The Illusion of Predictability: Why AI Investing is a Thumbsock

In this conversation, Oaktree Capital co-founder Howard Marks outlines the systemic risks within the current AI-driven market. His core argument is that we are seeing a period of irrational exuberance where traditional valuation metrics struggle because the future of AI is impossible to predict. The result is that investors are increasingly treating speculative lottery tickets as safe, high-growth assets. For the reader, the advantage lies in recognizing that while the potential upside of AI is high, current market pricing assumes a level of certainty that does not exist. Navigating this requires moving away from forecasting, which Marks views as a mistake, and toward building a portfolio based on a clear assessment of risk, uncertainty, and the inevitable return to market reality.

The Hidden Cost of This Time is Different

Marks compares the current AI frenzy to historical bubbles like the 1920s radio boom, the 1960s computer surge, and the 2000 internet bubble. The pattern remains the same: new, transformative technology attracts massive capital, leads to over-investment in infrastructure, and forces investors to pay prices that assume perpetual growth.

The way people get into trouble is by not being alert to the possibility. And this all seems incredibly relevant today on the day when SpaceX is set to go public... My favorite fortune cookie says that the cautious only mark right-grade poetry. So, you know, investing in these companies today could be a huge error. But it could be great poetry.

-- Howard Marks

The danger is not just the potential for loss, but the shift in how we define value. When profitability is ignored in favor of top-line revenue growth, the market becomes fragile. The idea that revenue growth justifies any price fails when projected forward, as it ignores the reality that competition among AI hyperscalers will eventually compress margins.

The Disruption of the Moat

Thirty years ago, value investors looked for moats, or defensive barriers like local newspaper monopolies that made businesses unassailable. Marks notes that AI has disrupted this concept. When software can write its own code, the traditional moat of proprietary software systems disappears.

This creates a feedback loop: as the world becomes less predictable, companies that once seemed safe are now vulnerable to AI-driven disruption. Investors must stop looking for companies that will never change and instead accept that the predictability of the world has permanently decreased.

I have never met anybody who thinks they can tell me what this world is going to look like five or 10 years from now. And so why should... the young person... who is laying the foundation for his investment portfolio... conclude that he is probably right when all these other people are?

-- Howard Marks

The Spectrum of Risk: Lottery Tickets vs. Hyperscalers

Marks argues that investors should calibrate their activities across a spectrum rather than seeking a single correct forecast. At one end, hyperscalers like Amazon, Google, and Microsoft offer established cash flows and moats, providing a lower-risk way to play AI. At the other end, startups and unprofitable AI firms represent lottery tickets, which are high-risk bets where the probability of total loss is high, but the potential for massive gain exists.

The systemic risk arises when investors mistake these lottery tickets for stable, high-growth assets. Marks warns that in the current environment, defined by 17 years of generally favorable conditions, many managers have succeeded through aggressiveness rather than skill. The real test will come when the market forces a sorting of those who were truly skilled versus those who were simply lucky.

Key Action Items

• Calibrate your position on the risk spectrum: Do not treat speculative AI startups as equivalent to established hyperscalers. (Immediate)
• Adopt a two-part forecast: When making an investment thesis, define both your forecast and your level of confidence in that forecast. If you are highly confident in an unspecifiable future, you are likely making a mistake. (Immediate)
• Prioritize intellectually prosaic sectors: For long-term stability, look toward industries less likely to be disrupted by AI, such as energy, food, timber, and home building, where physical reality limits the impact of intellectual automation. (Next 12 to 18 months)
• Accept the dentist reality: If you require a high success rate, do not pursue investing. Accept that even the most successful investors rely on a small number of great outcomes rather than consistent perfection. (Ongoing)
• Review your liquidity constraints: If you are invested in non-traded products like private credit, ensure you understand the terms of withdrawal. Disillusionment often follows when investors realize they cannot exit these positions during market stress. (Over the next quarter)
• Prepare for the tide to go out: Assume that the current favorable investment environment will not last indefinitely. Stress-test your portfolio against a scenario where capital is no longer cheap and abundant. (12 to 18 months)