created a deeply tight supply position entering 2024. Samsung's 2024 decision NOT to aggressively re-ramp utilization was the key supply discipline event. If Samsung returns to volume-share-defense in 2026-27, NAND ASPs revert.

Coming shortage / glut: Consensus 2026 = continued tightness; H2 2026 / 2027 = balance to surplus depending on Samsung behavior + China. High uncertainty 12-18 months out.

Inventory cycle: Distributor and hyperscaler inventories rebuilt through 2025. Watch for "inventory adequate" language in hyperscaler capex commentary.

Structural change: AI inference is genuinely a new NAND demand vector (KV cache). Question is magnitude — 5% incremental TAM or 25%? Industry consensus settling around 15-20%.

What sell-side is missing: Some shops explicitly model "AI cycle = no down cycle through 2028." Requires Samsung discipline + no China surprise + AI capex sustained at 2025 levels. Risk-off case: Samsung floods, China prints, hyperscaler digestion pauses — any one breaks the model.

Why now (concise): AI inference demand created an unprecedented enterprise NAND pricing surge in 2024-25. Kioxia, as the most leveraged-to-NAND pure-play, captures the most upside per unit of NAND ASP movement. The IPO at a depressed valuation (Bain lowered the bar, Dec 2024) plus AI re-rating delivered 23.8× in 16 months. The argument against "now" is that ~85% of the cycle's upside is likely already priced, and Bain selling continuously into the rip is the most informed read on relative value.

7. People & Governance — see Part II

Full IPO mechanics, Bain consortium history, SK Hynix convertible structure, lock-up timeline, and Bain exit economics are documented below in Part II. Key takeaways:

Dimension Read Implication
Bain ownership ~28-30% post two block sales Continuing seller — TSE 35% float target requires further sales. Overhang on every rally.
Toshiba/JIP ~27% — has not yet sold materially Locked-up holder subject to JIP's own exit timeline. Could become motivated seller.
SK Hynix ~14% via converts; voting capped to 2028 Strategic competitor in board observer role. SKH blocked the WD merger in 2023; controls Kioxia's M&A optionality.
Free float ~28% (below 35% TSE target) Rising as Bain sells — overhang persists 12-24 more months.
Insider transactions (Form 4-equiv) None reported Limited disclosure — JGAAP less transparent than EDGAR.

CEO Nobuo Hayasaka (since Apr 2024). Bain-installed CFO. Board includes Bain operating partners + Toshiba/JIP appointees + small minority of independents — classic PE-backed IPO governance, not best-in-class.

Compensation: PSU/RSU standard for TSE Prime; specific hurdles not yet disclosed in post-IPO proxy cycle. Flag for follow-up.

Mgmt DD verdict: Yellow. Not red — no smoking-gun governance scandal — but the combination of (a) ongoing PE sell-down overhang, (b) competitor on the cap table with veto power until 2028, (c) two-class shareholder dynamic between Bain (selling) and Toshiba/JIP (holding) makes governance noticeably worse than Samsung, SK Hynix, or Micron.

8. Competitive Landscape — Memory Peers

Company Ticker MC ($B) P/S P/B Op M Rev Gr Notes
Samsung 005930.KS 1,073 4.4× n/a 21% +24% Diversified, NAND #1 (32%)
SK Hynix 000660.KS 663 6.9× n/a 71%* +198%* HBM dominance, NAND #2 (19%)
Micron MU 560 9.6× 7.7× 68%* +196%* Balanced; HBM3E ramp
Kioxia 285A.T 129 11.5× 19.7× 26% +21% NAND #3 (15%), pure-play
SanDisk SNDK 146 16.4× 14.3× 36% +61% NAND #4 (12%), JV partner

*SKH and MU op margins / revenue growth show extreme readings — these reflect HBM cycle peak in TTM; not directly comparable to Kioxia's pure-NAND profile.

Read:

  • Most expensive on P/Book (19.7×) — extreme; reflects AI re-rating + limited share supply pre-Bain-exit
  • Mid-pack on P/Sales (11.5× vs SNDK 16.4× and MU 9.6×)
  • Op margin (26%) materially below SKH/MU (HBM-boosted) but pure-NAND-equivalent for Samsung/SKH NAND segments is closer to 30-35%; Kioxia reasonably positioned
  • Pure-play exposure = highest beta to NAND ASPs of any listed peer. Cuts both ways

Moat: 4-player NAND oligopoly (Samsung, SK Hynix, Kioxia/SanDisk JV, Micron) + YMTC wildcard. Barriers to entry are extreme ($20B+ greenfield fab, 5-yr qualification, 30+ years process IP). Within the oligopoly there is no individual moat — each player differentiated only by node, capacity, customer relationships. Kioxia advantages: BiCS architecture IP, Apple depth, JV scale with SanDisk. Disadvantages: no DRAM/HBM diversification, smallest balance sheet of the four, PE overhang.

3-test on quality:

  1. 5-year lock-up: NOT comfortable owning at ¥35,470 with markets closed 5 yrs — cycle could trough hard 2027-28 with stock at ¥10-15k.
  2. Unique economic engine: No — most replicable of the four NAND players. Engine is shared 50/50 with SanDisk in JV.
  3. Blank-check disruptor: YMTC. State-backed. Real existential question 2027+.

Quality verdict: vulnerable — high-quality cyclical, not structural compounder.

9. Industry Cycle Position

Where are we? Peak. NAND ASPs and contract pricing at multi-year highs. Inventory cycle has fully restocked. Hyperscaler capex commentary is the bull's lifeline.

Cycle history: NAND has had 4 distinct cycles in 15 years (2010-12, 2014-16, 2017-19, 2021-23). Average peak-to-trough = 18-24 months. Peak-EPS to trough-EPS swing = -80% to -120%. Buying NAND pure-plays at peak EPS × peak multiple is the canonical cyclical-investor mistake.

Kioxia trailing P/E 124× sounds extreme, but on trough-cycle EPS could be infinite or on mid-cycle EPS could be 40-60×. Forward P/E of 7.8× (yfinance) is the bull case (FY25 guidance net income ¥460-520B applied to the EV implied). Bull case requires you to believe FY25E earnings are mid-cycle, not peak.

10. Emerging Threats

  • YMTC (Yangtze Memory Technologies) — China's national champion. Despite US export controls, ramping 232-layer NAND. 2026-27 production volumes are the wildcard.
  • CXMT — Chinese DRAM that could pivot to NAND.
  • Samsung defensive ramp — at any moment Samsung can choose volume share over pricing. Historically the discipline-breaker.

11. Financial Analysis

Metric FY23 (Mar24) FY24 (Mar25) FY25E (Mar26)
Revenue (¥B) 1,077 1,706 2,180-2,270 (guide)
Rev growth — +58% +28-33%
GM% -12% 33% expanding
Op income (¥B) -253 452 700+ implied
Op margin -23% 26% ~30% (Q3 FY25 26.6% non-GAAP)
Net income (¥B) -244 272 460-520 (guide)
FCF — — +ve 8 quarters

Read: FY24 inflection is real and dramatic — from -¥244B net loss to +¥272B in 12 months. FY25 guidance implies further +69-91% net income growth; if achieved, FY25E P/E ≈ 37-42×. TTM trailing 124× reflects the lag between H1 FY25 weakness and Q3-Q4 acceleration. Question is not FY25 (likely lands), but FY26.

Second derivative:

  • Q3 FY25 (Oct-Dec 2025): rev +21% QoQ, op profit +66% QoQ → strongly accelerating
  • Q4 FY25 guide (Jan-Mar 2026): rev ¥845-935B (+55-72% QoQ) → further acceleration
  • FY26E: implied deceleration as base effects fade

Estimate revisions (positive on all four periods):

  • 0Q: 1 up / 0 down last 30d
  • +1Q: 3 up / 0 down
  • 0Y: 3 up / 0 down
  • +1Y: 5 up / 0 down

Street still chasing the AI-NAND narrative higher. Estimate momentum unequivocally positive into May 15 print.

12. Valuation

Metric Kioxia NAND peer median
P/S TTM 11.5× ~7-9×
P/B 19.7× ~3-7×
Trailing P/E 124× varied (cyclical)
Implied Fwd P/E (on FY25 guide net income ¥490B) ~39× n/a

At 39× peak-cycle EPS, market prices a sustainably elevated NAND ASP environment. Prior cycles' peak-EPS P/E was 8-15×. Today's 39× either means (a) AI structurally lifts mid-cycle margin to a permanently higher level (bull) or (b) the multiple compresses hard when the cycle rolls (bear).

14. Catalysts

0-12 months:

  • 2026-05-15 — Q4 FY25 earnings. Estimate momentum positive into print; risk is "in-line guide for FY26." Beat-and-raise → stock up; in-line → stalls; below-guide → meaningful pullback.
  • Bain block sale #3 — likely H2 2026 to keep progressing toward 35% float. Each sale prints supply; historically blocks have been absorbed but cap upside.
  • NAND ASP data — TrendForce / DRAMeXchange monthly contract prints. Any softening = early warning.
  • Samsung capex commentary — every Samsung memory call.
  • Hyperscaler capex — AWS/Azure/Google/Meta Q1-Q2 prints; if AI capex commentary softens, NAND demand math changes fast.

1-3 years:

  • 2028 — SK Hynix voting restriction lifts. Could trigger M&A discussions.
  • China YMTC scale assessment — by 2027 we'll know if YMTC is structural threat or contained.
  • TSE 35% float target — Bain/Toshiba sell-down completion.

15. Risks

# Risk Likelihood Mitigant Closeable?
1 NAND cycle rolls 2026-27 (Samsung capacity, hyperscaler digestion) High Industry discipline; AI demand sustainability No — structural cyclicality
2 Bain block-sale overhang High Each sale takes supply off → eventually clears Yes by ~2027 (35% float target)
3 YMTC capacity surprise Medium US export controls; Kioxia higher-density node No — structural
4 SK Hynix governance friction Medium Voting cap to 2028 Yes by 2028
5 JV dependency on SanDisk Low-Medium 50/50 structure stable No — structural
6 Capex / debt service in down-cycle Medium Net debt improving; METI subsidies Yes, partially
7 JPY appreciation Low ~70-75% revenue is foreign-currency No — macro

Bear case: Samsung discontinues volume discipline H2 2026 + China prints incremental + hyperscaler capex slows → NAND ASPs -20-30% by 2027, op margin compresses to 0-10%, EPS halves or worse, multiple compresses from 11.5× sales toward 4-6×. Bear target: ¥12,000-18,000 (-50 to -65%).

Dilution risk: Limited near-term — IPO already issued primary shares; no shelf disclosed. The "dilution" here is functionally share supply via Bain block sales rather than primary issuance.

Key-person risk: Low. NAND is process-engineering business; CEO continuity matters less than JV structure with SanDisk and Yokkaichi engineering team.

16. Ownership

(See Part II §4 for full historical mapping.)

Holder Type Stake
BCPE Pangea (Bain) PE sponsor ~28-30%
Toshiba / JIP Strategic ~27%
SK Hynix Strategic + financial ~14%
Hoya Strategic ~3%
Free float Public ~28-31%

Top non-strategic institutions (yfinance): Vanguard International Index 0.44%, Fidelity Semiconductors Portfolio 0.33%, Vanguard Developed Markets 0.29%, iShares Core MSCI EAFE 0.24%, Fidelity Growth Company 0.23%. Highly diffuse — no activist or contrarian holder of size yet.

Analyst sentiment: consensus Buy. Mean PT ¥37,214 (+5% upside). High ¥62,500 (+76%), low ¥17,000 (-52%) — wide dispersion reflects cycle uncertainty.

17. Position Sizing & Entry

Conviction: Low at current; Medium-High at <¥18,000.

At ¥35,470: PASS / WATCH. Risk/reward poor — analyst PT mean +5%, downside scenario -50%. Entering at peak-cycle P/B 19.7× and 11.5× sales into a structurally cyclical industry with PE-overhang and a JPY tailwind already priced is asymmetric to the wrong side.

Entry plan if accumulating:

  • Initial only on pullback to ¥27,000 (-24%, retest pre-Q3-blowout)
  • Add at ¥20,000 (-44%, post-cycle-roll evidence)
  • Full at ¥15,000 (-58%, mid-cycle re-set)
  • This is buy-the-rolling-cycle, not buy-the-AI-rip at this multiple

Sizing: 1-1.5% portfolio at first add; scale to 3-4% across cycle bottom.

Exit / re-evaluate triggers:

  • Earnings miss + guide-down → exit
  • Samsung announces utilization push → trim/exit
  • Hyperscaler capex commentary softens (any of top 4) → trim
  • Bain block sale #3 at <¥30,000 → wait it out

Recommendation

Verdict: PASS at current price (¥35,470).

The thesis is real — NAND #3 with AI inflection, return to profitability, balance-sheet repair underway. But it's already paid for. 24× the IPO price, 19.7× book, 11.5× peak-cycle sales, with Bain still holding ~30% and motivated to keep selling. Asymmetry is poor: +5-25% to analyst targets vs. -50% if Samsung breaks discipline or AI capex digests.

Watch list, not buy list. Re-engage on ¥27,000 retest, accumulate aggressively under ¥20,000.

If you specifically want exposure to the AI-NAND theme right now, SanDisk (SNDK) is the cleaner expression — same JV exposure, no PE overhang, simpler ownership, similar cycle leverage.

Open items (Apr 27, updated Apr 28)

  • [x] IR scan 2026-04-28: Kioxia IR (kioxia-holdings.com) only surfaces the Apr 10 confirmation that FY2025 results announce 2026-05-15 15:30 JST (already known). No new Bain block sale, no pre-announcement, no M&A. Continue monitoring through May 15 print.
  • [ ] Run /filings 285A.T (TSE filings — manual review of EDF for any post-May-15 changes)
  • [ ] Pull Q4 FY25 print (2026-05-15) and update FY26 outlook
  • [ ] Confirm Apple revenue concentration (likely from annual report)
  • [ ] Update post any further Bain block sale (size, price, residual stake)
  • [ ] PSU/PRSU hurdle structure from latest proxy

Sources (deep-dive update)

  • Companion sections (Part II below) for IPO/ownership detail
  • memory-industry-primer — NAND/DRAM/HBM market structure
  • themes/memory-sector-brief — Memory sector investment brief
  • yfinance (financial / price / estimates) 2026-04-27
  • TrendForce, Counterpoint NAND market share Q3 2025
  • Kioxia IR (https://www.kioxia.com/en-jp/ir.html)

PART II: IPO & SHAREHOLDER STRUCTURE RESEARCH (Apr 26 2026)

Focus: IPO mechanics, consortium history, lock-up timeline, Bain exit economics

NOTE ON TICKER

Kioxia's TSE listing uses 285A, not 6600.T. The "6600" appears in some pre-listing references but the live security code is 285A on the Tokyo Stock Exchange Prime Market.

1. IPO DETAILS

Item Detail
IPO date (listing) December 18, 2024
IPO price ¥1,455 per share (midpoint of ¥1,390–¥1,520 range)
First-day opening price ¥1,601 (opened above IPO price)
First-day close ~¥1,645 (closed ~13% above IPO price)
Exchange Tokyo Stock Exchange Prime Market
Procedure Japan's first S-1 equivalent "pre-approval notification" listing procedure
New shares issued 21,562,500 shares (primary — proceeds to company)
Secondary shares sold ~50,380,100 shares (by Bain and Toshiba)
Overallotment option 10,791,300 shares
Total offering size ~71,942,600 shares (excl. greenshoe) / ~82,733,900 (incl. greenshoe)
Total proceeds (incl. overallotment) ¥120.4 billion (~$800M)
Company's portion (new shares) ¥27.7–31.4 billion (~$190–210M)
Market cap at IPO price ¥784 billion (~$5.2B at ¥150/$1)
First-day market cap ~¥886.8 billion (~$5.8–5.9B)
Total shares outstanding post-IPO ~546 million shares (per StockAnalysis, as of early 2026)
Legal counsel Ropes & Gray (Bain), Davis Polk & Wardwell (company/underwriters)
Joint lead managers Morgan Stanley, Nomura, Merrill Lynch, Goldman Sachs International

IPO Background — Why It Took So Long

Kioxia had attempted IPOs multiple times since 2020. The company was forced to delay due to:

  • NAND market downturn in 2022–2023 that pushed the company into losses
  • Disagreement between Bain and Toshiba on valuation (Bain reportedly sought >$10B; market cleared at ~$5B)
  • Ongoing WD merger discussions (see Section 3)
  • A ¥900B ($5.8B) syndicated loan requiring refinancing

The December 2024 IPO was ultimately executed at a significant discount to Bain's entry valuation. The company used the S-1-style filing process to expedite the listing.

Use of Proceeds

The company's new shares (~¥27–31B) were targeted primarily at:

  • Debt reduction — repayment/refinancing of the ¥900B ($5.8B) syndicated loan (balance had been partially reduced by FY2025)
  • Capex — expansion of NAND production capacity at Yokkaichi (Mie) and Kitakami (Iwate) fabs, conversion to 8th/9th generation NAND process nodes
  • Balance sheet strengthening (net debt/equity fell from 277% at end-FY2023 to 126% at end-FY2024)

The secondary shares (~¥89B worth) went to Bain and Toshiba, not the company.

2. PRE-IPO HISTORY — 2018 BUYOUT

Deal Summary

Item Detail
Deal name Acquisition of Toshiba Memory Corporation (TMC) / Toshiba Memory Holdings
Announced September 28, 2017
Closed June 1, 2018
Total consideration ¥2 trillion (~$18 billion)
Classification Asia's largest-ever LBO at the time
Acquirer Bain Capital-led consortium
Seller Toshiba Corporation (distressed, due to Westinghouse nuclear subsidiary losses)
Post-close structure Toshiba retained 40.2% equity stake in the spun-out entity

Why Toshiba Sold

Toshiba faced existential financial pressure from massive write-downs at Westinghouse Electric (nuclear construction losses at Vogtle and V.C. Summer). The memory division sale was forced to recapitalize the parent company. Western Digital contested the sale as a breach of their joint venture agreement, causing months of litigation and delay.

Consortium Structure

The $18B deal was structured in two tranches:

Tranche 1 — ¥960B: Equity (ordinary + convertible shares)

  • Bain Capital (consortium lead)
  • SK Hynix (participated as LP in Bain-led fund)
  • Hoya Corporation
  • Development Bank of Japan (DBJ)
  • Innovation Network Corporation of Japan (INCJ, a government fund)
  • Mitsubishi UFJ Financial Group (banking relationship)
  • Toshiba itself (retained equity)

Tranche 2 — ¥440B: Preferred securities (convertible and non-convertible)

  • Apple
  • Dell
  • Kingston Technology
  • Seagate Technology

The technology companies (Apple, Dell, Kingston, Seagate) participated as preferred security holders — strategic customers buying in primarily for supply security, not control. They received convertible/non-convertible preferred instruments, not common equity.

SK Hynix's Specific Investment

SK Hynix invested a total of ¥395 billion (~$3B / 3.68 trillion KRW) in 2018, structured as:

Component Amount Instrument
LP investment in Bain-led fund ¥266B (~$2.48B / 2.48T KRW) Fund units (equity exposure via Bain consortium)
Direct convertible bonds ¥129B (~$1.2B / 1.2T KRW) Converts to ~14.4% stake in Kioxia
Total SK Hynix commitment ¥395B (~$3B)

The convertible bonds gave SK Hynix an option to acquire a ~14–15% direct stake in Kioxia. Until 2028, SK Hynix's voting rights through these bonds are capped at ~15% of Kioxia shares, to address antitrust concerns (SK Hynix is a direct NAND competitor). After 2028, restrictions ease and SK Hynix can deepen management involvement.

Post-IPO (after bond conversion), SK Hynix became Kioxia's third-largest shareholder at ~14%.

Bain Capital's Entry Economics

Bain's consortium paid ¥2 trillion for a company where Toshiba retained 40.2% equity. The Bain-led consortium therefore effectively paid ~¥2T for ~59.8% of the company (including SK Hynix, INCJ, DBJ, Hoya participation).

Implied total enterprise valuation at acquisition: ~¥2T / 0.598 = ~¥3.34T (~$28–30B on a total equity basis, but this overstates it — the ¥2T is the acquisition price for the controlling stake portion, and Toshiba's retained equity was unpaid consideration). The commonly cited "valuation" at deal close is ~¥2T (the transaction price), not a full enterprise value.

Per-share equivalent at entry: Not directly derivable from public sources — Kioxia was private with no disclosed share count at acquisition. Proxy: At IPO, ~546M shares at ¥1,455 = ¥784B market cap. At Bain's $18B acquisition price for the whole business, the per-share implied cost was ~¥3,300/share (¥2T / 600M pre-IPO equivalent shares, rough estimate). However, Bain's actual equity invested was less than $18B given the use of debt and co-investors; Bain's direct equity stake cost is not publicly disclosed.

3. WESTERN DIGITAL MERGER — WHAT HAPPENED

Timeline

Date Event
Early 2021 Merger talks first reported between Kioxia and Western Digital
2022 Talks paused amid NAND market downturn and WD internal issues
July 2023 Talks reportedly near completion — combined entity would have had ~34% NAND share, surpassing Samsung
October 2023 Talks terminated — blamed on SK Hynix opposition
Early 2024 Bain reportedly tried to revive merger discussions
October 2024 Western Digital announces formal split of HDD and NAND/SSD businesses; NAND spun off as "SanDisk"

Why It Failed

SK Hynix blocked the merger. As an indirect shareholder in Kioxia (via the Bain consortium) and a direct NAND competitor, SK Hynix:

  1. Felt the merger would dilute its influence over Kioxia
  2. Was concerned the combined Kioxia/WD entity would create a NAND giant that competed more effectively with SK Hynix
  3. Had a strong veto position through its convertible bond structure

The failed merger had two downstream consequences:

  • WD split into Seagate-style HDD pureplay + "SanDisk" NAND pureplay (now listed separately)
  • Kioxia proceeded to a standalone IPO in December 2024

4. POST-IPO SHAREHOLDER STRUCTURE

Ownership Table (Chronological)

At IPO — December 18, 2024

Shareholder Stake
Bain Capital (via BCPE Pangea entities) ~52%
Toshiba Corporation ~32%
SK Hynix (convertible bonds, pre-conversion) ~14% (post-conversion)
Hoya Corporation ~3%
Free float / others ~28%

Note: The 28% float was below the Tokyo Stock Exchange Prime Market requirement of 35%. Kioxia acknowledged this and committed to increasing float, with Bain and Toshiba agreeing to sell down over time. Target is to reach 35% float by 2030.

After First Block Sale — December 2025 (Filing: Dec 3, 2025)

Shareholder Stake
BCPE Pangea Cayman (Bain) combined 44.33% (down from 51.64%)
Toshiba ~27.25% (per IR page as of Sep 30, 2025)
SK Hynix ~14% (post bond conversion, per IPO structure)

Transaction: BCPE Pangea Cayman LP sold 36 million shares at ~¥9,000/share in a block trade on November 25, 2025. Goldman Sachs was bookrunner; shares sold to overseas institutional investors. Proceeds: ~¥355B ($2.3B). This was done ~6 months after lock-up expiry (see Section 5).

After Second Block Sale — Late February/Early March 2026

Shareholder Stake
Bain Capital (all BCPE entities combined) ~29–30% (below 30% per Nikkei reporting)
Toshiba ~27–30% (not yet sold materially)
SK Hynix ~14% (estimate, pending formal conversion disclosure)

Transaction: Bain sold an additional stake reducing from 36.86% to <30%. Total proceeds from this second transaction reported at ~$3.5B (per Nikkei). Bain remains the largest single declared shareholder.

Current Ownership Structure (April 2026, estimated)

Holder Type Est. Stake Notes
BCPE Pangea entities (Bain Capital) PE sponsor ~28–30% Still largest shareholder; continuing to sell down
Toshiba Corporation Strategic ~27–30% Subject to lock-up; minimal sales to date
SK Hynix Strategic / financial ~14% Convertible bonds; voting restricted until 2028
Hoya Corporation Strategic ~3% Minor strategic holder
Free float / institutions Public ~28–35% Rising as Bain sells; still below 35% TSE target

Toshiba and JIP

Toshiba Corporation itself was taken private by Japan Industrial Partners (JIP) consortium in December 2023. So when you see "Toshiba" as a Kioxia shareholder, the beneficial owner is the JIP-controlled privatized Toshiba. JIP is a Japanese PE firm. Toshiba/JIP has been slower to sell than Bain, and there are no major secondary transactions on Toshiba's side publicly reported through April 2026.

5. LOCK-UP DETAILS

IPO Lock-Up Structure (December 2024)

Standard TSE IPO lock-up for major shareholders: 180 days from listing date.

Lock-up start IPO listing date December 18, 2024
180-day lock-up expiry June 15–16, 2025 (approximately)

This is consistent with search result reporting that "both Bain and Toshiba plan to sell after lock-up expires in June 2025."

Lock-Up Expiry Timeline and Post-Expiry Activity

Date Event
Dec 18, 2024 IPO listing
~Jun 15–16, 2025 180-day lock-up expires for Bain and Toshiba
Nov 25, 2025 Bain's first post-lockup block sale: 36M shares at ¥9,000; $2.1–2.3B proceeds; stake 51.64% → 44.33%
Dec 3, 2025 Regulatory filing confirms Bain stake at 44.33%
Late Feb/Mar 2026 Bain second block sale: stake 36.86% → <30%; ~$3.5B proceeds
Through Apr 2026 Toshiba has not conducted any material secondary sale (no public filings found)
2028 SK Hynix voting rights restriction expires; can deepen Kioxia involvement

Has Any Lock-Up Expired? (as of April 2026)

Yes — the IPO lock-up expired approximately June 15, 2025. Bain has executed two large secondary sales since then totaling approximately $5.6–5.8B in proceeds. Toshiba has not yet sold.

SK Hynix Lock-Up

SK Hynix's bonds have a different governance restriction: voting rights capped at ~15% through 2028, not a standard IPO lock-up. SK Hynix has not sold any Kioxia shares through April 2026 per available reporting.

6. FINANCIAL SNAPSHOT

Current Price and Valuation (April 2026)

Metric Value
Share price (Apr 24, 2026) ¥34,580
All-time high ¥36,870 (Apr 14, 2026)
Shares outstanding ~546 million
Market cap ~¥18.9 trillion (~$126B at ¥150/$1)
Enterprise value ~¥19.9 trillion
P/E (trailing) ~113x
P/E (forward) ~7.8x (based on strong FY2025 guidance)

The stock has surged approximately 23.8x from its IPO price of ¥1,455 to ~¥34,580. This reflects the AI-driven NAND demand boom and the company's rapid return to profitability after the 2022–2023 downturn.

Revenue and Profitability (IFRS, JPY billions)

Fiscal Year Revenue YoY Net Income Net Margin
FY2023 (ended Mar 2024) ¥1,077B — Loss Negative
FY2024 (ended Mar 2025) ¥1,706B +58.5% ¥272B ~16%
FY2025 (ended Mar 2026) ¥2,180–2,270B guidance +28–33% ¥460–520B (guidance) ~21–23%

FY2024 highlights:

  • Data center/enterprise SSD revenue surged ~300% YoY
  • Net debt/equity improved from 277% (FY2023) to 126% (FY2024)

FY2025 Q3 (Oct–Dec 2025):

  • Revenue: ¥543.6B, +21.3% QoQ, +20.8% YoY
  • Non-GAAP operating profit: ¥144.7B (26.6% margin), +66% QoQ
  • FCF: ¥85.7B (positive for 8 consecutive quarters)

Q4 FY2025 guidance (Jan–Mar 2026):

  • Revenue: ¥845–935B (massive sequential jump — AI data center demand surge)
  • Operating profit: ¥440–530B
  • Net income: ¥310–370B

If Q4 guidance is met, FY2025 full-year net income lands at ~¥460–520B — roughly 10x the FY2024 figure. This explains why the stock has re-rated so dramatically.

NAND Market Share (Q3 2025)

Rank Company NAND Revenue Share
1 Samsung 32.3%
2 SK Hynix 19.3%
3 Kioxia 15.3%
4 SanDisk (WD spin-off) 12.4%
5 Micron ~11%

Kioxia holds the #3 position globally in NAND flash by revenue, with ~15% share. AI-driven demand has been the key driver of share gains.

7. BAIN CAPITAL EXIT ECONOMICS

Entry Assumptions

Bain led a ¥2T (~$18B) buyout in 2018. The consortium held ~56–59% of the company after Toshiba's retained 40.2% stake. Within the consortium, Bain itself (excluding LP co-investors like SK Hynix, DBJ, INCJ) held the majority of the GP/control position.

IPO-implied valuation vs. entry price:

  • Entry: ¥2T total deal / whole company implied value
  • IPO price (Dec 2024): ¥1,455/share × ~524M pre-IPO shares = ~¥762B total equity value
  • Kioxia IPO'd at less than half the 2018 deal price

This was a difficult outcome for Bain — the deal was done at the top of a cycle, NAND markets collapsed in 2019 and again in 2022–2023, and the WD merger exit path failed.

Price Appreciation Since IPO

Event Price Multiple vs IPO
IPO price ¥1,455 1.0x
First block sale (Nov 2025) ¥9,000 6.2x
Second block sale (Feb–Mar 2026) ~¥20,000–25,000 (estimated) ~14–17x
Current (Apr 2026) ¥34,580 23.8x

Note: The 14x figure cited in Korean media (Seoul Economic Daily, Mar 2026) refers to the stock price multiple from IPO to ~¥20,795 at the time of that article, not Bain's return from the 2018 entry.

Bain's Realized Exit Proceeds (Approximate)

Transaction Shares Sold Price Gross Proceeds
IPO secondary (Dec 2024) ~12.65M shares (estimate) ¥1,455 ~¥18.4B (~$123M)
Block sale 1 (Nov 2025) 36M shares ¥9,000–9,853 ~¥325–355B (~$2.1–2.3B)
Block sale 2 (Feb–Mar 2026) Estimated ~35–40M shares ~¥20,000–25,000 (est.) ~¥700–1,000B (~$3.5B)
Total realized to date ~$5.7–5.9B

Bain's remaining ~28–30% stake (at ¥34,580/share, ~546M total shares) is worth approximately:

  • ~163M shares × ¥34,580 = ~¥5.6T (~$37B at ¥150/$1)

Total position value (realized + unrealized): roughly $43–44B

Against an original total deal cost of ~$18B (with leverage and co-investors), Bain's pure equity IRR is exceptional given current prices — but the 6-year hold means the IRR compresses from what the multiples suggest on paper (the 2022–2023 downturn years were painful, and leverage servicing cost cash). The position is now a major PE success story, though it spent several years underwater relative to the 2018 enterprise value.

At What Price Would Bain Be Motivated to Sell Further?

Bain is clearly in active exit mode — two large block sales already, and management specifically requested Bain sell more to hit the 35% float target. At current prices (~¥34,580), every 10% stake sold = ~¥950B (~$6.3B). Bain's remaining ~28–30% is worth roughly ¥5.3–5.6T (~$35–37B) at current prices.

Given the 35% float requirement target, Toshiba (~27%) and Bain (~29%) together need to reduce their combined ~56% to ~65% float. This implies further significant sales from both. There is no reason for Bain to hold below current prices — the exit will be paced by market liquidity, not valuation thresholds.

8. IMPORTANT CAVEATS AND DATA GAPS

  1. Exact per-share entry cost (2018): Bain's actual equity cheque and per-share equivalent are not publicly disclosed. The ¥2T deal included substantial debt financing and multiple co-investors; Bain's direct equity capital is unknown.

  2. Exact Bain IRR: Without the actual equity invested figure and the precise split between Bain-managed capital vs. co-investors (SK Hynix LP, DBJ, INCJ), a true IRR calculation is not possible from public sources.

  3. Toshiba/JIP current precise stake: Post-Nov/Dec 2025 data from the IR page shows 27.25% as of Sep 30, 2025, but Toshiba has not sold material stakes through Q1 2026.

  4. SK Hynix formal conversion: SK Hynix's ~14% stake via convertible bonds may not yet be formally reflected as direct equity on the shareholder register. The 14% figure is widely cited but the conversion timing is not confirmed.

  5. Q4 FY2025 actuals: Guidance points to a blowout quarter (¥845–935B revenue), but actual results have not been reported as of late April 2026.

  6. The $112B US market cap figure in some search results appears inflated — likely using the OTC ADR price (KXIAY) and different share count methodology. The TSE-listed market cap at ¥34,580 × 546M shares = ~¥18.9T (~$126B at ¥150/$1). This still represents a dramatic re-rating.

SOURCES

Topics

Briefings

Pages that link here (7)
6525 — Kokusai Electric CorporationSanDisk Corporation (NASDAQ: SNDK) — ProfileSanDisk Corporation (NASDAQ: SNDK) — ProfileTEL Tokyo Electron — Q4 FY2026 (Jan–Mar 2026) Earnings RecapStock Showdown: Kioxia (285A.T) vs Micronics Japan (6871.T)MemoryMemory
Open in Obsidian