Kioxia Guided Flat Profit and Rising 2027 Bit Demand — and the Stock Fell 15.8%

Kioxia reported fiscal second-quarter results after the Tokyo close on Oct. 30, 2026, and the first session to price them was Nov. 2. Revenue of ¥2,480.0 billion was up 40.3% from the June quarter and above the ¥2,337.6 billion booked for the whole of fiscal 2025; operating profit was ¥1,996.2 billion at an 80.5% margin, net profit ¥1,355.7 billion, and diluted earnings per share ¥832.31 against the ¥778.57 the company had guided. December then came in sideways: revenue of about ¥2,500 billion and operating profit of about ¥2.0 trillion, which missed the roughly ¥2.35 trillion consensus by about 15%. The shares fell as much as 15.8% intraday, closed 13.3% lower at ¥11,680, and touched a six-month low.

The guide was not a demand call. Kioxia said NAND industry bit shipments grow 15–19% in 2026 and 20–24% in 2027, that its fiscal 2027 capacity is fully locked under long-term agreements, and that more than half of fiscal 2028 volume already sits under multi-year contracts. What it would not underwrite was price: the sequential rise in average flash prices slowed from about 70% in the April–June quarter to the mid-20% range in July–September, and TrendForce expects another 15–20% in the December quarter. Kioxia guided 2027 bit growth up to 20–24% while guiding profit flat — so the selloff priced the price curve, not the demand curve.

Strip out the currency drag and the guide is not flat. Kioxia assumed ¥157 per dollar against an average near ¥161, a headwind it put at roughly ¥55 billion of operating profit; add it back and December implies about ¥2.05 trillion of operating profit, some 3% above the September quarter, not zero. Consensus wanted ¥2.35 trillion, about 18% sequential growth. The gap between roughly ¥2.05 trillion and ¥2.35 trillion was the entire selloff, and it was a gap about the slope of the price curve rather than about Kioxia's ability to ship bits.

There is a quieter signal inside the guide. With contract prices rising 15–20% and revenue guided up only 1%, Kioxia's realized revenue per bit rose far less than the price index — either because the contracts it sells into lag the index, or because mix moved toward lower-priced products. Either way the index no longer converts one-for-one into revenue, and that is precisely what a market pricing peak-cycle earnings has to believe.

NAND's price curve is now set by its weakest buyer, and that buyer is no longer AI. TrendForce expects enterprise SSD bit demand to grow more than 80% in 2026 as inference deployments and agentic AI pull capacity into the data center, and enterprise SSDs are the only memory category where price increases are expected to accelerate in the December quarter. Client SSD runs the other way: PC makers are working down finished-goods inventory from the first half, cutting SSD capacity in mainstream models to defend bill of materials, and NAND suppliers have started quoting more flexibly. Kioxia expects consumer weakness to persist through 2027.

The mix gap is visible in the numbers. Micron's fiscal Q4 2026 data center SSD revenue reached nearly $10 billion, more than ten times a year earlier and over two-thirds of its $14.1 billion NAND line, while NAND was only 26% of the $54.23 billion quarter. SanDisk's fiscal Q1 2027 revenue of $11.03 billion, up 378% year over year, carried an 85.0% gross margin and a 79.6% operating margin. Enterprise SSD is where the pricing power now sits, and a NAND-only vendor with no DRAM business to cross-subsidise a client SSD price cut carries the exposure.

Capex is the one link that did not slow. Micron said it will spend more in fiscal 2027 than it previously planned, mostly construction spending to bring cleanrooms online beyond late 2028 with some equipment pulled forward, after $10.77 billion of net capex in the September quarter alone. Kioxia budgeted about ¥470 billion of capex for fiscal 2026, roughly ¥330 billion of it in the second half. Spending is shifting toward 2028 capacity, and the merchant toolmakers — Tokyo Electron and Advantest among them — are paid on schedules and cleanroom timelines, not on spot prices.

Downstream the chain has split into two markets. The hyperscalers are absorbing enterprise SSD supply at prices that are still rising, and they are the reason Micron's data center SSD line is more than ten times its year-ago level. The other end is the client market, where device makers defend bill of materials rather than buy more storage per unit. One price curve for AI infrastructure, another for the phones and notebooks sitting next to it — and only one of the two is still accelerating.

The dispersion is priced, not just traded. Kioxia trades on about 4.0x trailing EBITDA and 8.1x trailing earnings, with a 14% free cash flow yield, while Micron — which has a DRAM and HBM base Kioxia does not have — trades on 8.9x EBITDA and 11.7x earnings. The market pays roughly twice as much for AI memory with a DRAM base as for pure NAND at peak margin. That gap is the trade: the market is paying for duration where DRAM and HBM set the price, and discounting duration where a single product line does.

What 4x assumes is worth stating plainly. At an 80% operating margin, almost all of Kioxia's profit is price: cost of revenue was ¥426.9 billion against ¥2,480.0 billion of sales in the September quarter, and trailing depreciation of ¥307.5 billion barely dents ¥2,693.4 billion of net income. A 4x multiple on peak margin is a bet that the peak is real, and if the price curve bends further the same multiple becomes a trap rather than a bargain.

The demand call was never in dispute: Kioxia guided 2027 bit growth up, Micron guided December-quarter revenue to $61.5 billion, and enterprise SSD demand is compounding at more than 80% a year. What changed on Nov. 2 was the price of time. The market decided the shortage will now be paid for at a decelerating rate, and it repriced the names whose earnings sit one product line away from that decision. Own the bit curve through DRAM and HBM, and price NAND duration as the option it now is — with Kioxia at 4x trailing EBITDA, the trade is two-sided rather than one-sided, and being early on cycle timing is no longer automatically rewarded.