Oaktree co-founder Howard Marks recently issued a significant warning that investors may be underestimating the sheer scale and unpredictability of the AI revolution.
His core argument is a clear distinction between owning and lending in the age of AI. For lenders, the best-case scenario is getting their money back with interest—a fixed, linear return. However, they bear the downside risk if an AI-driven company is disrupted. In contrast, equity owners participate in the convexity of AI; their potential returns are non-linear and could be immense if the technology fundamentally reshapes an industry. The risk/reward profile, Marks argues, heavily favors owning a piece of the upside.
This isn't just theory; it's backed by a historic wave of investment. First, look at the capital expenditure (capex). In early 2026, Alphabet and Amazon alone signaled a combined $375–$385 billion in capex for the year, largely to build out their AI infrastructure. This level of spending by hyperscalers is a massive bet on future returns that are far from guaranteed, creating a scenario where the payoff, if it comes, will likely be captured by shareholders, not bondholders.
Second, the real-world impact on labor is accelerating faster than many expected. Marks pointed to Block's February 2026 decision to lay off nearly half its workforce, explicitly citing AI-driven efficiencies. The market's reaction was telling: Block's stock surged over 20%. This event serves as a stark example of AI not just creating value, but also causing significant disruption—a dynamic that introduces a level of volatility and risk that fixed-income investors are not typically compensated for.
Adding to this complexity is a shifting regulatory landscape, with new rules like the EU AI Act and U.S. chip export controls creating both opportunities and sudden risks. This environment, where supply, demand, and policy are all in flux, reinforces Marks' central thesis: linear extrapolation is a dangerous game in the world of AI. The debate is no longer about a potential bubble, but about correctly positioning for a technology whose impact is asymmetric, favoring those who own the future over those who simply finance it.
- Convexity: A term describing a non-linear relationship. In finance, it means an investment's value can increase at an accelerating rate, offering disproportionately large gains compared to the risk taken.
- Hyperscaler: A large-scale cloud computing provider, such as Amazon Web Services (AWS), Google Cloud, or Microsoft Azure, that can provide massive computing resources on demand.
- Capex: Short for capital expenditure, which are funds used by a company to acquire, upgrade, and maintain physical assets like property, buildings, or equipment.