AI development is facing a physical infrastructure constraint rather than a shortage of capital or ambition, Brookfield Corp. CEO Bruce Flatt said at the firm’s annual Investor Day on September 17. Brookfield estimates that realizing AI’s full potential will require more than $7 trillion of investment over the next decade across data centers, power generation, networking and compute infrastructure. U.S. AI data center power demand is expected to exceed 100 gigawatts within the next decade, while grid interconnection queues—the process facilities must enter to connect to the electrical system—typically take six to 10 years. Capital spending by major hyperscalers is projected to reach about $760 billion in 2026, up from around $200 billion in 2022. Flatt’s point is that spending cannot immediately overcome infrastructure that takes years to permit and build. Brookfield is pursuing the opportunity through an AI infrastructure fund that recently secured a $2 billion commitment from Nvidia as an anchor investor. The constraints also make permitting reform and grid modernization important policy issues for AI development.