Research
专题5月14日 · Morgan Stanley

Japan Quantitative Strategy: Echoes of the Dot-Com Bubble in Current AI-Driven Rally

中文EN

Japan Quantitative Strategy: Echoes of the Dot-Com Bubble in Current AI-Driven Rally

Core Conclusion

The current rally in Japanese equities, driven by AI/semiconductor and high-momentum stocks, exhibits factor return patterns nearly identical to the dot-com bubble of 1999–2000. While index-level valuations (NASDAQ forward P/E 24.9x, SOX 21.5x) remain below bubble extremes, individual stock valuations within the AI Enablers basket have surged, creating a divergence that mirrors the final stage of the dot-com mania. Investors are compelled to ride the trend but must prepare for a potential reversal within a six-month window.

What the Market Is Underappreciating

The market likely underestimates the structural similarity of this rally to the dot-com bubble—not just in index trajectory, but in the precise pattern of factor returns. Using Mahalanobis distance across 13 mid-level factors (value, momentum, quality, low volatility, size, etc.), April 2026’s factor return profile shows the shortest distances to months from June 1999 through February 2000. During that period, five months recorded a Mahalanobis distance below 8.0, the historical bottom decile. The value factor weakened (12M forward earnings yield, book-to-price, dividend yield all fell), while momentum (12×1M momentum, EPS revisions) surged—an exact replica of the current environment. This deep factor-level alignment is widely ignored, as most commentary focuses only on index price action.

Evidence Chain

Factor return structure is a near-match. The 13-factor Mahalanobis distance from April 2026 reveals that the closest historical matches cluster around the dot-com bubble. Value factors underperformed; momentum factors outperformed. Low volatility and small size factors also weakened in both periods. Quality factors (dividend payout, net debt-to-equity, asset growth, ROE) show some divergence, but the core value–momentum–low vol pattern is shared.

Price acceleration mirrors late 1999. The NASDAQ 100 and SOX indices have risen sharply since April 2026, following a trajectory similar to their steep climb from October 1999 to March 2000. In Japan, the Nikkei/TOPIX ratio breached 16x on April 24, 2026—an all-time high—echoing the rapid rise in that ratio during the dot-com bubble.

Index vs. stock valuation divergence is extreme. As of May 10, 2026, the NASDAQ 100 and SOX trade at 24.9x and 21.5x forward P/E, respectively—levels well below the 2000 peak and not historically stretched. However, within Morgan Stanley’s AI Enablers basket, the median and mean forward P/E have climbed sharply from 2025 levels, now at multi-year highs. This indicates that crowding is concentrated among individual names, not broad indices—a condition that preceded the dot-com bust.

Key Risks

  • Concentrated unwind risk. If sentiment shifts, the high-beta, high-momentum, semiconductor-heavy portfolio will face severe drawdowns, as in the 2000 dot-com collapse.
  • Valuation premium on unproven AI earnings. Individual stock P/Es have expanded rapidly. If AI revenue and margin forecasts disappoint, the premium will compress quickly.
  • External triggers. A rise in global interest rates or a geopolitical shock could accelerate the reversal before crowded positions are unwound.

Valuation and Trade Implications

Investors cannot afford to sit out this rally if it is indeed the largest bull run since the dot-com era. However, the historical precedent suggests the final stage lasted roughly six months (October 1999–March 2000). Therefore, a targeted six-month holding window from now is prudent. Within this window, investors should: (1) maintain exposure to AI/semiconductor and high-momentum factors, (2) actively monitor individual stock P/E levels and crowding metrics (e.g., beta polarization, momentum concentration), and (3) stand ready to reduce positions if Mahalanobis distance rises sharply or if low vol/quality factors begin to outperform—both early warning signs. Avoid chasing names with forward P/E in the top decile of the AI Enablers basket; instead, favor index-level exposure or diversified baskets.

Appendix: The 13 factors used in the Mahalanobis distance calculation are: 12M Forward Earnings Yield, Book-to-Price, Dividend Yield, Composite Growth, 12×1M Momentum, EPS Up-Down Revisions, Dividend Payouts, Net Debt-to-Equity, YoY Asset Growth, Asset Turnover, Return on Equity, Low Vol, and Small Size. For the full monthly distance series from 1990 to April 2026, refer to the original report.