The top of Japan's market capitalization list has changed again! From “storage is king” to “banks are king”, capital returns to the cash flow theme at a time of AI pullback

The Zhitong Finance App learned that the market value of the Japanese banking supergiant, Mitsubishi UFJ Financial Group Inc. (Mitsubishi UFJ Financial Group Inc.), jumped to the top of the Japanese stock market's highest market capitalization list. This is the first time since the formation of the current three major banking groups that bank stocks have topped the Japanese market capitalization list. Since this year, foreign investors have generally regarded the Japanese stock market as the “second battleground in the AI computing power infrastructure industry chain”. As a result, NAND storage giant Kioxia briefly topped the Japanese market capitalization list, and the core business of most companies in the top 15 market capitalization rankings is increasingly focused on the field of AI computing power infrastructure. However, recently, the AI computing power theme has fallen into sharp fluctuations due to overcrowded positions and excessive leverage. Kioxia's market value fell from the top to third place in the list.

According to data compiled by the agency, the stock price of Japan's largest commercial bank, Mitsubishi UFJ rose 2.3% on Monday to a record high of 3,541 yen per share, bringing its market capitalization to 42 trillion yen (equivalent to 259 billion US dollars). This scale exceeds the market value of Toyota Motor Corporation of about 41 trillion yen and the market value of NAND memory chip giant Kioxia about 36.7 trillion yen. The stock price of Sumitomo Mitsui Financial Group, another major commercial bank, also rose sharply by 1.6%, which also reached a record high.

Mitsubishi UFJ Financial Group climbed to the top of the Japanese market capitalization list. Essentially, the main axis of capital inflows to the Japanese stock market gradually changed from “scarce artificial intelligence computing power and high beta momentum” to “interest rate normalization and profit certainty.”

On July 13, the market value of the Mitsubishi UFJ rose to about 42 trillion yen, while Kioxia fell to about 36.72 trillion yen after falling 12.86% in a single day. After the Bank of Japan ended negative interest rates and raised the policy interest rate to 1%, loan interest rates rose faster than deposit costs, and bank net interest spreads shifted from long-term compression to expansion; Mitsubishi UFJ's net profit for the 2025 fiscal year reached 2.4272 trillion yen, setting a record for the third year in a row, and the 2026 target further rose to 2.7 trillion yen, of which the increase of about 170 billion yen is expected to come from rising interest rates in yen. In other words, the market is using higher valuations to reward a business model that can directly convert changes in macro interest rates into current profits, dividends, and repurchases.

Since the Bank of Japan ended its negative interest rate policy in March 2024, Bank of Japan stocks have continued to rise, mainly because large commercial banks can charge higher interest rates on loans. This is in stark contrast to the banking industry's plight over the past few decades. After the bubble burst in the early 1990s, banks were burdened with huge amounts of bad debt, and aggressive monetary easing policies to combat deflation made it difficult to raise interest rates on loans.

Since raising interest rates for the first time in decades in 2024, the Bank of Japan raised interest rates four more times, raising the policy interest rate to 1%. Higher interest rates enable banks to profit from the widening gap between loan interest rates and deposit interest rates, thereby increasing profitability. According to the latest calculation data from the Mitsubishi UFJ analyst team, for every 0.25 percentage point increase in the future benchmark interest rate, its annual net interest income will increase by 180 billion yen.

Under the “booming demand for AI computing power” narrative, Kioxia gave way to the “interest spread revival” led by the Mitsubishi UFJ

Kioxia's loss of top of the list indicates that the AI computing power industry chain has recently moved from “unlimited demand” to a stage of “testing the expected rate of change”. Its stock price fell from an annual high of 112,700 yen on June 22 to 67,100 yen on July 13, a retracement of about 40%; on the day of July 13, many leading AI computing power players such as South Korea's memory chip stocks were severely impacted by factors such as heightened geographical risk, rising oil prices, liquidation of leveraged transactions, and the market questioning whether the high growth in AI profits could continue, and the Korean stock market once again plummeted to collapse.

Kioxia's previous valuation has also taken into account multiple optimistic assumptions such as rising NAND prices, long-term supply shortages, the explosion of AI inference demand, and the company's leading NAND flash memory technology. Therefore, even if the fundamentals of the AI computing infrastructure industry have not been reversed, as long as positions are overleveraged or bullish positions are overcrowded, AI capital expenditure growth, storage price increases, or AI application cashout speed is marginally cooled, the stock price will experience severe valuation compression.

It is worth noting that the exchange of market capitalization rankings between the two companies does not mean “banks replace technology” or “the storage supercycle driven by AI computing power is over,” but rather a typical rotation of factors: capital shifts from semiconductor momentum stocks with crowded positions and highly dependent on forward growth in valuations to financial stocks with higher visibility in cash flow growth, benefiting from nominal growth and rising interest rates, and strengthening shareholder returns. The Kioxia transaction is a storage demand curve brought about by AI inference over the next few years, while the Mitsubishi UFJ transaction is the net interest spread repair and cash profit increase that has already occurred today; when market risk appetite declines and discount rates rise, the latter naturally dominates.

Kioxia previously topped the Japanese market capitalization list for a short time. The core is that after AI demand spread from training to large-scale inference, the storage bottleneck began to extend from high-bandwidth memory to high-capacity NAND. Retrieving enhanced generation, long-term context, vector databases, inference caches, and mass generated content requires a larger non-volatile storage capacity than traditional enterprise computing; at the same time, memory chip manufacturers such as Samsung Electronics, SK Hynix, and Micron previously prioritized investing in high-bandwidth memory and dynamic random access memory, leading to relatively insufficient investment in NAND, creating a supercycle of “sudden acceleration in demand and timely response to supply difficulties.” Kioxia's stock price once rose more than seven times during the year, with a market capitalization of more than 250 billion US dollars. This is the result of the market's revaluation from a cyclical consumer electronics storage vendor to a core supplier of AI inference infrastructure.