Tech Headlines

Kioxia's Comeback: How the AI Inference Era Reshapes Japan's NAND Memory Landscape

Japan's memory giant Kioxia, driven by surging demand for AI inference, has surpassed Toyota in market capitalization and is preparing for mass production of next-generation NAND flash memory. This marks the revival of Japan's semiconductor industry in the AI era.

From "Troubled Sample" to "AI Winner": Kioxia's Comeback

In July 2026, Japanese memory manufacturer Kioxia is about to hold a mass production ceremony for its next-generation NAND flash memory at its factory in northern Japan. Once seen as a symbol of the decline of Japan's semiconductor industry, the company has staged a stunning turnaround driven by the AI investment boom—its share price has surged more than sevenfold this year, pushing its market capitalization above $250 billion, surpassing automotive giant Toyota to become one of Japan's most valuable companies.

Kioxia's predecessor was Toshiba Memory, which invented NAND flash memory in the 1980s, but later, due to industry cyclical fluctuations and the operational difficulties of its parent company Toshiba, it was acquired in 2018 by a consortium led by Bain Capital for 2 trillion yen (about $12 billion). In the following years, memory prices remained low, and Bain had to postpone its IPO plan until the end of 2024. However, the wave of AI has completely reshaped the market landscape.

The AI Inference Era: A "Second Curve" for NAND Demand

In the early stages of AI development, market focus was on DRAM chips used for data training, especially High Bandwidth Memory (HBM). SK Hynix, as a pioneer in HBM, was once the biggest beneficiary of AI memory demand. However, as AI applications expanded from large-scale training to inference (the process of answering user queries), demand for high-capacity NAND flash memory began to explode.

Satoru Oyama, an expert who once served as a consultant at Tokyo Electron, pointed out: "Chip manufacturers have prioritized DRAM development to such an extent that they have sidelined NAND investment and development. They are completely unable to cope with the current NAND boom, so demand is now concentrated among a few companies that can supply."

This supply-demand imbalance has created a historic opportunity for Kioxia. As the inventor of NAND technology, Kioxia has deep expertise in 3D NAND. Its upcoming mass production of next-generation memory is likely to further consolidate its leadership in the AI inference-related memory market.

A "Catalyst" for Japan's Semiconductor Revival

Kioxia's rise is not just a corporate victory but is also seen as a key signal for the advancement of Japan's semiconductor strategy. Over the years, Japan's position in global chip manufacturing has been in decline, but the structural shift in memory demand triggered by AI provides room for Japanese companies to compete through differentiation. Unlike the fierce competition with South Korea and Taiwan in DRAM and logic chips, the NAND flash memory field still retains Japan's technological heritage and manufacturing capabilities.

Kioxia's success has also drawn investor attention to other Japanese semiconductor companies. Tokyo Electron, Shin-Etsu Chemical, and other equipment and material suppliers are also benefiting from the AI-driven capital expenditure boom. The Japanese government's "Semiconductor Revival Plan" is accelerating the build-up of domestic manufacturing capacity through subsidies and technology collaboration.

Future Challenges and Industry InsightsDespite the bright prospects, Kioxia still faces cyclical risks. The NAND market is highly volatile; once the enthusiasm for AI investment cools down or DRAM manufacturers shift to increasing NAND production, the supply-demand balance may once again become unbalanced. In addition, Kioxia's capacity expansion plan requires enormous capital, and it remains questionable whether its own funds are sufficient to sustain continuous investment.

However, from a broader perspective, Kioxia's story reveals an important trend: the deep infiltration of AI is reshaping the semiconductor value chain. The training phase relies on high-performance DRAM and HBM, while the inference phase places higher demands on high-density, low-cost NAND. This evolution of the "memory hierarchy" creates opportunities for companies with unique technological advantages to turn things around.

For Japan, Kioxia's recovery proves that even after losing its leadership in logic chips, it can still occupy an irreplaceable position in the AI era by leveraging its long-term accumulation in specific memory fields. This might be the starting point for Japan's semiconductor industry to rebuild its confidence.

Editorial marker · japantechreview

japantechreview frames this note through Japan Tech Review explains Japan technology, robotics, semiconductors, mobility, corporate innovation and s...: dates, names and status changes still need checking. Tech Headlines / Robotics & Automation / Semiconductor Japan explains the local editorial angle; Source links should be opened before the summary is reused.

Source links

  1. https://www.reuters.com/business/autos-transportation/kioxia-readies-next-gen-memory-mass-production-ai-boom-fuels-dramatic-comeback-2026-07-02/Primary source

Related articles

Back to channel