2026 Global Data Center Market Outlook

Overview

JLL's outlook of 5 January 2026 forecasts about 100 GW of new global data centre capacity in 2026-2030, roughly doubling capacity at a CAGR of about 14%. According to the record's figures, global capacity could reach about 200 GW by 2030, with a rise of 97 GW over 2025-2030. AI and cloud are identified as the drivers of this growth.

By region, the Americas account for about half of global capacity with a 17% supply CAGR, and about 90% of that capacity is in the US. The report says AI was about a quarter of workloads in 2025, could approach half by 2030, and that inference could overtake training around 2027. The regional split shows that capacity is geographically concentrated and that design assumptions may vary by region.

Power and cost topics bear directly on designers. The average wait for a grid connection in primary markets exceeds four years, which is pushing more behind-the-meter power and batteries. Average global construction cost rose from $7.7M to $10.7M per MW over 2020-2025 (7% CAGR), with a 2026 forecast of $11.3M per MW for shell and core. Tenant technology fit-out can reach $25M per MW for AI infrastructure. The document is a freely accessible industry outlook, and only the record's summary and key figures were used here. As a consultancy's market forecast, its figures are estimates rather than certainties and may vary by region.

Key points

1. Global capacity could reach about 200 GW by 2030; +97 GW over 2025-2030, roughly 14% CAGR.
2. The Americas hold about 50% of global capacity with a 17% supply CAGR; the US is about 90% of Americas capacity.
3. The average wait for grid connection in primary markets exceeds four years.
4. Average global construction cost rose from $7.7M to $10.7M per MW over 2020-2025 (7% CAGR); the 2026 forecast is $11.3M per MW (shell and core).
5. Tenant technology fit-out can reach $25M per MW for AI infrastructure.
6. AI was about a quarter of workloads in 2025, could be half by 2030, with inference overtaking training around 2027.

Using it in design

• In the cost module, use $/MW shell-and-core benchmarks (with year) as a sanity range for capex estimates, and show tenant fit-out separately.
• In the site module, warn that grid-connection lead times can exceed four years in primary markets.
• Offer on-site generation and BESS scenarios to bridge grid waits.
• Use the growth in AI share as context for density and cooling assumptions.

Who should read it

Data centre investors, developers and design teams making site, power and cost decisions.

Publication details

Authors

JLL Research

Publisher

JLL

Year

2026

Document language

English

Source type

Expert / community

Access

Free to read

Licence

Copyright retained; citation only — the publisher’s terms apply to the original document

Citation