AI Needs More Than Chips | Presented by CME Group

Summary

• The primary constraint on the AI boom is insufficient electrical power, not advancements in algorithms.
• Global data center electricity consumption is projected to more than double by 2026.
• A single AI-focused data facility can demand power equivalent to a small city.
• Data centers could consume 6% to 12% of U.S. electricity by 2030.
• Morgan Stanley forecasts a 49-gigawatt shortfall in available power access by 2028.
• Solutions to power constraints include chip efficiency gains, on-site power generation, and a nuclear revival driven by small modular reactors.
• Hyperscalers are expected to commit over $1 trillion in 2025 and 2026 alone to build out energy infrastructure.
• Short-term power constraints may slow AI growth but are predicted to ignite a concurrent energy revolution.
• The narrative asserts the AI boom will adapt through infrastructure investment rather than bust.
• Key market implication: Massive capital expenditure into energy infrastructure, particularly in nuclear and distributed generation, creating investment opportunities.
• Important uncertainty: The pace of infrastructure scaling and potential near-term slowdowns in AI deployment due to power limitations.
• Narrow niche observation: Small modular reactors represent a specific technological revival within the broader energy solution set.