The Hidden Risk Inside the AI IPO Boom

With SpaceX freshly listed and OpenAI and Anthropic both expected to follow at valuations above $1 trillion, investors are facing a genuinely difficult question that doesn’t have a clean answer: how much of the AI industry’s reported revenue reflects real, durable demand from end users, and how much reflects companies within the same tightly interconnected ecosystem buying from, investing in, and propping up one another.

The concern isn’t hypothetical. Several of the largest AI infrastructure deals reported this year involve chip makers, cloud providers, and AI labs simultaneously acting as suppliers, customers, and investors in each other’s businesses. That circularity can make headline growth numbers look stronger than the underlying economic reality, because a dollar of revenue booked by one company might simply be a dollar of capital that another company in the same circle just raised and redeployed.

Untangling that is genuinely hard even for professional analysts with access to detailed financial disclosures — let alone individual investors reading quarterly earnings summaries. That difficulty is exactly why concentrated bets on any single part of the AI stack, however compelling the story, deserve real caution. A company can be a legitimate long-term winner and still see its stock become significantly overvalued in the short term if the market hasn’t yet distinguished real demand from circular financing.

It’s a good example of why our own strategy caps any single high-conviction theme — AI, crypto, or otherwise — at a modest, deliberately limited slice of a portfolio, rather than letting excitement about a compelling narrative determine position size. Themes like this one can absolutely be part of a well-constructed plan. They shouldn’t be allowed to define it.