Brookfield Sees $7 Trillion AI Infrastructure Buildout Over Next Decade

Brookfield estimates the AI infrastructure buildout could require roughly $7 trillion over the next decade, driven by demand for power generation, compute infrastructure, and connectivity. The firm says funding will need to come from public and private capital markets rather than traditional bank financing alone, while it continues targeting contracted cash flows in infrastructure, real estate, and asset-backed credit. The comments reinforce growing investor focus on electricity availability, data center capacity, and related energy infrastructure.

Analysis

The most important implication is not the headline capex number, but that AI shifts infrastructure from a cyclical buildout to a quasi-utility financing regime. That favors owners of scarce bottlenecks — power interconnects, transmission, fiber routes, and entitled land — over pure compute vendors, because the binding constraint becomes grid access rather than GPU availability. In practice, this should compress returns for undifferentiated data-center landlords while expanding the value of vertically integrated platforms that can source power, finance assets, and lock in long-duration contracts.

A second-order winner is the capital stack itself. If bank balance sheets are insufficient, the funding mix shifts toward insurers, private credit, project finance, and hybrid real-asset structures, which is constructive for managers with permanent capital and liability-matched funding. That also means spreads on AI-adjacent infrastructure debt may stay tighter than fundamentals warrant, because investors will pay up for duration and contracted cash flows; the trade is in underwriting quality, not just thematic exposure.

The more underappreciated risk is localization: the buildout will likely concentrate in a handful of power-rich regions, creating regional price spikes for electricity, land, and labor while leaving many announced projects stranded in permitting. That argues for a months-to-years lens rather than trading the headline, because near-term enthusiasm can outrun interconnection timelines by 12-24 months. If power prices normalize, policy slows, or hyperscaler capex discipline tightens, the market will punish the more speculative parts of the theme first.

Contrarianly, the consensus may still be underestimating how much of the value accrues to non-sexy assets like gas peakers, transmission, cooling, and fiber rather than to the obvious AI winners. The better risk/reward is to own the bottleneck enablers and selectively fade the parts of the market pricing in flawless occupancy and perpetual demand growth. The theme is real, but the economics will be captured unevenly and with long lags.

Key Decisions for Investors

• Long BN on a 6-12 month horizon: best direct public-market proxy for scaled, contracted infrastructure capital. Use pullbacks to build; upside comes from monetizing scarce power-backed assets and recycling capital into higher-return opportunities. Risk: if AI capex shifts to self-funded hyperscaler balance sheets, fee-growth could lag.
• Long BE versus utilities-heavy peers for 3-9 months: the market may still underprice distributed power and on-site generation as the fastest path to AI load-shedding. Structure via call spreads to limit downside; catalyst is utility interconnect bottlenecks and grid delays. Risk: execution or policy friction on customer adoption.
• Pair trade: long BN / short a listed office-reit or non-core data-center landlord basket over 6-12 months. The thesis is that scarcity premium accrues to assets tied to power access and contracted cash flows, while commodity-like real estate faces margin compression. Use the pair to isolate the AI-infrastructure premium from broader rate sensitivity.
• Long selected utility or transmission exposure on 12-24 month timeframe, preferably names with visible capex pipelines and regulatory recovery. The setup is asymmetric because capex spend can re-rate equity via rate base growth even before load fully materializes. Risk: political pushback on rate increases or slower-than-expected interconnect approvals.
• Avoid chasing pure-play AI compute beneficiaries on the headline alone; if expressing bullish AI infrastructure, prefer the bottleneck chain over the compute narrative. If you need convexity, use call spreads rather than outright equity to protect against a near-term unwind in AI enthusiasm.