Is AI momentum starting to cool? BofA weighs in

Bank of America warned that the AI-driven momentum trade may be losing steam as end users push back on pricing and shift toward cheaper open-source alternatives. The bank said U.S. hyperscalers have underperformed the S&P 500 by almost 15% since January, signaling rising investor unease about returns. It also flagged semiconductors, capital goods, and mining as stretched, while favoring defensives such as staples if AI enthusiasm fades.

Analysis

The key market shift is not that AI demand is slowing, but that pricing power is breaking before utilization does. That matters because the first-order losers are not just the obvious “AI winners”; the second-order loser is the entire capex-financed ecosystem that assumed throughput growth would automatically convert into durable margin expansion. When end users can arbitrage to cheaper models, the valuation multiple on the whole AI stack should compress even if unit demand remains healthy, because revenue quality deteriorates faster than headline growth.

That creates a bifurcation inside technology: hardware and infrastructure names tied to the buildout can still print strong quarters, but the market will start to punish any hint of falling returns on incremental capital. The bigger risk is that hyperscaler underperformance becomes a signal for broader factor rotation out of momentum, which would spill into banks and other crowded high-beta leadership names. In that scenario, the unwind is less about AI itself and more about position de-grossing across the most consensus-long parts of the market.

The most interesting contrarian angle is that the selloff may initially overstate fundamental damage to the fastest-growing AI beneficiaries. If pricing competition forces model vendors to bundle more services or subsidize usage, the next leg of value capture likely shifts to compute-efficient picks-and-shovels and application-layer names with lower customer acquisition friction. But that transition is messy: over the next 1-3 months, multiple compression can arrive well before earnings estimates are revised, creating a window where expensive enablers underperform even if the long-term AI thesis remains intact.

Key Decisions for Investors

• Short SMCI on rallies over the next 2-4 weeks; use it as a high-beta proxy for AI capex sentiment. Risk/reward: favorable for a 10-15% downside move if the market starts discounting lower infrastructure ROI, but cover quickly if hyperscaler order commentary re-accelerates.
• Reduce BAC exposure tactically for 1-2 months or hedge with a bank ETF put spread. The crowding risk is the point: if momentum unwinds, financials can underperform despite decent fundamentals because they sit in the high-momentum basket.
• Pair trade: long APP / short SMCI for a 1-3 month horizon. APP has more resilient monetization optionality and less direct dependence on capex cycles, while SMCI is more exposed to any AI infrastructure multiple reset.
• Buy AAPL downside protection into the next 1-2 quarters via put spreads rather than outright shorting. The risk is not immediate demand collapse, but margin pressure from having to defend ecosystem pricing while AI monetization remains uncertain.
• Overweight defensive staples versus a broad market basket for the next 4-8 weeks if AI leadership continues to roll over. This is the cleanest hedge against a momentum reversal, with lower earnings sensitivity and better relative performance in de-grossing episodes.