TOKYO - Artificial intelligence stocks could regain momentum in the second half of 2027 as excessive market expectations ease and investors assess the impact of tighter U.S. data center regulations, according to Daiwa Securities strategist Yugo Tsuboi, who argues that the recent selloff represents a market adjustment rather than the beginning of a prolonged downturn in AI investment.
Tsuboi said the sharp decline in U.S. equities on November 5 was primarily driven by trading positions and risk management strategies rather than fundamental problems in the economy or financial system. The selloff reflected a combination of selling by value-at-risk-driven investment strategies, rebalancing by leveraged exchange-traded funds and concerns about market liquidity.
Technology-focused long-short funds had already been experiencing weaker returns since the summer as semiconductor stocks underwent corrections. When market volatility increased, some investors accelerated profit-taking, triggering additional selling as funds adjusted their exposure to comply with risk limits.
Leveraged ETFs may have amplified the market swings through automatic portfolio adjustments, Tsuboi said, noting that similar mechanisms have contributed to volatility in South Korean equities.
However, he emphasized that the downturn did not exhibit the characteristics of a broader financial crisis. During a genuine liquidity shock, investors often sell even traditionally safe assets to raise cash, causing simultaneous disruptions across multiple markets.
In the November selloff, interest rates remained relatively stable and there was little evidence of widespread selling of safe-haven assets, suggesting that the turbulence was concentrated in equity positioning rather than financial system instability.
Tsuboi pointed to the VIX volatility index and its futures term structure as important indicators of market stabilization. When short-term volatility expectations exceed longer-term expectations, a condition known as backwardation, markets are typically experiencing elevated stress. A return to a more normal futures structure would indicate that selling pressure is subsiding.
The correction has also reduced the unusually high expectations embedded in AI-related stock valuations.
In Japan, the Nikkei 225 had substantially outperformed the broader TOPIX index since around April, reflecting investor enthusiasm for semiconductor manufacturers, AI infrastructure providers and data center-related companies.
The gap between the two indexes has since narrowed significantly, suggesting that investors have begun reassessing the long-term earnings potential of businesses associated with AI.
Tsuboi said stock prices had previously incorporated expectations for developments several years into the future rather than simply reflecting earnings prospects for 2026 and 2027. The recent adjustment has brought valuations closer to levels consistent with more sustainable growth assumptions.
Nevertheless, a prolonged slowdown in AI investment could have consequences extending well beyond financial markets.
The International Monetary Fund examined several possible outcomes for AI development in its October 2026 World Economic Outlook, including scenarios in which investment accelerates alongside productivity improvements, expansion encounters supply constraints, or investor expectations collapse and trigger a market correction.
Under the pessimistic scenario discussed by Tsuboi, global economic output in 2027 could be approximately 1% below its baseline projection. With global growth projected at around 3.4%, such a deterioration could reduce the growth rate to roughly 2.4%, leaving the world economy expanding but at a noticeably slower pace.
Tsuboi considers such an extreme outcome unlikely, arguing that AI-related infrastructure investment is more likely to experience a moderate slowdown than a complete reversal. His outlook falls between the IMF scenarios involving supply constraints and a sharper collapse in expectations, although closer to the more moderate case.
One of the biggest uncertainties facing the industry is the growing political resistance to the rapid expansion of data centers across the United States.
Following the November 2026 midterm elections, several newly elected Democratic governors have advocated restrictions on new facilities, citing concerns over electricity consumption, infrastructure costs, environmental pressures and the distribution of economic benefits.
Tsuboi highlighted differences between proposals in Texas, Ohio and Georgia, arguing that the political response is far from uniform.
In Texas, the proposed approach would suspend approvals for new data centers until the state legislature establishes appropriate protections. Such a policy would be particularly restrictive for new developments, although facilities that have already received approval would not necessarily be affected.
Ohio has taken a more conditional approach, under which new data centers could proceed if developers meet requirements involving electricity costs, labor protections and public financial support.
These conditions include requiring operators to bear their full electricity-related costs, establish appropriate labor arrangements and repay subsidies or tax incentives if they fail to satisfy agreed commitments.
Georgia has also seen proposals for temporary restrictions on new data center construction, although Tsuboi noted that the incoming governor had previously supported data center development while serving as mayor of Atlanta, suggesting that a comprehensive prohibition may not be the intended outcome.
The composition of state legislatures could further limit the implementation of restrictive policies.
Republicans retain control of at least one legislative chamber in each of the three states discussed, according to Tsuboi's analysis, creating potential obstacles to prolonged or statewide prohibitions even where governors favor tighter controls.
Consequently, he expects regulatory pressure to produce a cooling-off period rather than bring U.S. data center investment to a complete halt.
Singapore provides a possible model for how such restrictions could evolve.
The city-state introduced a moratorium on new data center development in 2019 because of limitations involving land, electricity and other resources. However, authorities subsequently developed more selective approval procedures designed to balance environmental considerations with the country's need to remain competitive in the digital economy.
Data center development has since resumed under tighter conditions, demonstrating that restrictions can lead to more sustainable investment practices rather than permanently suppressing the industry.
Tsuboi expects the United States may follow a comparable path, establishing clearer rules governing infrastructure development while allowing the AI industry to continue expanding.
Geopolitical competition provides another reason why American AI investment is unlikely to slow dramatically.
The race between the United States and China to develop advanced AI technologies has intensified, with patent activity illustrating China's growing capabilities.
Although the United States remains a leader in important areas of AI development, Tsuboi warned that substantially reducing investment could allow China to close the technological gap or overtake American competitors.
AI infrastructure has therefore become an economic security priority as well as a commercial investment opportunity, making it difficult for the United States to abandon large-scale development.
Another factor supporting continued investor interest is the changing distribution of economic gains between labor and capital.
Tsuboi pointed to the long-term decline in labor's share of national income in the United States and other Group of Seven economies, arguing that improvements in productivity associated with cloud computing, information technology and automation have increasingly benefited companies rather than workers.
As AI adoption accelerates, the economic benefits of higher productivity may continue to flow disproportionately toward businesses and their shareholders.
For investors, this creates a dilemma. Selling all AI-related holdings may reduce exposure to short-term market volatility, but it could also mean missing the longer-term financial gains from technological advances.
Tsuboi therefore argued that maintaining some exposure to AI-related investments remains important even as valuations undergo further adjustments.
Historical market performance following U.S. midterm elections also provides grounds for cautious optimism.
Tsuboi examined previous periods in which a Republican president faced a Democratic-controlled Congress, a political configuration that he said had occurred eight times in the 21 postwar midterm election cycles included in his analysis.
In those historical cases, stock prices had risen by an average of approximately 20% over the following 160 trading days.
He suggested that divided government does not necessarily produce unfavorable stock market conditions, since political constraints can encourage more moderate policies and reduce the risk of abrupt changes.
Nevertheless, the immediate outlook for AI-related equities remains uncertain as investors await clearer indications of how U.S. data center regulations will develop.
Tsuboi expects the market to advance only gradually through the remainder of 2026 and into early 2027 while political and regulatory uncertainties continue to weigh on sentiment.
A more substantial recovery could begin during the second half of 2027 as expectations for AI investment improve and the Federal Reserve potentially adopts a more accommodative monetary policy.
The overall outlook, he said, is for a period of slower, sustainable gains followed by a stronger acceleration later in the year, with AI stocks undergoing a temporary cooling-off phase rather than entering a lasting decline.
Source: CNBC
