Brookfield Corp. Chief Executive Bruce Flatt says the artificial intelligence race is already decelerating, as developers are unable to build enough infrastructure to match the pace demanded by AI firms. In recent remarks, leaders of major AI companies, including Anthropic PBC CEO Dario Amodei and OpenAI CEO Sam Altman, suggested that AI labs must slow their technical advancement to address unpredictable risks and prevent advanced models from slipping beyond human control.
Flatt noted that the sector had already hit a wall prior to those statements due to hard limits on the speed at which data centers and other infrastructure can be delivered. "As an industry, we are not building nearly enough. We cannot even build a fraction of what everyone thinks they need," Flatt said at Brookfield's annual investor day on Thursday. "So when they say they plan to slow down, the reality is it would slow down anyway, because there simply is not enough computing power to deliver the scale everyone believes is necessary."
Brookfield estimates that AI infrastructure will require more than $7 trillion in capital investment over the next decade. Flatt described the slowdown as a positive development, saying it will "bring more discipline to the entire system." His assessment is backed by solid data. According to company guidance compiled by Statista, the combined capital expenditure of Meta, Microsoft, Alphabet, and Amazon is projected to reach roughly $760 billion in 2026, nearly quadrupling from about $200 billion in 2022. Among them, Amazon plans to spend around $220 billion, Alphabet up to $205 billion, Microsoft about $190 billion, and Meta approximately $145 billion.
Yet these massive investments are colliding with hard constraints on infrastructure construction. Flatt's point about "not being able to build" is not about a lack of funds but rather physical supply bottlenecks. Power shortages currently pose the most significant hurdle. In its 2026 data center forecast report, Uptime Institute pointed out that electricity will be the "defining constraint" on data center growth in 2026 and beyond, as AI-driven load growth intensifies pressure on already strained power grids, and the long lead times for large-scale power deployment will pose a substantial barrier.
Comments