--- title: "Kioxia's flat-price warning shows memory sellers lost the pricing war" description: "Kioxia posted a 22-fold profit jump and its first dividend, then told investors flash prices are going flat while capacity is sold out through 2027. The buyers locked their prices first, and that changes whose money is at risk in AI memory." publisher: "The Inference" section: "Chips" published: 2026-10-30T14:46:27.341Z modified: 2026-10-30T14:46:27.341Z canonical: https://theinference.org/article/kioxia-s-flat-price-warning-shows-memory-sellers-lost-the-pricing-war language: en keywords: "Kioxia, Memory, NAND, Sandisk, Intel, AI Data Centers" --- # Kioxia's flat-price warning shows memory sellers lost the pricing war > Kioxia posted a 22-fold profit jump and its first dividend, then told investors flash prices are going flat while capacity is sold out through 2027. The buyers locked their prices first, and that changes whose money is at risk in AI memory. The memory boom's most important number, the price of flash memory, has stopped rising on the supplier's own say-so. Kioxia, the Tokyo NAND maker that just posted a 22-fold profit jump and paid its first-ever dividend, [guided the December quarter to ¥2 trillion of operating profit](https://www.bloomberg.com/news/articles/2026-10-30/kioxia-s-profit-jumps-due-to-ballooning-ai-data-storage-needs), well below the ¥2.35 trillion analysts expected, and said its average selling prices rose only a mid-20% rate last quarter after 70% the quarter before [1]. Its executives said it is unclear whether prices keep rising in fiscal 2027 at all [1]. The stock fell 11% and now sits about 60% below its June peak [1]. ## Why a seller with sold-out capacity lost pricing power Because the buyers moved first. [Kioxia](https://theinference.org/markets/companies/kioxia-holdings-corporation) says long-term contracts with hyperscalers like [Meta](https://theinference.org/markets/companies/meta) and [Amazon](https://theinference.org/markets/companies/amazon) already cover well over half of its projected 2028 shipments, on top of capacity sold out through 2027 [1]. The people paying for NAND locked their price while the fear was theirs; now the risk of a price rollover sits on the seller's side of the table, and with it the multiple memory shareholders paid. That is the mechanism behind Friday: Kioxia down 11% [1], partner [Sandisk](https://theinference.org/markets/companies/sandisk) down 8% despite [beating estimates](https://finance.yahoo.com/markets/stocks/articles/sandisk-beats-fiscal-q1-earnings-103200762.html) [3], and [Intel](https://theinference.org/markets/companies/intel) down roughly 28% after [blaming memory prices for a PC slowdown](https://finance.yahoo.com/technology/article/intel-and-sandisk-have-very-different-views-about-whats-going-on-in-memory-chips-114638641.html) [2]. *Chart: **Sandisk's AI data-center revenue now rivals its edge business while consumer memory shrinks** Bar chart of Sandisk fiscal Q1 segment revenue: edge $5.83 billion, data center $4.80 billion, consumer $397 million, showing data center up 61% sequentially while consumer fell 29%.* *Sandisk fiscal first-quarter (ended October 2, 2026) revenue by segment, US dollars. Data center rose 61% sequentially, consumer memory fell 29%. Source: Sandisk fiscal Q1 results [3]. [3]* ## One industry, two opposite stories The tension inside the industry is now explicit. Sandisk's CEO told investors every NAND segment grows next year and "we see improvement from here" [2]. Kioxia's own guidance says prices go roughly flat. Both can be true only if volume, not price, carries the next leg of the boom, which means the sellers need buyers who keep consuming at today's prices. Intel's results suggest at least one buyer class, PC makers, cannot [2]. Kioxia plans a US listing next year to help fund [more than ¥5 trillion of capex with Sandisk](https://theinference.org/article/sandisk-and-kioxia-bet-31-billion-on-ai-storage-demand-and-sandisk-owes-half-the-fixed-costs) [1]. Investors should ask one question before they buy that supply: what price does it meet? If the contracts say the answer is "already agreed, and lower than spot," the reward for financing the buildout belongs to the hyperscalers who locked it in, not the memory shareholders now holding the stock. Start from the contracts, not the shortage. Readers deciding what to do with this should start learning how AI infrastructure actually gets financed, because the term sheets, the take-or-pay clauses and the capex schedules, not the hype cycle, now set who keeps the AI memory profits. ## Takeaway Kioxia's guidance shows flash price momentum is stalling because hyperscalers locked long-term prices early, shifting the risk of any memory-price rollover onto sellers. Volume, not price, must carry the next leg of AI memory, and investors financing new fabs should start from the contracts, not the shortage. ## Sources 1. [Bloomberg, Kioxia Warns Flash Memory Prices Stalling After Outlook Miss, 30 October 2026](https://www.bloomberg.com/news/articles/2026-10-30/kioxia-s-profit-jumps-due-to-ballooning-ai-data-storage-needs) 2. [Yahoo Finance, Intel and Sandisk have very different views about what's going on in memory chips, 30 October 2026](https://finance.yahoo.com/technology/article/intel-and-sandisk-have-very-different-views-about-whats-going-on-in-memory-chips-114638641.html) 3. [Yahoo Finance, Sandisk Beats Fiscal Q1 Earnings Estimates, Raises Outlook and Announces 8-for-1 Stock Split, 30 October 2026](https://finance.yahoo.com/markets/stocks/articles/sandisk-beats-fiscal-q1-earnings-103200762.html) --- Kioxia's flat-price warning shows memory sellers lost the pricing war — The Inference. Canonical: https://theinference.org/article/kioxia-s-flat-price-warning-shows-memory-sellers-lost-the-pricing-war