The AI revolution is driving a global datacenter capacity race. The world’s 11,000 datacenters, which grew 500% in 20 years, are just the beginning. A $3 trillion investment wave is now funding an even faster expansion, with Goldman Sachs projecting that current global capacity (59GW) will double by 2030.
The pace is “incredible.” This year alone, 7GW of new capacity will be completed, and work will start on 10GW more. This is fueling massive spending by “hyperscalers” like Google and Microsoft ($750bn in two years) and creating mega-projects like the $500bn “Stargate” venture.
This race is also creating “bubble” fears. Alibaba’s chair has warned of “excess” in the market, with “speculative” projects raising funds without customers. The Uptime Institute agrees, cautioning that “many” of the announced datacenters “will never be built” or will only be “populated partially.”
A $1.5tn “funding gap” for this expansion is being filled by “private credit,” which analysts warn could pose a “structural risk to the global economy” if these speculative projects fail.
This capacity race is also putting a massive strain on energy grids, with Goldman Sachs estimating a $720bn grid spending bill to meet the new demand. The world is racing to build the infrastructure for AI, but the speed and “debt-fueled exuberance” of the race have many worried about a crash.