David Rosenberg On Why The AI Boom Is Riskier Than You Think

David Rosenberg argues the current market rally is largely a single, highly correlated “AI trade,” with spillovers beyond tech into industrials, utilities, and financials. He links AI-driven corporate spending to knock-on effects for interest rates and the broader economy, but the piece provides no specific earnings or policy numbers. Overall, this is a market-structure narrative that may influence positioning rather than deliver concrete catalysts.

Analysis

The market is still pricing AI as a single factor, but the cash-flow winners are likely to remain narrow: semiconductor content, networking, power/thermal infrastructure, and a small set of data-center operators with contractual pricing power. The second-order effect is a transfer from software and other long-duration assets into businesses that sell the picks-and-shovels; most “AI exposure” labels will not convert into near-term EPS, and that gap is where multiple compression shows up first.

The more important macro risk is that the buildout is not rate-neutral. Heavy AI capex raises electricity demand, grid investment, and working-capital needs, which can keep input costs and bond yields sticky even if headline inflation cools elsewhere. That creates a two-way tension: beneficiaries like semis can keep earning, but the broader equity basket—especially utilities, REITs, and unprofitable software—can de-rate if the market starts treating AI spending as a higher-for-longer rates problem rather than a pure growth story.

Over the next 1-3 months, the key catalyst is earnings guidance, not product launches: if management teams keep lifting capex while revenue monetization stays back-half weighted, the trade becomes more crowded and more fragile. Over 6-18 months, the thesis is falsified if enterprise AI spending pauses, GPU lead times normalize faster than expected, or rate volatility rises enough to force portfolio de-risking. The contrarian take is that the market may be overbought on “AI breadth” and underweight the financing cost of that breadth; the real dispersion is likely between tangible infrastructure cash flows and narrative-only beneficiaries.

Key Decisions for Investors

• Overweight SMH vs. short IGV on a 3-6 month horizon: semis and networking are capturing the first wave of monetization, while app-layer software still faces valuation risk if AI adoption stays more capex-heavy than revenue-heavy. Falsify if software guides meaningfully re-accelerate or semis miss demand by one quarter.
• Long VST / CEG / ANET basket on dips for 6-12 months: these are among the few beneficiaries where AI demand can translate into durable pricing power and visible order flow. Risk is regulatory pressure on power prices or a faster-than-expected slowdown in data-center expansion.
• Short XLU or a regulated-utility-heavy basket against merchant power / grid names if rates back up over the next 1-3 months: higher electricity demand can be good for selective producers, but not for rate-sensitive yield proxies. Cover if 10Y yields break lower and utilities re-rate on defensive flows.
• Use put spreads on high-multiple, AI-adjacent software after earnings pops (1-3 months): upside is capped by the gap between narrative and FCF conversion, while downside can accelerate if guidance implies longer payback periods. Falsify if bookings and net retention inflect materially higher.
• Watch for a rotation trigger rather than forcing a broad market long: if AI capex commentary slows or yields rise, reduce exposure to equal-weight growth and small caps first, because they are most sensitive to multiple compression from a crowded single-factor trade.