Data-Centre Energy Use
Also known as: data center energy, data centre electricity consumption, cloud energy use, server energy consumption
Data centres are the physical facilities housing the servers, networking equipment, and cooling systems that power cloud computing, internet services, AI workloads, and digital communications.
They are significant energy consumers: the global data centre sector consumed an estimated 415 TWh of electricity in 2024, about 1.5% of global electricity demand according to the International Energy Agency, which expects that figure to roughly double by 2030. Energy efficiency, renewable energy sourcing, and cooling technology are central to reducing data centres' environmental impact.
Data-Centre Energy Use, explained at three levels
At home
When you use a website, stream a video, or send a message, a computer in a data centre somewhere is processing that request. Data centres are basically enormous buildings full of servers running 24 hours a day, seven days a week. They use a lot of electricity: both to power the servers themselves and to keep them cool (computers generate heat). The biggest data centres use as much electricity as a small city.
At work
Data centre energy consumption matters to businesses both as an operational cost and as a sustainability consideration. Cloud computing costs directly reflect data centre energy: compute, storage, and network pricing are all ultimately driven partly by energy costs. Organizations with sustainability commitments need to understand their cloud computing's carbon intensity and choose providers accordingly. Cloud providers now publish carbon footprints and offer renewable energy matching for enterprise customers.
At home
Data centre energy breaks down into three main uses: IT load (the servers, storage, and networking equipment doing actual computing), cooling (air conditioning, chillers, and cooling systems to remove heat from servers: typically the second largest energy use), and power conditioning (UPS systems, transformers, and distribution equipment). Modern hyperscale data centres achieve very high energy efficiency (PUE of 1.1-1.2) compared to traditional enterprise data centres (PUE of 1.5-2.0).
At work
Water use is an increasingly important dimension of data centre environmental impact alongside energy use. Air cooling is energy-intensive; evaporative cooling (which uses water to cool air before circulating it) is more energy-efficient but consumes significant water. A large data centre can consume millions of litres of water per day for cooling. In water-stressed regions, this creates tension between data centre energy efficiency and water sustainability: some hyperscale operators have committed to being 'water positive' (returning more water to local watersheds than they consume) by 2030.
At home
AI workloads are driving a step change in data centre energy demand. Training large AI models and running inference for popular AI services require high-density GPU computing, which is significantly more energy-intensive per server rack than traditional workloads. This has driven construction of specialized AI data centres with higher power density, more sophisticated cooling (direct liquid cooling for GPU clusters), and larger uninterruptible power supplies. A widely cited 2024 Goldman Sachs estimate put a ChatGPT query at roughly 10 times the electricity of a Google search, although figures published by OpenAI and Google in 2025 suggest a typical text prompt now uses about 0.3 watt-hours.
At work
From a practitioner standpoint, the most actionable levers for reducing cloud computing energy impact are: choosing regions with low-carbon electricity (cloud providers publish carbon intensity data by region), enabling carbon-aware scheduling (shifting flexible workloads to times when renewable energy is available), improving workload efficiency (fewer, more efficient server-side computations per user action), and right-sizing compute resources (not paying for and running more compute than the workload needs). These actions save both energy and money: energy efficiency improvements in data centres directly reduce cloud costs.
Real-world examples
At home
Carlos chose to store his photos in a cloud service that publishes detailed sustainability reports showing their data centres run on 100% matched renewable energy. He appreciated knowing his cloud storage was not contributing to fossil fuel electricity demand, even though he had no visibility into exactly which physical data centres his photos resided on.
At work
The cloud architecture team at a financial services firm used their cloud provider's carbon intensity dashboard to identify that their batch processing workload (which ran nightly) could be shifted to a lower-carbon availability region without performance impact. Moving the batch jobs reduced their estimated cloud-related Scope 2 emissions by 22% with no increase in cost.
Common confusions
- Not all data centres are the same: hyperscale and enterprise data centres differ dramatically in efficiency. A large hyperscale facility (Google, Amazon, Microsoft) typically operates at PUE 1.1-1.2, meaning nearly all the energy goes to computing. A traditional enterprise data centre may have PUE 1.5-2.0, meaning 50-100% extra energy is consumed in overhead. When assessing the environmental impact of cloud vs. on-premises hosting, compare the actual PUE values of the specific facilities, not generic averages.
- Data centre renewable energy claims require careful interpretation. 'Powered by 100% renewable energy' typically means the provider has purchased renewable energy certificates (RECs) matching their consumption: it does not necessarily mean the data centre is physically connected to a renewable source and never draws fossil fuel electricity from the grid. Hourly matching (where renewable generation is matched to consumption hour by hour in the same grid region) provides a stronger claim than annual matching through RECs.
- Data centres becoming more efficient does not mean their total energy use is falling. Individual data centres have become dramatically more efficient per unit of compute over the past decade, but the total number of compute operations performed has grown much faster. Total data centre electricity consumption globally has increased despite efficiency improvements, and AI workloads are accelerating this growth.
For analysts
In BA artefacts, data centre energy requirements surface in sustainability reporting systems, green IT policies, and cloud architecture governance. When documenting cloud infrastructure requirements, include: 'Workloads processing non-real-time batch data must preferentially run in regions with carbon intensity below [X] gCO2e/kWh as reported by the provider's carbon intensity API, with documentation of actual region selection and estimated carbon impact included in the quarterly sustainability report.'
History & origin
Data centres emerged from computer rooms of the 1960s and grew rapidly with the commercialization of the internet in the 1990s. Energy consumption became a major concern in the mid-2000s when data centres were consuming 1.5% of US electricity and growing rapidly. Hyperscale data centres (built by Amazon, Google, Microsoft, and Meta) from the 2010s brought dramatically improved energy efficiency through optimized cooling, server utilisation, and hardware design. The AI boom from 2022 drove a new wave of data centre construction, particularly for GPU-intensive AI training and inference workloads, reversing a period of plateauing energy growth.