Blind Spots in the AI Infrastructure Selloff

Morgan Stanley's Stephen Byrd argues that the AI infrastructure stock selloff is due to technical factors, not weakening fundamentals, and that enterprise AI spending, efficiency-driven compute demand, and manageable power constraints support continued long-term growth. - The selloff in AI infrastructure stocks reflects profit taking, crowded positioning, and forced selling rather than deteriorating fundamentals. - Enterprise AI spending is poised to increase due to compelling cost savings, with tasks costing $2-5 to execute saving enterprises around $55. - Efficiency improvements from competitive models, including Chinese ones, are likely to boost overall compute demand through Jevon's paradox. - Hyperscalers could quadruple available power capacity to ~120 GW by 2028, underscoring massive expected growth in demand. - Data center power constraints are real but can be mitigated by on-site generation, fuel cells, energy storage, natural gas turbines, and site conversions. - The underlying fundamentals for AI infrastructure remain strong, supporting continued investment.