Noah, Brad DeLong, and Dan Wang discuss the Anthropic IPO

Brad DeLong and Dan Wang joined me on Hexapodia, my Substack Live podcast, to discuss the Anthropic IPO. This was my debut episode of the new series, which will feature recorded conversations that I also send out as emails. Our first discussion was a bit disorganized, as I arrived late due to a time error. Despite that, I was pleased to hear Dan and Brad share their perspectives.

I began by outlining my "bull case" for Anthropic, which is not based on the idea of Anthropic dominating every industry or replacing human workers. Rather, I envision a change in the capital structure of the efficiently producing economy. In this scenario, to succeed in any industry, a company must partner with a frontier AI company that offers access to an internal proprietary model that is months ahead of any model derived from it.

This partnership would require tying up a significant portion of a company's capital structure, effectively making it a subsidiary of the AI company.

The idea of AI acting like oil, necessary for success in various industries, is not what I envision. Instead, I propose a world where companies must become subsidiaries of AI companies like Anthropic, OpenAI, or Google to thrive. Anthropic is currently the leading player in this oligopolistic ecosystem, but I expect other conglomerates to join the competition by creating their own subsidiaries within every industry.

This shift could lead to larger, concentrated AI-heavy firms, as human workers and AI become increasingly complementary.

Interestingly, a paper released by Google Deepmind researchers Imas, Koh, and Legg (2026) supports some of these dynamics. In their model, "When will advances in AI lead to economic transformation?" they explore how AI intelligence impacts organizations, considering that AI expertise is reproducible while human skills are in fixed supply.

As AI becomes more capable, organizations with human employees can still outperform those relying solely on AI. However, as AI coordination improves, it may break the traditional "Coasean ceiling" on firm size, leading to a market structure transformation towards large, AI-heavy conglomerates. I believe their model captures some of the key dynamics in our discussion.

Financial disclosure: I do not currently hold any Anthropic stock, but I anticipate purchasing it when the IPO occurs. Additionally, I own TypeSafe stock, which competes with Anthropic.