Morgan Stanley Quantitative Team: AI Chip Stocks Haven’t Bottomed Yet, Recommend “Reducing Positions on Rebounds”

The Morgan Stanley quantitative strategy team warns that the recent sell-off in semiconductor and data center-related stocks is not yet over, with further downside risks in the medium term. The team recommends that investors use any rebound as an opportunity to reduce positions in these sectors.

According to Chasing Alpha Trading Desk, on October 27, the Morgan Stanley quantitative strategy team led by Ukyo Haraguchi released a report noting that in the third and fourth weeks of October, concerns over financing and credit risks for AI infrastructure investments intensified in the market, triggering a sharp sell-off in semiconductor and data center-related stocks.

Although technical rebounds are possible in the short term, the team believes the medium-term correction has not yet fully played out, advising investors to "increase the scale of position reductions."

At the same time, the report recommends increasing allocations to low-volatility stocks and financially sound value stocks, and considers AI application enterprises as medium-to-long-term structural growth opportunities. The team also identified 50 Japanese listed stocks with high correlation to the AI applications basket for investors’ reference.

This strategic framework indicates that market focus is shifting from capital expenditure and infrastructure in the upstream AI supply chain to downstream AI adoption and applications.

The Morgan Stanley quant team had already warned in an October 9 report titled “Navigating Higher Challenges in the Late Stage of the AI Trade” that in a context of rising interest rates and high valuations, investors should not take the sustainability of the AI trade for granted, and should tactically increase allocation to low-volatility stocks while retaining exposure to financially stable value stocks to withstand potential reversals in a highly concentrated AI-driven market.

The team stated that both the speed and magnitude of the recent correction exceeded their expectations; however, the direction of market rotation—outperforming low-volatility and value factors—is in line with their previous judgment, with this trend evident in both the Japanese and U.S. markets.

In terms of valuations, the report pointed out that despite a significant pullback, the median forward P/E ratio for AI and semiconductor-related stocks remains at historically high levels. This means that price corrections alone have not been sufficient to alleviate valuation pressures.

(Median projected P/E ratio for AI and semiconductor-related stocks over the next 12 months: valuations remain elevated compared to historical standards despite the pullback)

The report also compares the current trajectory of the Philadelphia Semiconductor Index to that of the internet bubble in 2000, noting similar price patterns between the two periods.

(The trajectory of the semiconductor index today closely resembles that during the internet bubble period)

Morgan Stanley Research categorizes three major global AI-related baskets: AI Enablers, AI Infrastructure, and AI Adopters.

According to the report, the first two baskets have greater exposure to semiconductors and data centers and have seen significant declines in the recent correction. The AI Adopters basket—companies integrating AI into their services and operations—tracked similarly to the other two previously but has demonstrated notable relative resilience during this sell-off.

(Performance of global AI-related asset portfolios: Though recent pullbacks occurred, AI consumer companies remain resilient)

The team believes this divergence demonstrates that market focus is shifting: from the upstream AI supply chain (capital expenditure and infrastructure) toward the downstream (AI adoption and application).

The report notes that the AI trade has so far been dominated by semiconductor and data center-related stocks, but amid widespread concerns over the prospects for these stocks, AI application companies are increasingly likely to become the next main global AI trading theme.

The report also cites the opinion of Stephen Byrd, a Morgan Stanley thematic research analyst, recommending a "barbell strategy"—on one end, allocate to companies that address infrastructure bottlenecks (such as power supply), select semiconductor and infrastructure names; on the other, allocate to early-stage software enablers and AI application enterprises.

It is worth noting that the "enablers" category also includes infrastructure and security software providers, not just semiconductor and data center companies.

To help investors identify AI application opportunities in the Japanese market, the team systematically screened the constituents of the TOPIX 1000 index.

The method is: for each stock, calculate its beta to the AI Adopters basket (controlling for TOPIX returns), then subtract its beta to the AI Enablers basket, resulting in a "net AI application beta." The top 50 stocks by this measure were selected.

(Performance of high net beta AI adopter stocks)

Historically, this 50-stock portfolio performed relatively weakly in the first half of 2026, partly due to market concerns about a potential “SaaS endgame”, but performance has improved since the second half of the year.

The team notes that while short covering may have contributed to part of the rebound, combined with the resilience of the global AI Adopters basket, this trend deserves attention as it may be an early signal of market rotation toward AI application enterprises.

The 50 stocks selected are mainly concentrated in IT services and other industry sectors, including large-cap names such as Recruit Holdings, Nintendo, NEC, Nomura Research Institute, and BAYCURRENT, as well as mid-to-small-cap tech firms such as PKSHA TECHNOLOGY, Appier Group, and SANSAN.

In summary, the current allocation recommendations from the Morgan Stanley quant team can be distilled into three main themes:

• First, for semiconductor and data center stocks related to AI capital expenditures, take advantage of rebounds to gradually reduce holdings;
• Second, continue to increase allocations to low-volatility stocks and financially sound value stocks to manage market fluctuations;
• Third, focus on AI application enterprises as a medium-to-long-term thematic allocation, seeking structural opportunities with the net AI application beta screening approach.

The team emphasizes that AI applications provide a thematic stock-picking approach different from traditional style factors, and can complement the above factor allocation strategies.

As the AI trade enters a more challenging late stage, distinguishing between AI capital expenditure and AI application opportunities will be essential for investors to control risks and capture opportunities.