RESEARCH
AI Data Centers Consumed 415 TWh in 2024, Surpassing 1.5% of Global Electricity
Global data centers consumed roughly 415 TWh of electricity in 2024 — about 1.5% of total world electricity demand — with AI infrastructure identified as the primary accelerant. Independent modeling from the IEA, Gartner, LBNL, and Goldman Sachs confirms the scale but diverges on exact figures, reflecting methodological differences in how AI workloads are isolated from legacy infrastructure. The Delta Engine surfaces this spread: sources agree on the order of magnitude and growth trajectory, but the precise consumption estimate varies by up to 25% depending on scope and assumptions.
Findings
- The IEA estimates global data center electricity consumption at approximately 415 TWh in 2024 — roughly the combined annual electricity use of France and Germany combined.
- Gartner independent modeling implies a 2024 baseline closer to 387 TWh, rising 16% to 448 TWh in 2025 — a 7% divergence from the IEA figure and evidence that AI load isolation methodology drives material differences in estimates.
- Goldman Sachs frames data center demand in GW capacity terms, implying consumption above 500 TWh under broader scope assumptions — the largest single-source estimate reviewed.
- LBNL/DOE measured US data centers at 176 TWh in 2023 and projects 325–580 TWh nationally by 2028, suggesting US demand alone may approach or exceed current global IEA estimates within four years.
Delta Engine result
↔ Divergence Detected — Δ 0.4516 (threshold 0.05)
- IEA T5 95
- LBNL/DOE T5 65
- Pew Research T0 75
- Gartner T0 70
- Goldman Sachs T0 60
Evidence quality
avg 53 · min 25 · max 95 · spread 70