The Artificial Intelligence (AI) Sell-Off Has Gone Too Far. These Are the Stocks I'd Buy Before the Market Figures It Out.

Microsoft says its total addressable market will expand across the tech stack as AI demand accelerates, with Azure revenue up 39% year over year and commercial bookings up 230% on long-term OpenAI and Anthropic commitments. Brookfield Asset Management is positioning for AI infrastructure with a $100 billion program focused on land and energy and a $20 billion Qatar JV, while the stock offers a 4.4% forward dividend yield. The piece is broadly constructive on both names, emphasizing valuation support after each stock fell about 30% from recent highs.

Analysis

The market is still treating AI as a software capex story, but the bigger second-order trade is capacity scarcity: power interconnects, land banking, and cooling water become the binding constraints long before model demand normalizes. That shifts value creation away from pure compute beneficiaries and toward the owners of scarce inputs and the capital allocators who can finance them at scale. In that setup, Microsoft’s advantage is not just product share — it is balance-sheet capacity to pre-commit demand and lock in supply, which can widen the moat if smaller competitors get stranded by delivery delays.

For Microsoft, the key risk is not demand, but conversion: heavy AI investment can inflate depreciation and keep near-term FCF optically weak even while bookings accelerate. If the market remains skeptical into the next 2-3 quarters, that is usually when multiple compression bottoms and fundamentals start to re-rate. The main reversal trigger would be evidence that AI workloads are cannibalizing legacy cloud margins faster than they expand new revenue, or that enterprise adoption remains pilot-heavy rather than production-heavy.

Brookfield’s opportunity is less about AI itself than about monetizing the bottlenecks around AI. The winner here is the toll collector on infrastructure build-out, but the hidden loser is anyone relying on cheap incremental power and easy permitting; that includes smaller data-center developers and late-arriving hyperscale builds that have to bid up land, transmission, and water rights. The contrarian point is that the current market may be underpricing the duration of this scarcity cycle — it is likely a multi-year supply mismatch, not a one-year hype trade.

The main macro tail risk is rates: higher-for-longer would hit asset values, delay project financing, and compress Brookfield’s incentive fees, making the infrastructure thesis less linear than the AI narrative suggests. But if rates stabilize, the combo of fee growth plus asset scarcity creates a favorable asymmetry. In short, the better way to express AI exposure may be to own the enablers of deployment rather than the obvious model winners.

Key Decisions for Investors

• Long MSFT on a 3-6 month horizon into post-capex skepticism; target a re-rating as AI bookings convert to revenue, with downside limited by balance-sheet strength and upside from multiple expansion if production workloads scale.
• Long BAM as a structural AI infrastructure toll collector; add on any rate-driven drawdown, since project scarcity and fee growth should compound over 12-24 months even if near-term mark-to-market pressure persists.
• Pair trade: long MSFT / short a basket of smaller cloud or software names more exposed to AI spend without the same distribution or balance-sheet advantages; thesis is that capital intensity will concentrate share in the largest platforms over the next 2-4 quarters.
• Avoid chasing pure-play data-center developers until power interconnect and financing visibility improve; the risk/reward is worse than the article implies because permitting and utility bottlenecks can defer returns by 12-24 months.
• If rates back up materially, hedge BAM with a short-duration proxy or reduce size rather than exit outright; the core thesis is secular, but financing sensitivity can create a better re-entry point.