Kevin O’Leary estimates half of planned AI data centers in the US will never be built due to grid capacity limits, while Berkeley Lab finds over 70% of interconnection requests withdrawn. Taiwan Semiconductor Manufacturing Co. shares fell 4.25% as the shortage of new AI facilities threatens chip demand growth.
The US electrical grid was designed for one to two percent annual demand growth and lacks capacity for AI data center loads. Industry forecasts estimate $5.2 trillion of AI data center capex through 2030, yet over 70% of grid interconnection requests are withdrawn and half of planned projects are at risk of cancellation.
Microsoft signed a 20-year deal to restart the Three Mile Island nuclear plant, Amazon paid $650 million to co-locate with the Susquehanna nuclear station, Google backed small modular reactors, and Meta seeks up to 4 gigawatts of new nuclear capacity. These commitments total billions and reflect corporate strategies to bypass grid constraints.
With half of AI data centers unlikely to be built, demand growth for GPUs and advanced semiconductors may slow. Taiwan Semiconductor Manufacturing Co. could face reduced orders as hyperscalers delay or cancel facility expansions, pressuring manufacturing capacity utilization and revenue growth.
