source: techcrunch ai: data centers expected to use 4x more electricity by 2035
level: business
data centers are projected to consume one-fifth of all electricity generated in the us by 2035, a fourfold increase from today, according to a new bloombergnef report. the surge is fueled by ai compute demands, pushing total data center capacity to nearly 200 gigawatts over the next decade. nearly half of that capacity will be dedicated to ai training and inference, with the us hosting 64% of ai chips by power demand by 2033.
the forecast is 83% higher than bloombergnef's december estimate, reflecting rapid data center development. other groups have also raised projections: epri more than doubled its 2024 estimate, and s&p's forecast rose by over a third between october and april. most new data centers will connect to already strained grids, with pjm interconnection seeing 34% of its electricity go to data centers and ercot in texas 22%. pjm has struggled with connection requests, leading to a four-year pause on new source applications and a 76% electricity price increase over the past year.
globally, if ai adoption continues aggressively, data centers could create 1,935 terawatt-hours of new electricity demand by 2033, nearly matching india's annual usage. despite grid congestion, data centers still seek connection to pjm, representing 38% of charges in its latest capacity auction. the report highlights the growing tension between ai infrastructure expansion and electrical grid capacity.
why it matters: ai and data science professionals must consider the escalating energy demands of ai infrastructure, which could impact operational costs and sustainability goals.
source: techcrunch ai: data centers expected to use 4x more electricity by 2035