Wall Street has been said to follow a certain set of themes or plays that cycle every few years; for example, the environmental plays that dominated the 2010s with renewable energy and ESG mandates captured trillions in capital, which were then dethroned by artificial intelligence from 2022 onwards, raising Nvidia as the first $5 trillion company and driving unimaginable capital expenditure from hyperscalers. However, signs of another shift are now underway with SoftBank's complete exit from its $5.8 billion Nvidia stake and Michael Burry's aggressive short positions against strong AI stocks: Palantir and Nvidia. It might represent more than tactical repositioning, such as declarations that the AI valuation has entered bubble territory. Bank of America's October 2025 Global Fund Manager Survey reveals that 54% of institutional investors now believe AI stocks are overvalued, the highest scepticism level since the survey began tracking AI sentiment. Burry's specific thesis centres on accounting manipulation, where he argues that hyperscalers are understating depreciation by extending AI chip lifespans, likely overstating earnings by over 20%. What's particularly telling is that SoftBank isn't abandoning AI entirely but rotating from AI infrastructure (chips) to AI applications (OpenAI, with up to $40 billion committed), suggesting smart money believes the value capture is shifting from semiconductor manufacturers to companies that monetise AI directly. This mirrors the dot-com era when investors rotated from infrastructure plays like Cisco to application layer winners, or, to simply put it, from selling shovels to gold itself.
The developing investment themes arising from institutional capital flows point toward three distinct sectors positioned to absorb the rotation from pure-play AI stocks: energy infrastructure (particularly nuclear and grid modernisation), defence spending, and the picks-and-shovels supporting AI rather than AI itself. Nuclear power is seeing another renaissance, with the Nuclear Energy Institute forecasting $2.2 trillion in investment and representing 17% of global electricity generation by 2050 since data centre demands are prioritised more than the negative sentiment it gets. Defence spending has become irresistible to allocators, with the fiscal 2026 National Defence Authorisation Act calling for $924.7 billion and the SPADE Defence Index rising 90% since the Russia-Ukraine conflict began. Furthermore, the AI infrastructure bottleneck, particularly power grid capacity, has become the binding constraint, with utilities forecasting $1.1 trillion in capital investments through 2029 and data centres requiring 22% more grid-based power by year-end 2025. Goldman Sachs Research projects global power demand from data centres will rise 165% by 2030, creating a multi-trillion-dollar opportunity in transmission infrastructure, natural gas, and nuclear that doesn't depend on AI delivering its promised productivity gains. Besides energy plays, quantum computing is actively positioning itself as the "next AI", with stocks like Quantum Computing Inc. already delivering 2,275% returns in 2025, though this sector remains speculative compared to the tangible capital requirements of energy infrastructure.
To reiterate, Wall Street’s next dominant investment theme points to a capital rotation already in motion, as institutional investors grow wary of inflated AI valuations and look for more durable, fundamentally grounded sectors. Moves by major players such as SoftBank and Michael Burry signal the early stages of this shift, suggesting a broad repositioning away from overextended technology trades. Among emerging themes, energy infrastructure stands out as the most probable long-term leader through 2027–2030, supported by three reinforcing conditions: persistent supply-demand imbalances that sustain profitability regardless of AI’s trajectory, bipartisan political backing that ensures policy continuity, and scalable economics that amplify returns as capital inflows expand due to project size and natural monopoly characteristics. Defence spending offers a similar structural appeal but faces mean-reversion risks if geopolitical tensions subside, while quantum computing represents a smaller, high-upside frontier segment. The most effective portfolio approach is a barbell allocation—concentrating roughly 60–70% of rotation capital in energy infrastructure as the high-probability core theme, maintaining 10–15% exposure to quantum computing for asymmetric upside, and selectively holding AI infrastructure enablers such as data centre REITs and utilities that benefit regardless of market leadership shifts. Institutional behaviour suggests these rotations tend to persist longer than fundamentals alone justify due to benchmark pressures and herd dynamics, but as prominent exits gain visibility, the sector capable of absorbing $3–5 trillion in new capital while offering inflation protection and political durability is energy infrastructure, making it the most likely anchor of Wall Street’s next multi-year cycle.
Sources: Reuters, CNBC, Business Insider, NEI,
Photos: Unsplash
