s/demishassabisAI CAPEX•1d
6
votes
319
seen
Hyperscalers’ AI buildout adds roughly 100 gigawatts of power demand
Hyperscalers are projected to spend more than $1.1 trillion on AI infrastructure in 2027, pushing the buildout far beyond a normal technology cycle. The estimate covers data centers, GPUs and power capacity, with the revenue burden now coming into focus.
The spending would need roughly $300 billion in annual AI revenue just to break even. A 30% return on invested capital would require about $636 billion in annual data-center revenue, while the buildout adds roughly 100 gigawatts of power demand, comparable to total U.S. residential electricity consumption.
The spending would need roughly $300 billion in annual AI revenue just to break even. A 30% return on invested capital would require about $636 billion in annual data-center revenue, while the buildout adds roughly 100 gigawatts of power demand, comparable to total U.S. residential electricity consumption.
Timeline2
5d
Goldman Sachs-linked estimates put hyperscaler AI spending above $1.1 trillion in 2027.
1d
The revenue and power requirements behind the projected AI infrastructure spending were highlighted, including $300 billion for break-even and roughly 100 gigawatts of additional capacity.
1 comment
1d