Hyperscaler Data Center Capacity to Surge More Than 6x by 2035 as AI and Cloud Expansion Reshape Global Infrastructure

ABI Research forecasted that active hyperscaler IT load would rise from 24.37 gigawatts in 2025 to 147.13 gigawatts by 2035, based on a model of power and capacity scaling inside hyperscaler infrastructure. The projection points to expansion driven by capacity concentration rather than a proportional rise in facilities.

The forecast described a shift in how hyperscaler infrastructure grows, with power, cooling, and workload density scaling faster than physical site counts. ABI Research linked that pattern to AI build-outs, core cloud growth, storage demand, and enterprise migration workloads, while also citing energy generation and grid constraints as bottlenecks.

ABI Research reported that hyperscaler sites would increase from 3,182 in 2025 to 3,558 in 2035, with facilities above 10 megawatts accounting for nearly all meaningful growth, rising from 2,921 to 3,255 sites. Over the same period, available capacity was projected to expand from 33.96 gigawatts to 228.96 gigawatts, and under-the-roof capacity from 45.08 gigawatts to 276.63 gigawatts, based on its capacity-state modeling across active, available, and under-the-roof capacity.

ABI Research said energy access factors increasingly determine expansion as campuses scale into multi-hundred-megawatt deployments, with operators prioritizing high-capacity transmission access, long-term power agreements, and dedicated energy sources. Regionally, North America led, Asia-Pacific grew fastest, Europe consolidated into fewer high-capacity sites, and the Middle East and Africa advanced national-scale cloud campuses. “The hyperscaler market is no longer defined simply by how many facilities operators build, but by how much compute capacity they can concentrate into large, power-dense campuses,” said Leo Gergs, Principal Analyst at ABI Research.

Gergs also said, “AI is a major accelerator, but the broader story is that hyperscalers are simultaneously supporting next-generation AI clusters and the continued expansion of mainstream cloud and enterprise workloads.” ABI Research’s findings came from “Hyperscaler Data Centers Market Data Overview: 2Q 2026,” part of its “Cloud” research service, which introduced full capacity-state modeling and a redesigned workload split between AI and legacy compute to support cross-segment data center analysis.