Most AI Investors Are Looking in the Wrong Place

Most AI Investors Are Looking in the Wrong Place

Summary

~1.0 min read
AI's growth is constrained by power, not just chips. While investors focus on GPU demand, the speaker argues that AI is becoming a physical infrastructure buildout story, with massive electricity requirements that the system cannot add overnight. This creates a significant gap between AI demand and actual deployment. The best chips are useless if power cannot reach data centers through a complex chain from the grid to the rack. Many investors miss this crucial physical bottleneck. Companies solving these power delivery and support challenges, especially those close to deployment, may be undervalued. The speaker recommends focusing on specific bottlenecks rather than broad themes. Examples include Vertiv (VRT) for power and thermal support at the rack level, and Eaton for electrical infrastructure. Bloom Energy is mentioned as a higher-risk, optional extension for on-site power. Risks include overvaluation, other potential bottlenecks (cooling, networking), and lazy investing. The key is disciplined focus on companies directly solving the physical power constraint for AI deployment.