AI exuberance meets caution: is the trade turning into a bubble?
After months of record gains, Wall Street's tone has shifted from enthusiasm to concern. Senior voices are flagging stretched valuations, tight credit spreads, and risk-taking that looks late-cycle.
As one bank CEO put it, elevated assets have "further to fall." That doesn't rule out more upside first - it just raises the cost of being wrong.
What the smart money is signaling
Sentiment is nearing extremes. Bank of America's Global Fund Manager Survey named an "AI equity bubble" as the top tail risk for the first time. Cash balances dropped to 3.8%, brushing BofA's 3.7% "sell" threshold. Readings below 4% have historically marked peak risk appetite late in the cycle.
Positioning is crowded. State Street's Risk Appetite Index shows "Big Money" adding risk for five consecutive months. As DataTrek noted, absent a major shock, they're unlikely to pivot quickly.
Correlations are unusually low. Cross-sector correlations have fallen to their lowest level since the bull market began - a pattern that tends to show up when confidence runs too high and often precedes short-term pullbacks.
Corporate spend is accelerating
Companies are matching investor conviction with massive capex. Google announced a $15B data center hub in India. AMD shares jumped on a new chip partnership with Oracle. Walmart expanded an AI collaboration with OpenAI.
OpenAI has been locking in infrastructure deals across Broadcom, AMD, and Nvidia to diversify supply - a self-reinforcing investment loop that can amplify cycle risk. Michael O'Rourke of JonesTrading said, "I absolutely believe we are [in an AI bubble]," citing recent megadeals and OpenAI's roughly $1.5T build-out plans versus $13B in revenue and a lack of profitability as a point of disconnect.
The counterpoint: conviction, not mania
Others argue this is pricing to perfection, not mania. Optimism is high, FOMO exists, but balance sheets at leading tech firms are strong. The story is being priced ahead of execution, which raises the bar but doesn't equal euphoria.
Either way, earnings are the fulcrum. Consensus expects double-digit revenue and EPS growth through 2026 for Nvidia, Microsoft, and Alphabet - well ahead of the S&P 500. With high expectations, there's less room for upside surprises and more penalty for misses.
Practical playbook for finance teams
- Re-underwrite AI exposure: Map holdings by revenue sensitivity to AI capex and consumption (cloud spend, inference workloads, GPU supply). Tie positions to dated catalysts: earnings, product releases, and regulatory milestones.
- Right-size risk: Scale position sizes to implied volatility and expected drawdowns. Define hedge overlays (index collars, calendar spreads into earnings). Pair longs in cash-generative leaders with shorts/underweights in unprofitable AI proxies.
- Watch liquidity and credit: Track credit spreads and high-grade ETFs. If spreads widen while AI equities rise, treat it as a caution signal.
- Monitor correlation regime: Use breadth and cross-sector correlation data. Persistently low correlations after a strong run often precede pullbacks - reduce leverage and tighten stops.
- Maintain optionality: If fund cash is trending below 4%, consider raising marginal cash to regain flexibility without abandoning core exposure.
- Stress-test earnings: Model 5-10% top- and bottom-line downside for NVDA/MSFT/GOOGL and second-tier suppliers. Re-rate multiples under each scenario and size positions to survive a guidance cut.
- Track the capex cycle: Build a heat map of data center build-outs, power availability, and lead times. Watch GPU pricing, alternative accelerators, and delivery schedules for signs of supply normalization or demand fatigue.
- Assess concentration risk: Identify customers with outsized dependence on a small number of AI end-markets or vendors. Diversify where practical.
- Policy risk scan: Keep a live view on antitrust actions, AI safety rules, export controls, and data privacy. These can shift cost structures and timelines fast.
Signals to track over the next two quarters
- Sentiment/positioning: BofA fund manager cash below 4% and State Street risk appetite staying elevated.
- Market internals: Sector correlations, advance-decline breadth, and dispersion - especially around earnings windows.
- Infra bottlenecks: Data center power constraints, interconnect lead times, and utility commitments that can cap delivery schedules.
- Supply dynamics: GPU pricing trends, alternative chip adoption, and inventory signals across OEMs and cloud providers.
- Consumption vs. capex: Whether cloud/inference consumption growth keeps pace with capex intensity.
- EPS revision breadth: Up vs. down revisions across AI leaders and enablers; watch the second derivative.
Bottom line
The AI theme still carries real earnings power, but starting conditions matter. With sentiment hot and expectations high, the margin for error is thin. Let price, positioning, and cash flows guide exposure - add on weakness with proof of execution, hedge into crowded catalysts, and avoid chasing heat without a plan.
Further resources
Your membership also unlocks:
