Kioxia —
the memory maker that became Japan's most valuable company

From a sub-¥1 trillion listing to Japan's market-cap crown

Kioxia's ascent is one of the most dramatic re-ratings in modern Japanese equity-market history. The company listed in December 2024 with a market value below ¥1 trillion. By the close on 12 June 2026, its market capitalisation had reached approximately ¥44.36 trillion — about ¥520 billion above Toyota — making Kioxia Japan's most valuable listed company at that moment.

A fifty-fold expansion in market value naturally invites the word “bubble.” But market capitalisation alone cannot answer the valuation question. Investors must separate three things: the unprecedented near-term NAND profit cycle, the sustainable mid-cycle earning power that remains after prices normalise, and the share-price premium created by scarcity, index flows and momentum. Our model says the first two are strong enough to explain roughly today's price. It does not say the stock offers a large margin of safety.

First, what does Kioxia actually make?

The easiest way to understand Kioxia is to separate computing from memory, and then divide memory into two very different jobs. A GPU or CPU performs calculations. Memory either feeds those calculations at extreme speed or stores data economically for later use.

Kioxia's proprietary 3D NAND is branded BiCS FLASH. The company manufactures flash memory and turns it into finished products for several markets. Its high-performance TLC products store three bits per memory cell and are suited to performance-sensitive SSDs. Its higher-density QLC products store four bits per cell, lowering cost per terabyte and enabling very large enterprise drives - including the company's 245-terabyte class products.

The earnings are not normal - even for a memory up-cycle

Kioxia's FY2025 results, for the year ended March 2026, show why investors have been willing to reprice the company so aggressively. Revenue increased 37.0% to ¥2.338 trillion. Non-GAAP operating profit rose 93.4% to ¥876.2 billion, while free cash flow reached a record ¥395.0 billion. The net debt-to-equity ratio fell from 126% a year earlier to 39%, helped by record earnings, refinancing and the redemption of preferred shares.

The fourth quarter was the real shock. Revenue nearly doubled sequentially to ¥1.003 trillion. Non-GAAP operating profit increased more than fourfold to ¥599.1 billion, equal to a 59.7% operating margin. Excluding low-margin joint-venture-related sales, adjusted gross margin reached 70%.

Crucially, this was not a volume story. Kioxia said its blended US-dollar selling price more than doubled quarter on quarter, while bit shipments declined by around 10%. In plain English, customers paid dramatically more for less physical NAND output. That is the signature of a severe shortage - and the reason investors should not extrapolate the quarter mechanically forever.

Management's June-quarter guidance is even more extreme: ¥1.75 trillion of revenue and ¥1.30 trillion of non-GAAP operating profit, a 74.3% margin. The company is also planning ¥450 billion of gross capital expenditure, launching tenth-generation BiCS FLASH, expanding eighth-generation output and repaying ¥400 billion of senior loans early.

Why we see the current price as broadly fair value

Our valuation deliberately avoids relying on a single peak-year multiple. For a cyclical memory producer, a low forward P/E during the best part of the cycle can be a trap: the denominator is temporarily enormous. We therefore combine scenario-based DCF, near-term EBITDA and P/E methods, and more conservative mid-cycle earnings methods.

The integrated result is ¥81,134, effectively flat versus the market price. The probability-weighted DCF is slightly lower at ¥78,068. This is the core conclusion: the market price is explainable by the earnings and cash-flow outlook embedded in current expectations. It is not necessary to assume irrational speculation to reach today's valuation.

But “not a bubble” should not be confused with “low risk.” The model's base DCF is only ¥74,221, and the mid-cycle EV/EBITDA method produces ¥65,383. The more optimistic valuation methods depend on applying multiples to FY2026-FY2027 earnings that are likely to represent the strongest part of the cycle. Kioxia may be fundamentally fair and technically overheated at the same time.

The peak-P/E illusion

At ¥81,200, the model shows a forward P/E of only around 8.5 times FY2026E earnings and 7.7 times FY2027E earnings. That looks cheap, but the normalised multiple is closer to 12.9 times. This is standard memory-cycle behaviour: P/E looks lowest when earnings are closest to a peak, then rises as profits normalise. The right question is not “Is 8.5 times cheap?” It is “How much of those earnings survive the next supply response?”

How the stock can still reach approximately ¥141,000

The model's full Bull-case DCF is ¥140,774 per share, or about 73% above the current price. This is not the central case and should not be presented as a conventional 12-month target. The model assigns the DCF Bull case a 25% probability, while its separate extreme-bull 12-month scenario is ¥130,000. In other words, the ¥141,000 outcome is a multi-year, through-the-cycle valuation scenario — not a forecast that the current quarterly earnings simply continue unchanged.

Where Base and Bull actually separate

FY2026 is not the main dividing line. Both scenarios assume a historic pricing windfall: the Base case uses a 209% increase in constant-mix NAND price per bit and an 80.9% EBITDA margin, versus 225% and 84.0% in Bull. The gap becomes economically meaningful in FY2027 and then decisive in the normalisation years. Bull assumes that pricing fades more slowly, shipment growth and product mix stay stronger, and margins settle well above the Base case even after the shortage peak has passed.

The five conditions that decide the Bull case

This hierarchy matters. The central operating question is whether mid-cycle EBITDA margin converges toward roughly 40% or remains above 46%. That outcome is produced by the interaction of price, cost and industry supply — not by any single headline about AI. Kioxia does not manufacture HBM; its HBM connection is indirect. Heavy HBM investment can restrict competitors' NAND spending, while AI inference can increase demand for Kioxia's enterprise SSDs.

What investors should monitor each quarter

The most useful scorecard is: (1) quarter-on-quarter NAND contract-price and blended ASP momentum, (2) Kioxia bit-shipment growth, (3) enterprise-SSD revenue mix, (4) cost-per-bit reduction and BiCS10 yields, (5) supplier capex, utilisation and announced capacity, and (6) inventory days. Together, these indicators show whether the business is tracking toward Base normalisation or Bull-level margin durability before the annual financial statements make it obvious.

The bear case is not theoretical

NAND is one of the most cyclical markets in semiconductors. High prices encourage suppliers to raise utilisation, accelerate technology migrations and add capacity. Customers then rebuild inventory, demand slows, and the same operating leverage that created extraordinary profit works in reverse.

The model assigns a 40% probability to a ¥72,000 base scenario, a combined 25% probability to bear outcomes of ¥45,000 or below, and a combined 35% probability to bull outcomes of ¥100,000 or above. The expected 12-month price is ¥76,800. This is a useful reminder: a stock can be close to intrinsic value while still offering an unattractive short-term probability distribution after a huge rally.

The bottom line

Kioxia is not merely a story stock. It is producing record revenue, extraordinary margins and rapidly improving free cash flow at a moment when AI-driven storage demand is colliding with constrained NAND supply. The current share price is therefore fundamentally defensible. Our integrated fair value of roughly ¥81,100 says the market has largely priced the central case correctly.

That conclusion is less bullish than it sounds. At today's price, investors are no longer being paid generously for ordinary execution. They are underwriting unusually high FY2026-FY2027 earnings and assuming the eventual margin normalisation is manageable. The model's own base DCF sits below the share price, while the 12-month expected value is ¥76,800.

For existing shareholders, the most important indicators are not the headline P/E or daily price action. Watch NAND contract-price momentum, Kioxia's bit shipments, enterprise-SSD mix, cost per bit and BiCS10 yields, industry capacity and utilisation, and inventory days. Those indicators will decide whether ¥81,200 proves to be a fair mid-cycle price, an early stop on the way to roughly ¥141,000, or the top of another memory cycle.

For informational purposes only. This is not investment advice or a recommendation to buy, sell or hold any security. Company figures are drawn from Kioxia Holdings' FY2025 Financial Results presentation dated 15 May 2026. Market price and valuation-model data are as of 12-14 June 2026. Forecast revenue, earnings, margins, scenario probabilities and fair values are analyst estimates, not official company projections. Kioxia provides quarterly guidance and does not provide the full-year FY2026 guidance used in the model. Memory pricing, foreign exchange rates, technology yields, customer demand and supply additions can change materially. Always verify figures against official company filings and conduct your own research.