AI’s Next Stress Test

Summary

Morgan Stanley's Stephen Byrd and Ariana Salvatore assess the AI investment cycle's durability amid token maxing fears and political pushback on data centers. Byrd argues enterprise AI use cases remain highly profitable, compute demand will outstrip supply, and Jevons paradox will drive capex higher, supporting AI capex beneficiaries. He also sees Bitcoin miners benefiting from data center grid constraints. Salvatore expects federal support for AI to persist, with local opposition leading to conditional, not blocked, data center builds.

• Enterprise AI use cases deliver ~$55 of benefit for a few dollars of token cost, making token maxing concerns overblown.
• Both frontier and open source AI models will continue to grow, with high-end models retaining value for complex tasks.
• Stephen Byrd sees Jevons paradox playing out: cheaper models will expand adoption and drive AI capex higher.
• Ariana Salvatore says a US federal data center ban is unlikely because of strategic competition with China.
• Local political opposition is growing and will force data centers to adopt conditional buildouts and off-grid solutions.
• Bitcoin miners with existing grid access are experiencing strong demand for data center capacity.
• Off-grid data centers will increasingly rely on natural gas turbines, fuel cells, and large-scale energy storage.