How To Hedge Your AI Risk

The AI boom is growing riskier: Spending has accelerated, credit is doing more of the work and the payoff remains difficult to determine.

Microsoft, Meta, Amazon, Oracle and Alphabet are committing their operating cash flow to data centers, chips and other AI infrastructure to boost supply of “compute” (ability of AI to crunch data). Some outlays are pushing free cash flow below zero. Strong balance sheets gave these hyperscalers room to borrow, but easy financing is supporting new and riskier projects with unproven business models.

The consequences reach far beyond technology. Borrowing to build compute magnifies hyperscaler spending and channels revenue to semiconductor makers, hardware suppliers, Wall Street (not Main Street) banks, industrial companies, utilities and materials producers. The effect has lifted cyclical stocks and highly leveraged businesses, leaving major indexes dependent on continued AI capital spending.

The bullish case for equity markets assumes credit remains available, competition and costs stay contained, and hyperscaler cash flow grows 15% to 20% annually through 2030. That outlook leaves little room for error. If demand develops more slowly than compute capacity, credit and equity markets would be at risk of a steep reversal.

Returns on all this investment (ROI) remain difficult to measure precisely. Much demand comes from technology firms and venture-backed startups building applications that businesses and consumers might eventually use. Corporate adoption is rising, but implementation remains uneven and often depends on software and expertise still being developed.

History offers a guide. Transformative technologies begin with an infrastructure surge, pass through a shakeout and create their greatest value through applications that reach the real economy—businesses and consumers. A downturn in the current surge could lower computing costs and clear the way for stronger applications and platforms. Market leadership may then shift from compute infrastructure toward platform and application companies that drive measurable productivity gains for the economy and achieve high ROI for businesses.

Bottom line: The risk is not that AI fails to transform the economy, but that markets have priced the transformation too narrowly. Prudent investors should consider retaining exposure to hyperscalers and infrastructure beneficiaries through broad indexes while diversifying into high-quality, dividend-paying companies less dependent on AI spending. Selective exposure to applications and platforms could provide access to the next phase of AI adoption and deployment without turning a portfolio into a one-theme wager on AI infrastructure.