Howard Marks Flags Uncertainty in AI Investing

Oaktree co-founder Howard Marks said market optimism has prevailed since late 2022, with investors increasingly overlooking negative news, warranting greater caution. He said AI enthusiasm is clear but that uncertainty around profitability, valuations and labor-market effects prevents a firm conclusion on whether current exuberance is irrational.

Analysis

The actionable signal is not an AI short by itself; it is that broad risk assets may have become increasingly dependent on continued multiple expansion and benign macro data. In that regime, the most vulnerable exposures are long-duration equities where earnings estimates already embed several years of AI-driven operating leverage—particularly software and infrastructure names with elevated capex, weak free-cash-flow conversion, or customer-concentration risk. A modest disappointment in hyperscaler capex commentary or enterprise AI monetization can trigger a rapid de-rating even if underlying revenue remains positive.

Over the next 1-3 months, positioning risk is most likely to surface around quarterly guidance rather than headline AI demand. Watch whether MSFT, AMZN, GOOGL and META show incremental revenue or margin evidence sufficient to justify continued AI infrastructure spend; a widening gap between capex growth and disclosed AI revenue would pressure suppliers whose valuations assume an uninterrupted buildout, including NVDA, AVGO, ANET and VRT. Conversely, a broad correction would likely be led by the highest-beta AI basket rather than diversified mega-cap platforms with durable cash flows.

The contrarian point is that elevated enthusiasm does not establish a near-term top: index concentration and systematic inflows can sustain premium valuations longer than fundamentals alone imply. The cleaner expression is selective downside protection and relative-value shorts, not outright bearish beta. This caution thesis is falsified if upcoming earnings show accelerating AI revenue monetization, stable hyperscaler FCF despite capex, and upward—not merely maintained—forward EPS revisions across the semiconductor and infrastructure complex.

Key Decisions for Investors

• Do not initiate a broad S&P 500 short on sentiment alone. Instead, add 3-6 month Nasdaq-100 downside hedges via QQQ put spreads, financed selectively with upside call overwrites; target approximately 2:1 payoff if QQQ declines 8-12%, and reassess after the next hyperscaler earnings cycle.
• Establish a relative-value basket: long MSFT and META versus short a diversified high-multiple AI-infrastructure basket (for example, equal-weight NVDA, ANET and VRT) over 1-3 months. The thesis is that monetization and balance-sheet resilience matter more than capex exposure if AI expectations reset; exit if infrastructure names receive material upward EPS revisions alongside improved FCF guidance.
• Reduce exposure to unprofitable or low-visibility application-software names marketed on AI optionality until customer adoption converts into disclosed recurring revenue. Set an earnings-season watch item for AI revenue contribution, net-retention trends and sales-cycle length; absent those data, this is a risk-control action rather than a short recommendation.
• Monitor the ratio of semiconductor/infrastructure forward EPS revisions to hyperscaler capex guidance. A sustained deterioration in revisions while capex remains elevated is a high-conviction catalyst for the relative-value short; accelerating revisions and expanding disclosed AI revenue would invalidate it.