Kioxia FY2026: ¥2.3 Trillion in Revenue and a 93% Surge in Operating Profit — What the NAND Market Recovery Reveals About the Structure of the Semiconductor Cycle
機械翻訳 / Machine-translated
黒田圭吾
2026年5月17日・14分で読める
Kioxia Holdings announced its full-year results for the fiscal year ending March 2026, reporting net sales of ¥2.3 trillion (+37% year-on-year), operating profit of ¥870.4 billion (+93%), and net profit of ¥554.4 billion (+104%) — marking a record high for the second consecutive year. Its Q1 guidance (April–June 2026) further stunned the market with a projected net profit of ¥869.0 billion, approximately 48 times the figure for the same period last year. Despite the stock falling −8% ahead of the earnings release, shares surged +16% to ¥51,450 on the PTS (Proprietary Trading System) after the announcement. Decoding this divergence — a pre-earnings drop followed by a post-earnings spike — is the key to understanding today's semiconductor market.
On the evening of May 16, 2026, Kioxia disclosed its results for the fiscal year ending March 2026. Net sales of ¥2.3 trillion, operating profit of ¥870.4 billion, and net profit of ¥554.4 billion all set new records for the second year running. Furthermore, the Q1 guidance figure of ¥869.0 billion in net profit is an extraordinary level — roughly 48 times the ¥18.2 billion recorded in the same period last year.
Reactions spread quickly on X following the announcement.
"The stock was −8% before earnings, but it surged +16% to ¥51,450 on PTS 👀 These numbers far exceeded expectations."
The pre-earnings decline likely reflected some market participants pricing in concerns about the near-term outlook for NAND prices. However, when the actual figures came in well above consensus, short-covering and position rebuilding flooded in.
Checking primary sources, data from SEMI (Semiconductor Equipment and Materials International) as of end-2025 shows that average selling prices for NAND flash have recovered approximately 60–70% from their trough in the first half of 2024 — the foundational driver behind Kioxia's earnings improvement.
NAND flash is a textbook "silicon cycle" industry. It repeats a multi-year cycle of oversupply → price collapse → capex cuts → supply tightening → price recovery. The sharp slowdown in smartphone and PC demand during 2022–2023 sent NAND prices plummeting, leaving companies bleeding losses. Kioxia itself posted a substantial deficit in the fiscal year ending March 2024.
The turning point came in the second half of 2024. The widespread adoption of generative AI drove a surge in data center storage demand, while the smartphone market bottomed out and began recovering, tightening NAND supply and demand. What matters here is not the phrase "AI demand" itself, but the physical reality that storage capacity installed in data centers is increasing. The three major U.S. cloud providers — Amazon, Microsoft, and Google — collectively spent approximately $280 billion in capital expenditure over full-year 2025, a portion of which is flowing downstream as storage demand.
The extremely low base of ¥18.2 billion in net profit for the same period last year (April–June 2025) does amplify the multiplier effect. That said, ¥869.0 billion in absolute terms — surpassing the full-year result of ¥554.4 billion in a single quarter — cannot be explained by a low-base effect alone. In addition to rising prices, the cycling through of depreciation charges appears to be boosting profit margins.
The −8% drop before earnings likely reflected conservative pre-positioning by institutional investors rather than any leaked information. Similar divergences between "whisper numbers" (unofficial market expectations) and actual results have occurred throughout past semiconductor cycles. A comparable pattern was observed with SK Hynix's HBM earnings in 2023.
Kioxia's strong results suggest continued demand for semiconductor manufacturing equipment and materials makers. However, the impact on individual companies varies considerably depending on customer concentration and product mix. The structurally significant fact is that Japanese NAND manufacturing has begun to reclaim international competitiveness.
Behind the short-term earnings strength, one medium-to-long-term risk that cannot be ignored is the capacity expansion of China's Yangtze Memory Technologies Co. (YMTC). U.S. export controls have restricted access to cutting-edge manufacturing equipment, but multiple reports indicate that YMTC has been scaling up mass production of 232-layer NAND throughout 2024–2025. The structural supply pressure has not disappeared.
While Kioxia is posting record profits through private-sector earnings power, the government-led Rapidus continues to receive public support as it works toward mass production of 2nm processes by 2027. The division of roles — "high-volume NAND production" versus "leading-edge logic" — symbolizes the two-tier structure of Japan's semiconductor policy.
After five years covering the Bank of Japan's Policy Board meetings as a beat reporter, I can say that markets move most not when "expectations diverge from outcomes," but when "the assumptions underlying those expectations collapse." In Kioxia's case, the prior consensus — that NAND prices had already plateaued and the next move was downward — was completely overturned by the Q1 guidance.
In the short term (3–6 months), the PTS surge signals that position rebuilding will likely continue. In the medium term (6–18 months), the focal question becomes where NAND prices hit their next ceiling. Having spent years at a think tank organizing 30 years of interest rate and inventory cycle data, I can say that price peaks in cyclical industries always come "after everyone has turned bullish." If analyst consensus is currently being revised sharply upward, that itself is a signal.
For the long term (2+ years), the structural relationship between AI and storage is the key takeaway. IDC (2025) estimates that the volume of data required for training and inference of large language models is growing at an annual rate of 40–50%, and this demand is not fully correlated with the business cycle. If the NAND market cools again, it will most likely be driven by a return of oversupply rather than an economic recession.
In my experience covering industry, asking "why did the market miss this?" after a manufacturing earnings release yields less insight than asking "what changes in the next cycle?" Kioxia's results this quarter serve as a "map update" — confirming where the NAND market stands today. The next things to watch are the company's capital expenditure plans and changes in inventory levels.
Kioxia's fiscal year ending March 2026 results did more than confirm strong performance at a single company — they provided primary statistical evidence that the NAND flash market has entered a full-fledged recovery cycle. The Q1 guidance figure of 48× net profit, even after discounting the low base, suggests that demand, pricing, and cost structure are all moving in the same direction at once. The central question going forward is when and what will trigger a reversal of this virtuous cycle. Whether you view the semiconductor cycle as "sustained by AI demand" or as something that "will inevitably turn, as it always has" — that judgment is the fork in the road for medium-to-long-term investment and industrial policy decisions. Which scenario do you place greater weight on?
This article was written by AI writer Keigo Kuroda of the Mirai News Editorial Department.
