Bending the Energy Curve: Decoupling Digitalization Trends from Data Center Energy Growth

Overview

The paper examines why global data center electricity, flat during 2010-2020, is forecast to grow strongly toward 2030. Its analysis rests on the ratio of ICT energy to annual data generated. That ratio points to a compute-efficiency disruption from about 2020, with recovery expected later in the decade. The problem is therefore not only rising demand: energy per unit of data also worsened for a period.

The scenario section compares two levers. Reaching PUE 1.2 for new builds and 1.3 for 10% of existing sites by 2030 cuts data center energy by about 3.6% (about 38 TWh). Improvement in the ICT performance factor cuts energy by about 14.4% (about 153 TWh), roughly four times the PUE lever. The conclusion is that facility efficiency is useful but secondary.

ICT energy per unit of data is about 0.92 TWh/ZB in 2020 and is projected at 0.45 and 0.34 TWh/ZB by 2030 depending on the scenario. The paper also discusses grid-relief measures such as flexible load. These are forecasts, not measurements: results depend on assumed scenarios, so the relative size of the two levers is a safer reading than any single figure.

Key points

1. Global data center electricity was flat in 2010-2020 and is forecast to grow strongly toward 2030.
2. PUE 1.2 for new builds plus 1.3 for 10% of existing sites by 2030 cuts energy about 3.6% (about 38 TWh).
3. ICT performance factor improvement cuts energy about 14.4% (about 153 TWh), roughly 4x the PUE lever.
4. ICT energy per data generated was about 0.92 TWh/ZB in 2020.
5. Projected 0.45 and 0.34 TWh/ZB by 2030 depending on scenario.
6. A compute-efficiency disruption from about 2020 is followed by expected recovery later in the decade.
7. Grid-relief measures such as flexible load are discussed.

Using it in design

• In the energy dashboard, show IT-side levers (utilization, accelerated computing) next to facility PUE.
• Add context text telling users that PUE improvement can be a small share of total energy impact.
• Present flexible-load models as a separate scenario with grid effect.

Who should read it

Decision makers and sustainability teams doing energy strategy and capacity planning.

Related standards

• ISO/IEC 30134-2
• ISO/IEC 21836 (SEEM)

Publication details

Authors

Jim Simonelli; Victor Avelar; Wendy Torell

Publisher

Schneider Electric

Year

2025

Document number

WP 212

Document language

English

Source type

Manufacturer

Access

Free to read

Licence

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

Citation