Investors Are Confusing AI Infrastructure Spending With Profitability

David Trainer, CEO of the financial research firm New Constructs, argues that the current market frenzy ignores a fundamental economic reality: AI infrastructure is a cost, not a competitive moat. Building out data centers and securing chips provides the tools for adoption, but it does not create the proprietary workflows or unique value propositions that drive long-term profit. According to Trainer, the true competitive edge lies in the ownership of closed-loop datasets—information that competitors cannot easily replicate or access through public models.

This shift in focus toward proprietary data echoes sentiments from industry leaders like Hims & Hers CEO Andrew Dudum and investor Chamath Palihapitiya, who suggest that value will accrue in the application layers rather than the foundational models themselves. Palantir serves as a case study for this approach, utilizing operational workflows to leverage data for specific client needs. However, even with strong demand, valuation remains a hurdle. For instance, Palantir’s current stock price implies aggressive profit growth that may be difficult to sustain over the long term. Ultimately, Trainer suggests that AI will mirror the evolution of electricity—a ubiquitous utility where the real winners are not those who build the infrastructure, but those who successfully convert it into specialized services that customers are willing to pay for.