AI Drives Global Data Center Power Surge: 17% Growth in 2025
Multiple major institutions independently documented a 16–17% surge in global data center electricity consumption in 2025, reaching between 448 and 485 TWh. The IEA reported approximately 17% growth — more than 5× faster than overall global electricity demand — while Gartner independently measured 16% growth to 448 TWh. AI-optimized server workloads were the primary driver, growing at 3–5× the rate of conventional servers. The remaining gap in absolute figures (~37 TWh) reflects differing methodological baselines, not underlying disagreement about the trajectory.
claim: Global data center electricity consumption grew 16–17% in 2025, reaching 448–485 TWh, with AI-optimized server workloads driving the fastest growth — outpacing overall global electricity demand growth by more than 5×.
Sources · prominence score
Evidence Quality
Tier Mix
Pipeline Warnings
- Unknown source host — defaulted to T? (lowest credibility)CredibilityScorer · gartner.com
- Unknown source host — defaulted to T? (lowest credibility)CredibilityScorer · goldmansachs.com
- Unknown source host — defaulted to T? (lowest credibility)CredibilityScorer · pewresearch.org
- Unknown source host — defaulted to T? (lowest credibility)CredibilityScorer · wri.org
- insufficient_candidatesAlgox:topK · 5/6
- ephemeral_signing_keyResearchProtocolAdapter · UVRN_EXPANSE_PRODUCER_PRIVATE_KEY not set — signed with a one-time ephemeral key
Findings
- IEA reported global data center electricity grew 17% in 2025, reaching ~485 TWh — more than 5× the rate of overall global electricity demand growth (3%).
- Gartner independently reported 16% growth for 2025, estimating 448 TWh — a near-identical trajectory with a ~37 TWh methodological gap reflecting different scope baselines.
- AI-optimized servers consumed an estimated 93 TWh in 2025 (21% of total data center power, per Gartner) and are forecast to triple their share to 44% by 2030.
- Goldman Sachs forecast US data center power demand growing from 31 GW in 2025 to 66 GW by 2027 — a doubling in two years — as hyperscaler capex topped $400 billion annually.
Utility planners, grid operators, and energy consultants building 10-year capacity plans need to know whether AI infrastructure demand forecasts from different institutions align. UVRN surfaces how much agreement — or divergence — exists across major institutional sources before those forecasts inform capital decisions.
- Run ID
- run-104
- Agent
- [email protected]