The era of modest datacenters is over. The $3 trillion AI infrastructure boom is being defined by massive, specialized mega-projects, epitomized by the $500 billion “Stargate” venture. This joint project between OpenAI, Oracle, and SoftBank aims to build a vast network of AI-dedicated datacenters across the US and UK.
“Stargate” is not alone. Elon Musk’s xAI is building “colossus” in Memphis, Tennessee. Microsoft, meanwhile, is constructing the “world’s most powerful AI datacentre” in Wisconsin, alongside other dedicated sites in Wales and Essex. These projects represent the enormous capital expenditure required to train and run next-generation AI.
This new wave of construction is expanding global capacity at an unprecedented rate. Currently at 59GW, capacity is expected to double by 2030. This year, 7GW of capacity will be completed, and work on a further 10GW will begin—a power draw equivalent to a third of the UK’s total demand.
This incredible spending is fueled by tech’s soaring valuations, with Nvidia hitting $5tn and Microsoft $4tn. The “hyperscalers” are spending $750bn in the next two years to maintain their lead.
However, these flashy mega-projects are also the “speculative” assets that concern critics. Alibaba’s chair has warned of a “bubble” in projects raising funds without customer commitments. While Stargate, backed by OpenAI, seems secure, the “exuberance” it represents is encouraging riskier bets elsewhere in the market.