Howard Marks’ 2025 Memo: Assessing AI Market Bubble Risks

The report integrates Howard Marks’ (Oaktree Capital) memo content [0] to assess the AI market’s bubble risk. Marks frames bubbles as events driven by irrational exuberance, distinguishing between two types: “inflection bubbles” (e.g., railroads, internet, AI) that accelerate technological progress despite short-term investor losses, and “mean-reversion bubbles” (e.g., subprime mortgages) that only destroy value [0]. The current AI market, characterized by AI accounting for the majority of U.S. corporate capex, S&P 500 gains, and GDP growth (with Nvidia briefly reaching a $5T valuation [0]), exhibits parallels to past inflection bubbles, such as speculative startup valuations (e.g., Thinking Machines Lab’s $50B pre-product valuation [0]) and investor FOMO. However, key differences exist: leading AI firms like Nvidia and Microsoft have reasonable P/E ratios compared to 1999 internet stocks, indicating more grounded fundamentals [0]. The memo also identifies specific risks, including aggressive debt financing in AI infrastructure, circular deals (e.g., OpenAI/Nvidia transactions recalling 1990s telecom accounting practices [0]), and rapid technological obsolescence of AI assets [0].

• Bubble Classification Matters
  : Marks’ distinction between inflection and mean-reversion bubbles challenges the binary “bubble/no bubble” narrative, suggesting AI could drive long-term progress even if a speculative correction occurs [0].
• Balanced Fundamentals vs. Speculation
  : While established AI leaders show solid fundamentals, startup excesses (e.g., $2B seed rounds for product-less firms [0]) indicate pockets of irrationality in the market.
• Societal Impacts as Market Drivers
  : AI’s labor-saving potential and the associated risk of mass job loss, alongside challenges with universal basic income (UBI) solutions [0], could become catalysts for policy changes that impact AI investment.

• Risks
  :
  • Financial Instability
    : Aggressive debt and special purpose vehicle (SPV) use in AI infrastructure could lead to market corrections if lending standards tighten [0].
  • Speculative Losses
    : High valuations of pre-revenue AI startups may result in significant investor losses in a downturn [0].
  • Societal Unrest
    : Job displacement from AI could fuel regulatory interventions that restrict AI development or increase operational costs [0].
• Opportunities
  :
  • Long-Term Technological Progress
    : Historical inflection bubbles (railroads, internet) demonstrate AI’s potential to drive decades of economic growth despite short-term volatility [0].
  • Investment Discipline
    : Marks’ memo may encourage selective investing in AI firms with proven revenues and sustainable models [0].

Howard Marks, co-founder of $180B asset manager Oaktree Capital [0], released a memo in December 2025 addressing AI bubble concerns, drawing on parallels with historical bubbles and distinguishing between bubble types. The AI market currently dominates U.S. market gains, with Nvidia as the world’s first $5T company [0]. While the memo identifies risks like speculative valuations and aggressive debt, it acknowledges AI’s real demand and potential for long-term progress. Key gaps include the absence of a quantitative bubble-burst probability, limited discussion of non-U.S. AI markets, and specific investment strategies beyond “moderate positioning” [0].