The global rush to build AI infrastructure, a $3 trillion endeavor, is running into a massive physical barrier: the power grid. The sheer number of datacenters being built to power AI tools is creating an unprecedented demand for electricity, which may require a separate infrastructure boom just to keep the lights on.
The scale of the expansion is staggering. The world’s current datacenter capacity is 59 gigawatts (GW). According to Goldman Sachs, this is expected to double by the end of 2030. This year alone, work is expected to start on 10GW of new capacity—an amount equal to roughly one-third of the UK’s entire power demand.
This rapid, power-hungry expansion carries a secondary infrastructure cost. Goldman Sachs estimates that $720 billion in grid spending is needed to meet this new energy demand. This cost is separate from the $3 trillion being spent on the datacenters themselves, highlighting the massive, cascading infrastructure needs of the AI revolution.
Tech giants are building at a furious pace. Microsoft is constructing the world’s most powerful AI datacenter in Wisconsin. Elon Musk’s xAI has its “colossus” project, and the $500bn Stargate venture is planning a network of massive AI-dedicated sites.
While some question if this is a bubble, the physical constraints are undeniable. Datacenters are already massive power consumers, and AI-specific workloads are even more intensive. The $3 trillion bet on AI is therefore also a $720 billion bet that our global energy grids can scale fast enough to support it.
AI’s Power Problem: $3Tn Datacenter Boom Requires $720Bn Grid Overhaul
