The $3 trillion AI spending spree is not just a digital phenomenon; it’s a massive physical one. This year, work is expected to start on an estimated 10 gigawatts (GW) of new datacenter capacity—a power draw “representing roughly a third of the UK’s power demand.”
This explosive growth is adding to an existing global capacity of 59GW, which Goldman Sachs expects to double by 2030. This pace is fueled by “hyperscalers” (Amazon, Meta, Google, Microsoft) spending $750bn in two years, and mega-projects like the $500bn “Stargate” venture.
This physical footprint comes with an enormous secondary cost. Goldman Sachs estimates $720 billion will be needed in grid spending just to meet the new energy demand. This highlights the immense strain the AI boom is placing on global infrastructure.
While the market celebrates Nvidia’s $5tn valuation, some warn this physical build-out is “speculative.” The Uptime Institute cautions many announced datacenters “will never be built.” Alibaba’s chair, too, sees a “bubble” in projects raising funds without customer commitments.
Whether boom or bubble, the AI revolution’s demand for power is real. The $3tn investment in datacenters is inseparable from the $720bn cost to power them, creating one of the largest infrastructure challenges of the 21st century.
1