How much energy and water do AI data centres use, and why are communities pushing back?
Data centres are energy-hungry buildings, and studies suggest smart design and pairing them with wastewater plants could cut their water and carbon footprint, while US politics is turning against them.
Covers: Estimates of electricity and water use by data centres, especially those running AI workloads, the drivers of growth, and the main community concerns such as local grid strain, water competition, noise and land use. It does not cover individual company investment decisions or step-by-step advocacy tactics.
Also answers: How much water do AI data centres use? · Why are people protesting data centres? · AI data centre energy and water consumption · Data centre water usage and community backlash
- One page for this question6 other ways of asking lead here
- 5 independent sourcesEvery claim links to what supports it
- 4 connected pages4 changed this week
- Clean discussionScreened before anything appears
The short answer
Interpretation AI-prepared starting mapData centres are buildings that house the servers, storage and cooling equipment behind digital services and AI training. The evidence here is uneven: one detailed study of a single water-cooled facility in Sweden found that over 99.9% of its energy use went to data-processing processes rather than support systems, a share the authors call much higher than…12345
Read the full explanation and caveats
At a glance
The picture in numbers
Live · updated just now
99.9%
100 in every 100
- Freshwater saved1,300 million m³
- CO₂-equivalent cut84 million m³
- Cost savings95.4 million m³
- Carbon-aware scheduling32%
- Renewable-backed, lower-PUE shift79%
The evidence behind it
5 sources- Other studies and data3
- Background2
Published in 2026
| Source | Kind | Year |
|---|---|---|
| How backlash against data centres could shape the US midterm elections | Background | 2026 |
| Global data-water symbiosis reduces AI infrastructure's carbon and water footprint. | Other studies and data | 2026 |
| Pursuing a hierarchical taxonomy in a water-cooled data center room. | Other studies and data | 2026 |
| Quantifying and Disclosing the Environmental Footprint of AI in Research: Life Cycle-Informed Framework and Open-Access Calculator Development Study. | Other studies and data | 2026 |
| Data center (Wikipedia) | Background | Unknown |
The community around it
No one has added to this page yet. Firsthand experience, a newer study or a different reading of the numbers would show up here, credited to you.
Test what you remember
Fill in the blank: ?% of one Swedish facility's energy use went to data processing, not support systems
Drag the slider to fill in the blank
Did this answer your question?
Be the first to voteIn brief
In one closely measured Swedish facility, over 99.9% of energy use went to data-processing processes rather than support systems, a share the authors call unusually high for a production sector.2
Evidence-backedPairing data centres with wastewater treatment plants is modelled to save about 1,300 million m³ of freshwater and 84 million tonnes of CO₂-equivalent a year, with the largest potential in the US, Japan, China, the Netherlands and the UK.3
Evidence-backedDesign choices matter: carbon-aware scheduling cut mid-sized cluster emissions by 32%, and a shift to renewable-backed, lower-PUE infrastructure cut operational emissions by about 79%.4
Evidence-backedOpposition to data-centre development in the US now spans both major parties and is seen as potentially shaping the midterms.5
Evidence-backed
Your turn
Have your say
See where others stand. Join free to add your perspective. One answer per account.
How do you feel about this?
No votes yetQuick questions from connected pages
What it means for you
Which fits you?
Pick the situation closest to yours. Each answer says what it rests on.
If you want to compare the energy performance of similar data centres
process-level indicators of the kind developed for the Swedish water-cooled room can be used to benchmark facilities with comparable infrastructure and monitor where energy goes.2
Evidence-backedIf you are siting or planning a data centre near a wastewater treatment plant
the pairing model suggests using treated effluent for cooling and recovering return heat for sludge drying and anaerobic digestion, with the largest modelled potential in the US, Japan, China, the Netherlands and the UK.3
Evidence-backedIf you run a mid-sized compute cluster and want to cut emissions
carbon-aware scheduling is modelled to reduce emissions by about 32%, from roughly 665 to 450 kg CO₂e per month.4
Evidence-backedIf you are choosing infrastructure for a large-scale AI deployment
renewable-backed, lower-PUE infrastructure is modelled to cut operational emissions by about 79%, and the water–carbon trade-off becomes more important to report.4
Evidence-backedIf you are reporting or assessing an AI project's footprint
disclosing energy use, emissions, water use, hardware assumptions and regional context makes the figures reproducible and comparable.4
Evidence-backedIf you are following US electoral politics around data centres
opposition is coming from both Democrats and Republicans, so a project's local politics may not split along party lines.5
Evidence-backedThe full explanation
Interpretation AI-prepared starting mapData centres are buildings that house the servers, storage and cooling equipment behind digital services and AI training. The evidence here is uneven: one detailed study of a single water-cooled facility in Sweden found that over 99.9% of its energy use went to data-processing processes rather than support systems, a share the authors call much higher than in other production sectors. A modelling study estimates that pairing data centres with wastewater treatment plants worldwide could save about 1,300 million m³ of freshwater a year, cut roughly 84 million tonnes of CO₂-equivalent annually and save about US$95.4 billion a year, with the biggest potential in the US, Japan, China, the Netherlands and the UK. A separate framework reports that carbon-aware scheduling cut emissions in a mid-sized cluster by 32% (about 665 to 450 kg CO₂e/month) and that moving to renewable-backed, lower-PUE infrastructure cut operational emissions by about 79%. On politics, the BBC reports that candidates from both major US parties are opposing data-centre development amid a growing backlash ahead of the midterms.12345
- Evidence 15
- Interpretation 7
The full story · 3 chapters
01
How much energy data centres use
AI summary:A Swedish data-centre study found over 99.9% of energy went to data processing, but one facility's result isn't a sector average.
Evidence-backed: A data centre is a physical facility for storing, managing and disseminating data, including the computer systems and associated components that house IT infrastructure and train artificial intelligence. US law defines it as any facility that primarily contains electronic equipment used to process, store and transmit digital information, including a facility within a larger structure that uses environmental control equipment to keep that equipment running.1
Evidence-backed: A time-resolved study of a water-cooled data-centre room in Sweden, operated by Ericsson with a detached in-rack cooling system, broke energy use into six processes. Four relate to data processing and two to support functions. Over 99.9% of total energy use was connected to data-processing processes, which the authors describe as significantly higher than in other production sectors. The study's value is in benchmarking: it offers indicators that let similar facilities be compared and their energy use monitored process by process.2
Interpretation: The single-facility result says little about the sector as a whole. Energy intensity depends heavily on cooling design, climate, workload and how efficiently the facility is run, so a figure from one Swedish room should not be read as a typical or average number.2
02
Water use, cooling and emissions
AI summary:Modelled pairing with wastewater plants could save water and emissions, and design choices like scheduling and renewables cut footprints further.
Evidence-backed: Cooling is where water enters the picture. A global analysis found a strong spatial overlap between data centres and wastewater treatment plants, which makes national-scale pairing possible: treated effluent is used for data-centre cooling, and the heat returned is recovered to support sludge drying and anaerobic digestion. The authors estimate this would cut greenhouse gas emissions by about 84 million tonnes of CO₂-equivalent a year, conserve about 1,300 million m³ of freshwater a year and deliver net annual cost savings of about US$95.4 billion. They identify the US, Japan, China, the Netherlands and the UK as having the greatest mitigation and water-saving potential.3
Interpretation: These are modelled savings from a proposed infrastructure arrangement, not measurements of current consumption. They describe what could be saved if data centres and wastewater plants were paired at scale, which is a different claim from how much water data centres use today.3
Evidence-backed: A life-cycle framework for reporting AI's environmental footprint, with an open-access calculator, tested three scenarios. In a small laboratory setting, optimisation cut emissions from about 0.07 kg CO₂e to 0.05 kg CO₂e and improved the proposed grade from C to B. In a mid-sized cluster, carbon-aware scheduling cut emissions from about 665 kg CO₂e/month to 450 kg CO₂e/month, a 32% reduction. At large scale, shifting to renewable-backed, lower-PUE infrastructure cut operational emissions by about 79%, while making the water–carbon trade-off more important to report. The authors argue for routine disclosure of energy use, emissions, water use, hardware assumptions and regional context.4
Interpretation: The calculator results show that the size of a facility's footprint and the size of the achievable saving both depend on design choices: scheduling, power source and cooling efficiency. The 79% figure applies to operational emissions under a specific infrastructure shift, not to total life-cycle emissions.4
03
Why communities are pushing back
AI summary:US candidates from both parties are opposing data-centre development ahead of the midterms, though local grievances aren't detailed here.
Evidence-backed: The BBC reports that Democrats and Republicans alike are coming out against data-centre development, amid a growing backlash across the US, and that this could shape the midterm elections.5
Interpretation: That report establishes that opposition is politically broad and electorally salient, but it does not set out the specific local grievances behind it. The concerns named in this page's scope, such as grid strain, competition for water, noise and land use, are not documented with figures or case detail in the sources available here.5
Interpretation: The technical evidence does point to plausible sources of local tension. A facility that sends over 99.9% of its energy use to computing still draws that power from the local grid, and water-based cooling creates demand for water that the wastewater-pairing study treats as a problem worth solving. The pairing proposal is itself a response to the water and emissions pressure data centres create.23
Before you go
What to remember
Try to recall each hidden figure before you reveal it. Remembering, not rereading, is what makes it stick.
In one closely measured Swedish facility, over of energy use went to data-processing processes rather than support systems, a share the authors call unusually high for a production sector.
Pairing data centres with wastewater treatment plants is modelled to save about million m³ of freshwater and 84 million tonnes of CO₂-equivalent a year, with the largest potential in the US, Japan, China, the Netherlands and the UK.
Design choices matter: carbon-aware scheduling cut mid-sized cluster emissions by , and a shift to renewable-backed, lower-PUE infrastructure cut operational emissions by about 79%.
Your reading
0 of 3 chaptersThis answer keeps changing
When new evidence or a better source comes in, this page is updated (it's on version 2, last changed 38 minutes ago). Follow it to be told when that happens.
Up nextHow do data centres use and affect local water supplies?How do data centres use water, and what effects do their water withdrawals and discharges have on local water supplies and communities?Ask this Sylo
Still wondering about something?
Answers come only from this page's reviewed material, with citations, and say plainly when the page doesn't cover it yet.
More on AI
Everything on AI ›
Photo: Maxim KlimashinWhy are tech and AI stocks falling in 2026?
Photo: AbsolutVisionWhy is USA Today suing OpenAI over training data?
Why are eight US states suing OpenAI over ChatGPT safety?
Image: METRThe future of work: AI, shorter weeks and healthier desks
How are AI tools, shorter working weeks and new workplace habits changing knowledge work, and what does the evidence say?
Photo: Resume GeniusHow do AI job applications backfire for graduates?
How is AI regulated compared with other industries?
How does the regulation of artificial intelligence compare with how other industries are regulated?
Behind this page
Who's adding to it, where it comes from, how it changed and what would make it better. Always open to everyone.
Discussion
Sources
Numbers match the citations in the article. A working link isn't proof that a page supports a claim; check the quoted passage and date.
- 1Data center (Wikipedia)WikipediaPublished Oct 16, 2026Checked Oct 17, 2026
“A data center is a physical room, building, or facility for storing, managing, and disseminating data and information, including computer systems and associated components, housing IT infrastructure, and training artificial intelligence. The United States' act of Congress, the Energy Independence and Security Act of 2007, defines a data center as "any facility that primarily contains electronic equipment used to process, store, and transmit digital information." This includes "a free-standing structure" or "a facility within a larger structure, that uses environmental control equipment to maintain the proper conditions for the operation of electronic equipment." According to IBM, data centers date back to the 1940s, with the U.S. military's Electrical Numerical Integrator and Computer (ENIAC) as an early example. Data centers have different models depending on their workloads.”
- 2Pursuing a hierarchical taxonomy in a water-cooled data center room.Scientific reports (Milić et al.)Published Aug 24, 2026Checked Oct 17, 2026
“The methodology combines hierarchical classification of energy flows with quantitative analysis based on time-resolved operational data, allowing for detailed analysis of dynamic system behaviour and allocation of energy use across processes, applied to a DC room with a detached in-rack water cooling system in Sweden operated by Ericsson. The results of this paper present a novel hierarchical taxonomy for a water-cooled DC room with six defined EEU processes, of which four are related to data processing processes and two to support processes. Notably, over 99.9% of total EEU is connected to data processing processes which is a significantly higher figure compared to other production processes in other sectors. Moreover, the established EnPIs can facilitate benchmarking between DCs with similar infrastructures and for monitoring the energy use of various EEU processes. The proposed methodology is transferable to other DC environments with comparable system boundaries and data availability, and its reproducibility is supported by the structured and data-driven methodology, which enables replication of the analytical approach given comparable data availability.”
- 3Global data-water symbiosis reduces AI infrastructure's carbon and water footprint.Environmental science and ecotechnology (Wang et al.)Published Apr 26, 2026Checked Oct 17, 2026
“The analysis reveals a strong global spatial co-occurrence between data centres and WWTPs, enabling optimized national-scale pairings in which treated effluent is used for data centre cooling and the return heat is recovered to support sludge drying and anaerobic digestion. This symbiotic approach reduces greenhouse gas emissions by approximately 84 million tonnes of CO2 equivalent annually, conserves approximately 1300 million m3 of freshwater, and provides net annual cost savings of approximately US$95.4 billion. The greatest mitigation and water-saving potential lies in the United States, Japan, China, the Netherlands, and the United Kingdom. These findings establish data-water symbiosis as a readily scalable infrastructure solution that decouples AI from its carbon and water footprints. WWTPs are poised to evolve from disposal facilities into critical energy-coupling hubs, enabling efficient thermal and water exchange across urban systems and accelerating progress towards multiple Sustainable Development Goals.”
- 4Quantifying and Disclosing the Environmental Footprint of AI in Research: Life Cycle-Informed Framework and Open-Access Calculator Development Study.JMIR AI (Mozafari et al.)Published Aug 18, 2026Checked Oct 17, 2026
“In the small laboratory scenario, optimization reduced total emissions from approximately 0.07 kg carbon dioxide equivalents (CO2e) to 0.05 kg CO2e and improved the proposed label from C to B. In the midsized cluster scenario, carbon-aware scheduling reduced emissions from approximately 665 kg CO2e/month to 450 kg CO2e/month-a 32% reduction. In the large-scale scenario, shifting to renewable-backed, lower-PUE infrastructure reduced operational emissions by approximately 79% while increasing the importance of water-carbon trade-off reporting.ConclusionsThe calculator provides a practical and transparent method for reporting AI environmental footprints using auditable parameters and publication-ready outputs. Routine disclosure of energy use, emissions, water use, hardware assumptions, and regional context can improve reproducibility and support more equitable and sustainable AI research evaluation.”
- 5How backlash against data centres could shape the US midterm electionsBBC NewsPublished Oct 14, 2026Checked Oct 17, 2026
“Democrats and Republicans alike are coming out against data centre development, amid growing backlash across the US.”
How it changed
Published 1 time since Oct 17, 2026.
- Version 2Oct 17, 2026Live now
AI-prepared Starting Map from live research.
- First published version.
Help improve it
The brief is open about what's uncertain. These are the specific gaps that new material would fill.
Open questions
Which specific local concerns, such as grid strain, water competition, noise or land use, are driving the backlash in particular US districts, and how do they differ between projects?
No answers yet
What are the best current estimates of total electricity and water consumption by AI data centres, as distinct from modelled potential savings?
No answers yet
What practical, regulatory or cost barriers stand between the wastewater-pairing proposal and deployment at the scale the study models?
No answers yet
Would routine disclosure of energy, emissions, water use and regional context, as the calculator authors propose, change community opposition or trust?
No answers yet
Around this topic
Sylos connect: narrower topics report up to broader ones, so what's learned in one place shows up where it matters.