Kioxia readies next-gen memory mass production as AI boom fuels dramatic comeback

Kioxia is set to begin mass manufacturing next-generation 10th-generation BiCS Flash NAND at its Kitakami (Iwate) fab, aiming to capitalize on the AI-driven memory demand shift from DRAM to high-capacity NAND. The stock has surged more than sevenfold this year, with market cap exceeding $250B and now larger than Toyota’s, reflecting a major turnaround. Kioxia is also considering a stock split and plans to list American depositary shares on a U.S. exchange in early FY2027 (starting April 2027).

Analysis

This is less a one-off fab event than a signal that NAND is being re-rated as AI infrastructure, not a commodity input. The biggest winner is the closest listed analog to the Kioxia/Sandisk ecosystem: if inference-driven storage demand persists, pricing power can extend beyond the usual inventory pop and support multiple expansion for SNDK. TOELY also screens as an indirect beneficiary because a genuinely tight advanced-memory cycle forces process/tool spending to stay elevated even if the broader semiconductor capex tape is choppy.

The second-order loser is the "HBM-only" consensus trade. If capital rotates into NAND, some of the air can come out of the most crowded DRAM/compute beneficiaries, while lagging memory peers face a harder catch-up path because node transitions and capex timing matter more than headline AI demand. The key risk is supply response: if competitors accelerate investment, today's scarcity narrative can morph into a 6-18 month oversupply setup, especially once these nodes are broadly qualified.

Near term, the market can keep bidding the theme for 1-3 months on scarcity and any upbeat commentary around utilization or pricing. The thesis weakens if hyperscalers show inventory normalization rather than sustained inference load growth, or if the public-market enthusiasm around an eventual U.S. listing pulls valuation forward faster than earnings do. The contrarian view is that the move may still be under-owned in listed proxies, but the headline-market-cap comparison is noise; only realized ASPs, mix, and capex intent will determine whether this becomes a durable re-rate or another cyclical spike.

Key Decisions for Investors

• Buy SNDK on pullbacks over the next 1-4 weeks as the cleanest listed proxy for an AI-driven NAND re-rating; target 15-20% upside if pricing commentary stays constructive, cut if utilization/ASP data flatten.
• Initiate a small long TOELY position as a second-order capex beneficiary; this works best if the next earnings cycle confirms that advanced-memory node upgrades are driving tool demand rather than just a one-name scarcity trade.
• Do not chase TM as a proxy for this story; treat any Japan-market sympathy bid as cosmetic unless there is evidence of spillover into broader semiconductor supply-chain orders.
• Set an alert for any public guidance that NAND ASPs or lead times are normalizing; that would be the first sign the scarcity premium is peaking and would argue for trimming SNDK/TOELY exposure.