Financials
Financials — What the Numbers Say
Kioxia is a NAND flash memory pure-play that IPO'd on the TSE Prime on December 18, 2024 at ¥1,601. Eighteen months later the stock changes hands at ¥78,140 (June 5, 2026) — a 48-fold revaluation that briefly made Kioxia the second-most valuable company in Japan behind only SoftBank. The financial statements both justify and challenge that move: revenue has nearly tripled in two years, operating margin has swung from -23.5% to +37.2%, and free cash flow has flipped from deeply negative to ¥395B — but the entire reset rests on the AI-driven NAND up-cycle, and Q4 FY2026 alone delivered 68% of the year's operating profit. The single financial metric that matters most right now is the trajectory of operating margin in FY2027: management has guided 74% Q1 FY27 OP margin, and if that holds the current 76× trailing P/E collapses to ~10× — if it does not, the market has paid full price for a cycle peak.
Revenue FY26 (¥M)
Operating Margin FY26
Free Cash Flow FY26 (¥M)
ROE FY26
P/E (TTM)
P/E (Forward NTM)
Price / Book
Market Cap (¥M)
Enterprise Value (¥M)
The core tension in one line. Trailing P/E of 76× says expensive; forward P/E of 8.2× says cheap. Both are right — the question is whether the Q4 FY26 surge to 59.5% operating margin is the new normal or a cycle peak. Memory is one of the most cyclical commodities in tech; treat the gap as the size of the bet.
How to read this page
- Revenue and Margins — how scale and pricing turned a ¥253B loss into a ¥870B profit.
- Cash Quality — whether ¥554B of GAAP profit became cash (it did, mostly).
- Balance Sheet — the deleveraging from a 15.7% equity ratio to 37.9% in two years.
- Returns and Allocation — first-ever shareholder return policy lands FY2027.
- Segments — where the AI uplift is showing up (SSD and Storage).
- Valuation — what ¥42.7T market cap implies for the next NAND cycle.
- Peers — Kioxia vs Micron, SK Hynix, Samsung, Sandisk, Western Digital.
- Watchlist — the next five prints that will settle the debate.
1. Revenue, Margins, and Earnings Power
NAND memory is sold by the bit. Earnings power depends on two things — how many bits Kioxia ships, and the price per bit. Both swung violently between FY2023 and FY2026.
Annual revenue and operating profit (last five fiscal years)
The pattern is textbook memory cycle. FY2022 was the post-COVID peak (¥1.53T revenue, 22.7% non-GAAP OP margin). FY2023 to FY2024 was a brutal trough — NAND oversupply, hyperscaler digestion, and Chinese demand weakness collapsed revenue 30% in two years and pushed operating margin to -23.5%. FY2025 was a violent recovery (revenue +59%, margin +50pp), and FY2026 added another +37% revenue growth on top of that as the AI/data-center demand wave hit just as supply discipline tightened. Across two years revenue has 2.17×'d and operating profit has gone from -¥253B to +¥870B.
Margin structure over time
Operating leverage is extreme. Between FY2024 trough and FY2026 peak, revenue grew 2.17× while operating profit went from a ¥253B loss to a ¥870B profit. Each incremental yen of revenue dropped roughly 76% to operating profit on the way up. This is the signature of a high fixed-cost manufacturer — wafer fab depreciation and headcount do not flex with volume, so the entire delta in price × volume falls to the bottom line. The same lever works in reverse: in the FY2023 → FY2024 down-cycle, revenue fell 16% and operating profit fell ¥185B in a single year.
Quarterly trajectory — where we are in the cycle
The quarterly path is not smooth. After a strong first half of FY2025 (peak margin 34.5% in Q2), Kioxia walked through a sharp mini-correction — Q4 FY2025 margin collapsed to 10.7% on softer hyperscaler orders and inventory destocking. The rebound began in Q2 FY2026 and accelerated dramatically in Q4 FY2026: ¥1.0T of revenue in a single quarter (3× the prior quarter), 59.5% operating margin, and ¥408B of net income — more than the entire FY2025 net income in one three-month period. Management has guided Q1 FY2027 operating margin to 74%, the highest in the company's history.
This is what makes valuation hard: if you annualize the Q4 FY2026 run-rate, revenue is ¥4.0T and operating profit is ¥2.4T — but the data does not yet exist to know whether Q4 was a step-change or a spike.
2. Cash Flow and Earnings Quality
Free cash flow (FCF) is the cash a business generates after running its operations and after the capex needed to maintain and grow capacity. For a fab-owning memory maker, capex is the single biggest cash drain; FCF tells you whether the company funds growth from its own wallet or someone else's.
Net income vs. operating cash flow vs. free cash flow
Earnings quality has been steadily good in the up-cycle. In FY2024, the trough year, OCF was actually a healthy ¥195B positive despite a ¥244B GAAP net loss — because depreciation and amortization of ¥346B is non-cash and working capital released ¥146B as inventories were drawn down. The catch was capex: ¥304B on PPE meant FCF was -¥109B. In FY2025 and FY2026, OCF exceeded net income (OCF/NI of 1.75× in FY25 and 1.11× in FY26), which is exactly what you want to see in a recovery: depreciation continues to flow as cash, and tax payments lag profits.
What moves the cash gap
Three observations: (1) depreciation alone runs ~¥300B per year — a useful proxy for true maintenance capex burden; (2) interest expense more than doubled in FY2025 (¥82B vs ¥32B), because Kioxia refinanced at higher rates ahead of the IPO; (3) government grants of ¥44B in FY2025 (under Japan's semiconductor support program) are a real, ongoing tailwind to the investing line.
3. Balance Sheet and Financial Resilience
Two years ago Kioxia had a balance sheet built for a cyclical trough — 15.7% equity ratio, ¥1.1T of borrowings, and a current portion of debt (¥826B) larger than total cash (¥188B) that signalled refinancing pressure into the IPO. Today the balance sheet looks completely different.
Equity build and leverage trend
What changed
The capital structure transformation has three drivers:
- IPO proceeds (Dec 2024) brought ¥30B of fresh equity and refinancing optionality.
- ¥272B of FY2025 retained earnings rebuilt equity from the FY2024 loss.
- Active deleveraging in FY2025 — Kioxia repaid ¥266B of long-term borrowings and ¥126B of short-term net borrowings, cutting interest-bearing debt by ¥334B in one year.
Equity-attributable-to-owners rose from ¥450B → ¥738B → ¥1.40T across the three reported year-ends. Equity ratio has moved from 15.7% → 25.3% → 37.9%, and management has stated the target of a net-cash position by end of Q1 FY2027 with equity ratio rising toward ~60%. Interest coverage (operating profit / finance costs) flipped from -3.7× in FY2024 to +4.3× in FY2025 to ~10× in FY2026 — refinancing risk has effectively disappeared as a near-term issue.
Balance-sheet risk has largely been retired. With ¥471B of cash, near-net-cash by Q1 FY27, and operating profit covering interest by ~10×, the catalyst for capital-return policy (progressive dividends from FY2027, up to 50% of net cash flow) is now financially supportable. The financing-risk overhang that defined Kioxia in 2022-23 is no longer the binding constraint.
4. Returns, Reinvestment, and Capital Allocation
Returns on capital tell you whether management is creating value or just deploying more of it. ROE is profit divided by shareholders' equity; ROA is profit divided by total assets. For a capital-heavy fab business, both matter — high ROE without ROA can hide leverage; high ROA confirms underlying asset productivity.
Return ratios
Kioxia is now generating 51.9% ROE and 23.7% ROA on its FY2026 results. Industry economics make ROA the more telling number: 23.7% on a ~¥3.7T asset base is best-in-class for memory and above where Micron, SK Hynix and Samsung have historically printed at their cyclical peaks. Whether it lasts depends entirely on NAND ASP — at FY2024 prices the same asset base produced a -11.8% ROA. The asset base does not get cheaper; only the ASP changes.
Capital allocation — what the company did with the cash
Through FY2025, capital allocation was almost entirely capex and debt repayment — no dividend, no buyback, and just ¥30B raised in the IPO. Capex stepped down meaningfully from ¥304B (FY24) to ¥224B (FY25), which is what tipped FCF positive. The aggressive debt paydown of ¥324B in FY25 explains why total borrowings dropped ¥334B.
The capital allocation policy now changes. At its Investor Day on June 2, 2026, management committed to:
- Capex, research and development, and human capital first — fund growth at hurdle rates.
- Excess FCF returned to shareholders via a progressive dividend starting FY2027.
- Up to 50% of net cash flow as shareholder return (dividends, with flexibility for special dividends or buybacks).
- Disciplined mergers and acquisitions — only AI-adjacent inorganic targets that clear return hurdles.
There is no buyback yet, no historical dividend record to underwrite, and per-share book value has compounded from ¥868.86 (FY24) → ¥1,367.49 (FY25) → ¥2,561.74 (FY26) — a 1.95× in two years entirely through retained earnings rather than buybacks. For a freshly-listed cyclical, the cash policy is the right one. The risk is that it has not yet been tested through a down-cycle.
5. Segment and Unit Economics
Kioxia is a single reportable segment (Memory). The useful disaggregation is by product application and by geography, both disclosed in the Annual Securities Report for FY2025.
Revenue by application — where the AI uplift is showing up
SSD and Storage — enterprise/cloud SSDs and data-center storage — almost doubled (+92%) in FY2025 and now contributes 58% of revenue versus 48% the year before. This is the AI fingerprint. Smart Devices (mobile UFS/eMMC) grew 34%, consistent with smartphone unit recovery but no breakout. Management's stated target is 60% data-center revenue by FY2028, which would lock in the highest-margin, lowest-volatility part of the NAND mix.
Geographic mix and customer concentration
Customer concentration is real. Three customers contribute 40% of revenue — Apple alone is 17.6%. Apple is the mobile-NAND anchor; Dell is the SSD/data-center anchor; Sandisk's 11.6% reflects the Flash Ventures JV, where Kioxia and Sandisk jointly produce wafers at Yokkaichi and Kitakami and Sandisk buys a portion of the output. That JV is a strategic asset (shared capex) and a counterparty risk (any disruption to the JV agreement post-Sandisk-spinoff would be material).
Geographically, 89% of FY2025 revenue is outside Japan, with the US the single largest end market at 44.5%. JPY weakness has been a real tailwind: at the period-end FX rate, FY2025 revenue was ¥1.7T (£24% of which would have been a different number at 2022 yen).
6. Valuation and Market Expectations
This is the most important section in the report. The investor question is not "is Kioxia cheap" — it is "what does the market price already assume?"
Current multiples vs the cycle
The multiple split is the story. Trailing P/E is 76× — punishing — because FY2026 net income was depressed by H1 weakness and only Q4 delivered the explosive 59.5% margin. Forward P/E is 8.2× — extremely cheap — because consensus expects Q4 FY2026's run-rate (or higher) to hold. Both numbers are right; only one will turn out to be the right way to look at the stock. The fork is operating margin in Q1 and Q2 FY2027.
How the market values memory peers right now
Among the memory peer set, Kioxia trades at the highest P/E and P/B. SK Hynix — the AI-memory leader with the HBM franchise — trades at just 19.6× trailing P/E and 8.9× P/B, less than one-third of Kioxia's multiples on either basis. Samsung trades at 26× / 4.6×. Sandisk, Kioxia's JV partner with identical NAND exposure but a US listing, is the closest comp on P/B (18.9×) and trades at 60× P/E.
The peer-level read on Kioxia: it is priced as if AI-NAND is the single most attractive sub-segment of memory, which is plausible (Sandisk supports the read), but it is also priced as the cycle-peak winner with no normalization discount, which is harder to defend against SK Hynix's 19.6× multiple.
Analyst consensus
The dispersion is the warning. A spread between high and low price targets of ¥17,000 to ¥200,000 — that is almost 12× between the bear and bull cases on the same set of facts — is what you would expect in a stock where the cycle direction, not the company quality, is what's being debated. Recent analyst action has been positive (Morgan Stanley upgraded to Overweight in August 2025), but the average target sits only +10% above the current price, suggesting most of the easy upside has already been priced.
A simple bear / base / bull frame
7. Peer Financial Comparison
A side-by-side on the metrics that actually matter for a memory cyclical: growth, margin, returns, leverage, and valuation. Peer financials reference latest reported values; Kioxia is FY2026 (year ended March 2026).
The peer-gap that matters. Kioxia's operating margin of 37.2% and ROE of 51.9% beat what Micron, Samsung and SK Hynix are likely to print at this cycle stage — Kioxia simply rode the cycle harder because it is more concentrated (NAND pure-play, no DRAM/HBM dilution). The premium P/B vs SK Hynix is the price you pay for that concentration. Whether the premium is deserved depends on whether the AI-NAND mix shift is durable; the AI thesis says yes for SSDs into data centers, the cycle says be careful when everyone is paying the top.
8. What to Watch in the Financials
What the financials confirm — and what they contradict
The financials confirm that Kioxia is at the right place at the right time: a NAND pure-play in the middle of an AI-driven data-center capex surge, with operating leverage that compounds in the up-cycle (¥1.5T → ¥2.3T revenue produced ¥-253B → ¥870B of operating profit). They confirm that the balance-sheet rebuild is real (equity ratio 15.7% → 37.9%, interest coverage -3.7× → ~10×). They confirm that earnings are converting to cash at a respectable rate (OCF / NI ≈ 1.1× in FY26).
What they contradict is the implicit-pricing assumption that the Q4 FY2026 result represents a new steady state. The data does not yet support that — one quarter does not break a cyclical pattern that has been intact for 30+ years in the NAND industry. The 8.2× forward P/E only works if Q4's 59.5% margin holds; the 76× trailing P/E is what you get if it does not.
The first financial metric to watch is the Q1 FY2027 operating margin print (early August 2026). Management has guided 74%; the question is what it actually delivers and what guidance for the back half of FY2027 looks like. A print above 50% with rising guidance moves the forward P/E story toward base/bull. A print below 35% with cautious commentary on H2 FY27 ASPs would reset the entire valuation case.