The $3 trillion AI datacenter boom has a massive, hidden price tag: a $720 billion bill for grid spending, according to Goldman Sachs. The “incredible” expansion of datacenters to power AI is putting an unprecedented strain on the world’s energy infrastructure.
The AI build-out is “rapid.” Global datacenter capacity (59GW) is set to double by 2030. This year alone, 7GW of new capacity will be completed, and work will start on 10GW more. This 10GW alone represents a power draw “roughly a third of the UK’s power demand.”
This energy demand is the physical-world consequence of the digital boom. While “hyperscalers” like Google and Microsoft spend $750bn on the datacenters themselves, and mega-projects like “Stargate” are planned, none of it works without power.
The $720bn grid cost is a “further infrastructure cost” that is often overlooked in the “boom or bubble” debate. Even if some of the “speculative” datacenters “will never be built,” the “healthy” expansion by Big Tech is already creating massive energy needs.
This “power problem” is a fundamental constraint on the AI revolution. The industry is not just betting $3 trillion on AI; it’s also betting $720 billion that our global energy grids can scale fast enough to support it.