This may be the most important signal of AI market peaking — learning from Japan's lessons during the internet bubble period.

The AI rally continues to deepen, but questions about a bubble still linger over the market.

In a research report released on July 6, Morgan Stanley's Japan quantitative strategy team proposed that the key to determining whether the current Japanese AI rally is approaching a turning point lies not in valuation levels, but in marginal shifts in capital structure.

After reviewing the history of the U.S. internet bubble in 2000, the team found that the Nasdaq tech rally truly peaked not simply because of inflated valuations, but because domestic professional institutions in the U.S. reduced positions and exited simultaneously; correspondingly in the Japanese market, foreign institutions are the core capital force driving overall market trends.

For investors, the most noteworthy warning signal for Japanese stocks is foreign capital flows: once foreign investors shift from sustained net buying to net selling, and domestic funds such as retail investors and local investment trusts are forced to absorb large amounts of selling pressure, Japan's market capital structure will replicate the high-risk pattern seen before the U.S. internet bubble burst.

However, Morgan Stanley also notes that the current Japanese market does not yet exhibit the hallmark retail-investor overheating seen in the late stages of the U.S. internet bubble. Although the current Japanese AI and semiconductor rally has shown common bubble-phase characteristics such as capital concentration and outperformance by high-growth stocks, key indicators such as retail fund flows, domestic fund subscription volumes, and margin trading ratios have not reached the overheated levels of the U.S. bubble peak in 2000. This means that the current AI rally still lacks the core trigger for a substantive turning point.

When the Internet Bubble Truly Ended, Institutions Left First, Retail Investors Took Over

Conventional wisdom holds that the U.S. internet bubble was driven by frenzied retail buying. However, after reviewing several classic academic studies, Morgan Stanley found that this was not the case.

During the rapid expansion of the bubble from 1997 to 2000, institutional investors were the largest buyers of tech stocks, accounting for 63.6% of active buying, far exceeding retail investors and mutual funds. Hedge funds in particular took the most aggressive tech stock positions, but notably, they did not short in advance due to high valuations; instead, they continued to participate in the rally, gradually reducing positions before the market peaked.

The real turning point came after the market peaked in March 2000. As institutional funds began to exit en masse, their share of new buying in tech stocks plummeted to 36.4%; meanwhile, direct retail buying rose to 49%, and along with 14.6% flowing in through mutual funds, the market gradually shifted to a pattern where retail investors took over institutional selling, and tech stocks entered a sustained decline.

Thus, Morgan Stanley believes that the bursting of the bubble was not caused by "smart money" recognizing the valuation bubble in advance, but rather by institutional investors collectively choosing to exit at the same time.

Four Changes That Could Trigger Synchronized Institutional Withdrawal

The report suggests that changes in stock supply, interest rates, earnings expectations, and market sentiment could all act as catalysts for institutional repositioning.

Stock Supply: Financing Pressure Is Building

During the internet bubble, a flood of IPOs, lockup expirations, and insider selling were seen as key catalysts for the market peak. Currently, there has been a notable increase in IPOs, secondary offerings, and convertible bond financing in the AI supply chain. Although it is still too early to determine whether this will evolve into systemic risk, the pressure from stock supply warrants continuous attention.

Interest Rates: Growth Stocks Are More Sensitive to Financing Conditions

The peak of the internet bubble coincided with the Fed's rate hike cycle. In contrast, the market currently expects no further rate hikes this year and anticipates a rate cut resumption in March 2027. However, Morgan Stanley points out that if interest rates rise significantly above current expectations, high-valuation AI growth stocks could still face valuation pressure.

Earnings: Watch for Slowing Earnings Revisions

Unlike the internet bubble era, current AI leaders have stronger earnings foundations, and continuous upward earnings revisions have been a key support for the rally. However, Morgan Stanley reminds that more important than the level of earnings itself is whether the trend of upward revisions begins to slow, as this often affects stock prices first.

Market Sentiment: Are Bad News Being Amplified?

In the later stages of the internet bubble, the market became more sensitive to negative news about regulation, finances, and lockup expirations. Similar signs are now appearing in the AI supply chain, where stocks are showing increased volatility in response to IPO, capex, and regulatory news. Changes in sentiment deserve attention.

What You Really Need to Watch Is Whether Foreign Capital Starts to Retreat

Compared to the four potential catalysts above, Morgan Stanley believes the real determinant of whether the rally reaches a turning point is capital flows themselves. Before the internet bubble burst, the Japanese market saw sustained net selling by foreign investors while investment trusts continued to buy, with domestic capital gradually absorbing the supply.

In contrast, the current Japanese market shows a completely different pattern. Foreign investors remain strong buyers, and investment trusts have yet to show significant sustained inflows; retail investors overall remain net sellers, and while margin buying balances have reached their highest level since 2002, their ratio to total market capitalization is not significantly above historical levels, nor has there been the leverage overheating seen at the end of the internet bubble.

In other words, although the current AI rally shares many similarities with the late stages of the internet bubble—such as growth stocks leading and capital concentrating in a few leaders—the retail-driven overheating has not yet appeared, and the market structure remains importantly different from the eve of the 2000 bubble burst.

Risk Disclaimer and Terms

        Market risk exists; investment requires caution. This article does not constitute personal investment advice, nor does it consider the specific investment objectives, financial situations, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their particular circumstances. Investment based on this content is at one's own risk.