The Scale Of The Infrastructure Build
The numbers describing the current AI infrastructure build are large enough that they can lose meaning through repetition, but the investment implications are concrete. AI-related capital expenditures across major hyperscalers exceeded $360 billion in 2025, reflecting multiyear commitments to data centers, computing infrastructure, and model development. These investments are being funded through a combination of strong operating cash flows and long-duration debt issuance — a pattern that signals confidence in the return profile of the spending cycle rather than speculative excess. The commitment horizon is measured in years, not quarters.
BlackRock Investment Institute's 2026 midyear outlook identifies the AI buildout as an accelerating theme and notes that whichever model architecture ultimately wins the competitive race for AI superiority, the physical infrastructure requirements remain constant: power, memory, chips, and data centers are scarce inputs that will shape investment opportunities regardless of how the software and model layer evolves. This framing is useful because it shifts attention away from picking winners among AI software companies toward identifying the constrained physical resources that all AI applications require.
Power And Energy As The Binding Constraint
Energy demand is emerging as the most significant near-term constraint on the AI buildout, and investors who focus exclusively on semiconductors and software miss a critical part of the investment landscape. Data centers consume enormous amounts of electricity, and the expansion of AI computing capacity is creating demand growth that existing grid infrastructure was not designed to accommodate. Morgan Stanley Research has identified the Future of Energy as one of its four key 2026 investment themes specifically because of this intersection between AI adoption and electricity demand.
Natural gas has attracted particular attention as a bridging fuel for data center power generation, with its relative reliability, scalability, and cost profile making it more practical than intermittent renewable sources for baseload data center requirements. Utilities with significant exposure to data center customers are seeing a re-rating of their growth prospects. Grid infrastructure companies — transformers, transmission equipment, grid management software — are benefiting from a secular demand cycle that extends well beyond the AI-specific buildout. For investors looking to access the AI theme without direct semiconductor exposure, energy and power infrastructure represent a complementary and arguably less crowded expression of the same secular trend.
Positioning Without Overpaying For AI Exposure
The challenge for investors entering the AI theme now is valuation. Technology and AI-related equities have been leading the market, and optimism about the capital expenditure cycle is already reflected in the prices of the most obvious beneficiaries. J.P. Morgan's research team has flagged AI skepticism as a key risk, noting that with valuations running hot and significant hype in the market, episodes of volatility or temporary retracement cannot be ruled out even if the fundamental CapEx trajectory continues.
The practical response for investors who want AI exposure without paying peak multiples for the most crowded names is to look at the infrastructure layer — power, cooling, connectivity, and real estate in the form of data center REITs — which tends to trade at more modest valuations than pure semiconductor or software plays while still capturing meaningful exposure to the same spending cycle. S&P 500 consensus earnings growth of 24% for 2026 reflects AI's broad economic contribution, but investors need to stay selective about the price paid for that growth.