Jason Thomas, Head of Global Research & Investment Strategy at Carlyle, argues that investors are applying outdated valuation frameworks to AI companies, failing to account for their dramatic shift towards capital intensity. Previously "asset-light" businesses like Amazon, Alphabet, Microsoft, Meta, and Oracle have significantly increased their physical assets (PP&E) by 50% to 200% since the end of 2023 due to aggressive "compute" capacity acquisitions. This change has led to a decline of 600 to 1300 basis points in their cash return on equity.
Thomas points out that before the Global Financial Crisis (GFC), a company's price-to-book ratio was a strong predictor of future returns. However, the post-GFC era saw the rise of "asset-light" businesses rich in intangible assets where market value no longer correlated with physical assets. These companies, benefiting from network effects and zero marginal costs, were valued highly despite low book values. Now, with AI necessitating substantial capital expenditures, these valuations are being questioned. Investors still value these companies as if nothing has changed, even though their business models have fundamentally shifted.
Carlyle's analysis suggests that if the new physical assets acquired by these AI-focused companies were valued at cost, and a 10x price-to-book multiple applied to the rest of the business, their market capitalizations would be approximately half of what they are today. This highlights a potential overvaluation, as the market currently asks shareholders to effectively pay $1 billion for every $100 million in data center assets acquired. Thomas concludes that while being bullish on AI technology is reasonable, skepticism about the valuations of companies at the forefront of this shift is also warranted, especially as returns on physical assets typically do not exhibit the same scale-free growth as software and digital platforms.
