Skip to content

GCRI · Artificial Intelligence and Autonomous Systems

Compute and energy

Training and running AI requires specialised chips, data centres, electricity, cooling water and network capacity. Demand is growing quickly and concentrating in particular places, with effects on power grids, water resources and emissions. This page sets out how AI compute uses energy and water, how those uses are measured, and what evidence serves the grid operators, planners and policymakers who must accommodate them.

01Ilaphu lokuzoba

Indlela umsebenzi oqhuba ngayo.

Iinkqubo · ezijikeleze esi sifundoNgumzekelo

Model registry

Compute and energy · Model registry · Release gate

One model blocked at gateGCRI · Artificial Intelligence and Autonomous Systems

Models

M-904 triage assistantPublic serviceIphezulu19 of 24 suitesKuthintelwe
M-890 flood nowcastScienceIphakathi10 of 12Released
M-955 document searchInternalIphantsi7 of 8Released
M-242 agent plannerAgenticIphezulu13 of 16Evaluating

Model card

Intended use
triage support, human decides
Out of scope
eligibility decisions
Known limits
lower accuracy on rare cases
Umnini
Model owner

Incident log

  1. D-28INCHallucinated citation in summary
  2. D-6MITRetrieval grounding added
  3. D-4INCPrompt injection via uploaded file
  4. D-1MITTool permission narrowed

Evaluation scores

vs release threshold

Task accuracy71 of 85 requiredRobustness71 of 75 requiredFairness gap (lower is better)3 of 5 requiredRefusal on misuse91 of 95 required
Model registry: the screen’s structure, terms and states for Compute and energy.Example names and illustrative values; a design, not a record or a live system.

Fig.Model registry: example names and illustrative values, not a record or a live system.

02The question

Compute and energy: how the risk develops.

Compute concentration, efficiency and demand. The dossier below sets out how the risk develops, what is measured, where the data come from, how it is modelled and who decides.

Indlela yokusebenza

Training and inference both use energy

Training a large model uses a large amount of electricity once; inference uses less per query but runs continuously at scale, and can dominate over a model’s life.

Indlela yokusebenza

Data centres strain local grids

Large data centres need connections comparable to industrial plants. Clusters in one region can outpace grid and generation build-out.

Indlela yokusebenza

Cooling uses water

Evaporative cooling and electricity generation consume water. Where water is scarce, data centre siting competes with other uses.

Indlela yokusebenza

Efficiency and demand move together

Hardware and software efficiency improve quickly, but lower costs can raise total use. Net energy demand depends on both.

03Ijografi

Energy now in East Asia.

Night lights, population density and land cover over East Asia, from NASA GIBS. Each layer carries its own date.

Imephu esebenzisanayo idinga i-JavaScript; uluhlu lweeleya nemithombo zingezantsi.

Iileya 3

Energy now, for this page

Malunga nale leya
Ubungqina
Kuqwalaselwe
Umthombo
VIIRS Black Marble (VNP46A2) (NASA LAADS DAAC)
Umboneleli
NASA LAADS DAAC
Ilayisensi
Idatha evulekileyo ye-NASA
Uhlelo
imihla ngemihla; Ukukhanya okulinganiswe ngesathelayithi, kulungiselelwe ukukhanya kwenyanga nomoya
Okutsha ngexesha lokwakha
2026-10-12
Ucaciso
igridi emalunga ne-500 m
Ukungaqiniseki
Ukukhanya kwasebusuku okulinganiswe ngesathelayithi komhla oboniswayo; izibane zingumbonakaliso wokusetyenziswa kombane nokuqhawuka, hayi umlinganiselo wobonelelo okanye wokufikelela.
Kufikelelwe nge
this site's data relay, from gibs.earthdata.nasa.gov
Malunga nale leya
Ubungqina
Kuqikelelwe
Umthombo
Gridded Population of the World v4.11 (NASA SEDAC (CIESIN, Columbia University))
Umboneleli
NASA SEDAC (CIESIN, Columbia University)
Ilayisensi
CC BY 4.0
Uhlelo
null; Census-based estimate, static (2020)
Ucaciso
igridi emalunga ne-1 km
Ukungaqiniseki
Uqikelelo olusekelwe kubalo-bantu lwabantu kwikhilomitha-sikwere nganye ngo-2020, luboniswe njengemeko yokuchaphazeleka; asilonani labantu abachatshazelwe yiyo nayiphi na ingozi.
Kufikelelwe nge
this site's data relay, from gibs.earthdata.nasa.gov
Malunga nale leya
Ubungqina
Modelled (latest month GIBS publishes)
Umthombo
MERRA-2 (NASA GMAO / GES DISC)
Umboneleli
NASA GMAO / GES DISC
Ilayisensi
Idatha evulekileyo ye-NASA
Uhlelo
monthly; Reanalysis model; latest month GIBS publishes
Okutsha ngexesha lokwakha
2026-07-01
Ucaciso
igridi emalunga ne-2 km
Ukungaqiniseki
Reanalysis model estimate of monthly mean incoming sunlight over land; an indicator of solar resource, not a site yield assessment.
Kufikelelwe nge
this site's data relay, from gibs.earthdata.nasa.gov
Malunga nale leya
Ubungqina
Kuqwalaselwe
Umthombo
VIIRS NOAA-20 corrected reflectance (NASA LANCE)
Umboneleli
NASA LANCE
Ilayisensi
Idatha evulekileyo ye-NASA
Uhlelo
imihla ngemihla; Umfanekiso wesathelayithi
Okutsha ngexesha lokwakha
2026-10-14
Ucaciso
igridi emalunga ne-250 m
Ukungaqiniseki
Umfanekiso wesathelayithi. I-raster yesathelayithi okanye yemodeli njengoko umboneleli eyipapasha. Amafu, izikhewu zokugubungela neempazamo zokufumana zishiya izikhewu nengxolo; amaxabiso zizalathisi, awayomilinganiselo kwindawo ethile.
Kufikelelwe nge
this site's data relay, from gibs.earthdata.nasa.gov
Malunga nale leya
Ubungqina
Kumodelwe (unyaka wamva nje opapashwa yi-GIBS)
Umthombo
MODIS Terra and Aqua isigubungelo somhlaba (MCD12Q1) (NASA LP DAAC)
Umboneleli
NASA LP DAAC
Ilayisensi
Idatha evulekileyo ye-NASA
Uhlelo
ngonyaka; Ukuhlelwa ngesathelayithi; unyaka wamva nje opapashwa yi-GIBS
Okutsha ngexesha lokwakha
2024-01-01
Ucaciso
igridi emalunga ne-500 m
Ukungaqiniseki
Iiklasi zesigubungelo somhlaba ngesathelayithi zonyaka oboniswayo, kuquka imihlaba elinyiweyo; ukuhlelwa okuneempazamo ezaziwayo, hayi irejista yeefama.
Kufikelelwe nge
this site's data relay, from gibs.earthdata.nasa.gov
Malunga nale leya
Ubungqina
Kuqwalaselwe (amafu ashiya izikhewu)
Umthombo
VIIRS NOAA-20 (NASA LANCE)
Umboneleli
NASA LANCE
Ilayisensi
Idatha evulekileyo ye-NASA
Uhlelo
imihla ngemihla; Ukufunyanwa kwedatha ngesathelayithi; amafu ashiya izikhewu
Okutsha ngexesha lokwakha
2026-10-14
Ucaciso
igridi emalunga ne-1 km
Ukungaqiniseki
Satellite skin temperature of the ground on the day shown, an indicator of cooling demand; not air temperature or a heat warning.
Kufikelelwe nge
this site's data relay, from gibs.earthdata.nasa.gov

Okutsha okukhoyo

NASA GIBS / EOSDIS · ASTER GDEM (METI, NASA) · Blue Marble (NASA) · © abanegalelo kwi-OpenStreetMap (ODbL) · Natural Earth

Kulayishwa imifanekiso yamva nje ye-NASA.

Rhuqa ukuze uhambise. Ctrl okanye ⌘ uze ujikelezise, chophela okanye usebenzise amaqhosha ukusondeza; beka isalathisi phezu kwento ukuze ubone amaxabiso aphantsi kwaso; ukubuyisela kubuyela kummandla.

Energy now, framed on East Asia.Satellite and model products as NASA publishes them, each with its own date. They are indicators, not assessments of any place.

Data: NASA GIBS · ASTER GDEM · Blue Marble · © OpenStreetMap contributors · Natural Earth · NASA EONET · NASA/NOAA VIIRS · NASA JPL OPERA DIST-ALERTSources and licences

  • We acknowledge the use of imagery provided by services from NASA’s Global Imagery Browse Services (GIBS), part of NASA’s Earth Science Data and Information System (ESDIS). Colours redrawn from the published colour map; values averaged into hexagonal cells where shown (modified). NASA open data
  • We acknowledge the use of imagery provided by services from NASA’s Global Imagery Browse Services (GIBS), part of NASA’s Earth Science Data and Information System (ESDIS). Colours redrawn from the published colour map; values averaged into hexagonal cells where shown (modified). NASA open data
  • VIIRS Black Marble night-time lights, NASA Goddard Space Flight Center; distributed by the NASA Level-1 and Atmosphere Archive and Distribution System (LAADS DAAC); imagery via NASA GIBS. Colours redrawn from the published colour map; values averaged into hexagonal cells where shown (modified). NASA open data
  • Center for International Earth Science Information Network (CIESIN), Columbia University. 2018. Gridded Population of the World, Version 4 (GPWv4): Population Density, Revision 11. NASA Socioeconomic Data and Applications Center (SEDAC). Licensed under CC BY 4.0. Colours redrawn from the published colour map; values averaged into hexagonal cells where shown (modified). doi:10.7927/H49C6VHW. CC BY 4.0
  • MERRA-2 reanalysis, NASA Global Modeling and Assimilation Office; distributed by the NASA GES DISC; imagery via NASA GIBS. Colours redrawn from the published colour map; values averaged into hexagonal cells where shown (modified). NASA open data
  • VIIRS near-real-time products (flood, land surface temperature, snow cover, corrected reflectance), NASA LANCE; imagery via NASA GIBS. Colours redrawn from the published colour map; values averaged into hexagonal cells where shown (modified). NASA open data
  • MODIS land products (vegetation indices, gross primary production, land cover), NASA LP DAAC and LANCE; imagery via NASA GIBS. Colours redrawn from the published colour map; values averaged into hexagonal cells where shown (modified). NASA open data
  • ASTER GDEM is a product of METI and NASA (ASTER Global Digital Elevation Model V003, NASA LP DAAC); imagery via NASA GIBS. Colours redrawn from the published colour map; values averaged into hexagonal cells where shown (modified). Cite as: NASA/METI/AIST/Japan Spacesystems and U.S./Japan ASTER Science Team (2019). ASTER Global Digital Elevation Model V003. doi:10.5067/ASTER/ASTGTM.003. NASA open data
  • Yenziwe nge-Natural Earth. Idatha yemephu ye-vector ne-raster esimahla @ naturalearthdata.com. Yeyoluntu
  • © OpenStreetMap contributors (ODbL), via NASA GIBS (coastlines, boundaries, roads and place names). Colours redrawn from the published colour map; values averaged into hexagonal cells where shown (modified). ODbL
  • Blue Marble, NASA Earth Observatory; imifanekiso ivela kwi-NASA GIBS. Imibala izotywe ngokutsha ukusuka kwimephu yemibala epapashiweyo; amaxabiso afakwe umndilili kwiiseli ezineheksagoni apho aboniswa khona (iguqulwe). Idatha evulekileyo ye-NASA
  • Imeko kwigloubhu nakwimephu ethe tyaba: unxweme, imida, iindlela namagama eendawo avela kwiileya zesalathiso ze-NASA GIBS (idatha ye-OpenStreetMap, ODbL, © abanegalelo kwi-OpenStreetMap); imifanekiso ye-Blue Marble njengesiseko esikhethwayo; umhlaba we-3D nokufundwa kokuphakama okuvela kwisalathiso semibala sokuphakama se-ASTER GDEM (iiklasi ezisuka kwi-5 m ukuya kwi-200 m, zandisiwe ukuze zifundeke). Imida izotywe njengoko ipapashiwe kwaye ayibonisi kuma kuwo nawuphi na ummandla.
  • Iziganeko ezenzeka ngoku: NASA Earth Observatory Natural Event Tracker (EONET); iziganeko zingumqondiso nje kwaye zigcina amakhonkco azo emithombo yokuqala.
  • Observed (EO): NASA/NOAA VIIRS (Suomi NPP). We acknowledge the use of imagery provided by services from NASA’s Global Imagery Browse Services (GIBS), part of NASA’s Earth Science Data and Information System (ESDIS).
  • Observed (EO): NASA JPL OPERA DIST-ALERT (contains modified Copernicus Sentinel data 2026). We acknowledge the use of imagery provided by services from NASA’s Global Imagery Browse Services (GIBS), part of NASA’s Earth Science Data and Information System (ESDIS).
  • NASA GIBS: We acknowledge the use of imagery provided by services from NASA’s Global Imagery Browse Services (GIBS), part of NASA’s Earth Science Data and Information System (ESDIS).
  • Unyaka odlulileyo, ubushushu bemihla ngemihla (ERA5 nge-Open-Meteo): Iqulethe ulwazi oluguqulweyo lwe-Copernicus Climate Change Service 2026.
  • Unyaka odlulileyo, ubushushu bemihla ngemihla nokukhanya kwelanga (NASA POWER): Le datha ifunyenwe kwi-NASA Langley Research Center (LaRC) POWER Project exhaswa ngemali yi-NASA Earth Science/Applied Science Program..
  • Idatha yemozulu ivela kwi-Open-Meteo.com (https://open-meteo.com/), inelayisensi phantsi kwe-CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).
  • Optional layers (10 sources in the layer list: live, forecast, observed and context) print their provider credit on the map and here when switched on; forecasts are model output and derived indices are estimates.

Iileya nemihla

  • Night-time lights, daily: VIIRS Black Marble (VNP46A2), NASA LAADS DAAC · 12 Oct 2026
  • Population density, 2020: Population: CIESIN, Columbia University, Gridded Population of the World v4.11 (2018), NASA SEDAC, CC BY 4.0 · static
  • Incoming sunlight at the surface over land, monthly mean: MERRA-2, NASA GMAO / GES DISC · July 2026
  • True colour: VIIRS NOAA-20 corrected reflectance, NASA LANCE · 13 Oct 2026
  • Uhlobo lwesigubungelo somhlaba (IGBP), ngonyaka: MODIS Terra and Aqua isigubungelo somhlaba (MCD12Q1), NASA LP DAAC · 2024
  • Land surface temperature, day: VIIRS NOAA-20, NASA LANCE · 13 Oct 2026

04Across the domains

AI compute couples to energy and water.

Compute and energy analyses are designed for the Foresight Outlook series. The IEA’s Energy and AI report (2025) gives scenario ranges for data-centre electricity demand.

  • Measured energy and water use of computing facilities.
  • Grid connection studies.
05Simulator

A cascade through interdependent networks.

A hazard footprint takes out co-located assets in every layer; fragmentation, dependency between layers and overload of the survivors then spread the failure. The 3D view separates the layers by depth. Teaching model with stated illustrative values, not a prediction. The map on this page shows East Asia; the model itself is not tied to a place.

Izilawuli

Assets per layer
18
Hazard footprint
8% of area
Assets that depend on another layer
60%
Spare capacity (tolerance)
30%

Iziphumo ezifundwayo

Assets hit by the hazard
6%
Assets still working
20%
Amplification of the initial damage
14.3 times
Amanyathelo okusombulula
4
Amanani, iindlela nemithombo
InyatheloData centres working (%)Networks working (%)Power working (%)
08994100
14444100
262278
362239
462233

Iindlela zokusebenza

  1. Interdependent networks Buldyrev and others (2010), Catastrophic cascade of failures in interdependent networks, Nature 464
  2. Overload cascades Motter and Lai (2002), Physical Review E 66
  3. Betweenness load Brandes (2001), Journal of Mathematical Sociology 25
  4. Multilayer networks Kivela and others (2014), Journal of Complex Networks 2
  5. Infrastructure interdependencies Rinaldi, Peerenboom and Kelly (2001), IEEE Control Systems 21(6)
  6. Uncertainty fan (parameter Monte Carlo) Spiegelhalter, Pearson and Short (2011), Visualising uncertainty about the future, Science 333
  7. One-at-a-time sensitivity (tornado) Saltelli and others (2008), Global Sensitivity Analysis: The Primer, Wiley

Idatha

The model runs on the teaching values set by the controls. To calibrate it for a place, read it against these admitted layers and series, which this page’s map can draw:

  • Major roads Natural Earth
  • Railways Natural Earth
  • Seaports Natural Earth
  • Airports Natural Earth
  • Urban extents Natural Earth
  • Night lights tonight, outages (VIIRS Day/Night Band) NASA/NOAA VIIRS (Suomi NPP) · register nasa.gibs.context-live-layers
  • Natural events (NASA EONET) NASA EONET · register nasa.eonet.events
A cascade through interdependent networks. Illustrative teaching model: it runs on the values set by the controls, not on measured data. Methods: Interdependent networks (Buldyrev and others (2010), Catastrophic cascade of failures in interdependent networks, Nature 464); Overload cascades (Motter and Lai (2002), Physical Review E 66); Betweenness load (Brandes (2001), Journal of Mathematical Sociology 25).
06Izalathisi

What is measured, and how well.

Each indicator carries its unit, its cadence, the kind of source behind it and the evidence label a value would carry. No values are shown: a programme supplies them under its own authority.

IndicatorUnitCadenceSource typeIleyibhile
Data centre electricity consumptionterawatt hours per yearannualEnergy agency estimatesKUQIKELELWE
Power usage effectivenessratio of facility to IT energyannualOperator disclosuresDECLARED
Water usage effectivenesslitres per kilowatt hourannualOperator disclosuresDECLARED
Grid connection requests from data centresmegawatts requestedquarterlyGrid operator queuesPublished statistics
Training compute of notable modelsfloating point operationsper modelCompiled model databasesKUQIKELELWE
Energy per inferencewatt hours per queryper measurementMeasurement studiesKUQIKELELWE
07Isilinganisi sokuphindwa nothotho lweziphumo

Model compute and energy as a system. Run the cascade.

Assembled from the Nexus Standards area profile AI and autonomous systems risk standards: the hazards it names, the systems and sectors that carry the loss, the controls that dampen it and the data that observe it. Choose the components, pick an initiating hazard and step through the cascade. The Map view places the run on an example geography with open data. Every coupling weight is a stated assumption you can change. A teaching model, not a forecast, an assessment or a certification. To model a place of your own choosing, open the full System Builder.

Iingozi

  • Dam failure
  • Sea-ice loss
  • Coastal erosion (set off downstream)
  • River flood (set off downstream)
  • Flash flood (set off downstream)
  • Ukufuduswa kwabantu abaninzi (kubangela iziphumo ezilandelayo)
  • Biodiversity loss (set off downstream)
  • Ukuqhawuka kombane okuxhaphakileyo (kubangela iziphumo ezilandelayo)
  • Water supply shortfall (set off downstream)
  • Arctic winter sea-ice collapse (set off downstream)
  • Polar vortex disruption (set off downstream)

Izilawuli

  • NS-0 Umgaqo-siseko nemigaqo
  • NS-10 Imigaqo-nkqubo namagatya (izibophelelo ezifundeka ngoomatshini: inkcazo, ukubopha, uvavanyo nokunyanzeliswa)
  • NS-11 Ukuqinisekiswa nokubekwa kweliso (ubungqina, iziqinisekiso, uqwalaselo lokuhambelana ngaphandle kokuthemba)
  • NS-12 Software, model and AI/ML artefacts (supply chain, model cards, dataset sheets)
  • NRS-H100 Hazard family: Robotics, autonomy and autonomous weapons (NHS-1 tech.autonomy-weapons)
  • NRS-H53 Hazard family: Cyber-physical disruption (NHS-1 cyber.cyber-physical)
  • NRS-H61 Hazard family: Cascading failure (NHS-1 sys.cascade)
  • NRS-H113 Hazard family: Financial-system fragility (NHS-1 soc.financial-system)
  • NRS-A8 AI-ready risk data and model packaging
  • NRS-H54 Hazard family: Signal and platform disruption (NHS-1 cyber.platform)

Uqwalaselo

  • OSV open source vulnerability database (listed for reference)
  • Our World in Data datasets (listed for reference)
  • Research and development expenditure as a proportion of GDP

Umzekelo wejografi

  • North America: 243 hexagonal cells
  • Izibane zasebusuku namhlanje, ukuqhawuka kombane (VIIRS Day/Night Band), NASA GIBS
  • Amanzi aphezu komhlaba phantsi kwamafu, 30 m (OPERA DSWx-S1 radar), NASA GIBS
  • Isantya semvula, isiqingatha seyure esidlulileyo (IMERG), NASA GIBS
  • Vegetation disturbance alerts, 30 m (OPERA DIST-ALERT), NASA GIBS
08Imithombo

Where the evidence comes from.

Named public datasets and institutional records, each with its licence class. Sources by agreement are used only under their terms and never republished.

UmthomboWhat it givesLicence class
IEA Energy and AI analysis and data centre electricity estimatesGlobal and regional data centre electricity demandLicence per product (to check)
Epoch AI data on notable models and computeTraining compute and hardware trendsOpen, CC BY 4.0
TOP500 and Green500 listsPerformance and energy efficiency of supercomputersPublic, TOP500 terms
Grid operator interconnection queues and planning reportsRequested connections and load forecastsPublic records, terms per operator
Operator sustainability reportsEnergy, water and emissions disclosuresPublic documents, terms per operator
09Uphawu lwecandelo

The sector signature of compute and energy.

The sectors compute and energy touches and how strongly they depend on each other: what a study has to observe and model together.

Khetha iikona · beka isalathisi phezu kwemigcaubudlelwane obu-6 / 15

Uphawu

Amandla namaqabane awo

Amandla anika iziseko zophuhliso umbane wezothutho nonxibelelwano (0.8), amanzi ukumpompa nokucoca (0.5) nokutya isichumisi, oomatshini nokugcina kubanda (0.5).

W · E · I · F · H · B

Amacandelo
4 kwezi-6
Ubudlelwane
6 kwezi-15
Unxibelelwano
0.46 isabelo sesixa sisonke

Ubudlelwane obomeleleyo obusebenzayo

  • Energy → Infrastructure · power for transport and telecoms0.8
  • Water → Energy · cooling water and hydropower0.6
  • Amanzi → Ukutya · unkcenkceshelo0.8
  • Iziseko zophuhliso → Amanzi · imibhobho, amadama nezindlu zokucoca0.6
Amaxabiso okufundisa kummandla wesixeko oqhelekileyo, awayomilinganiselo yayo nayiphi na indawo. Unxibelelwano yimbumba yamandla obudlelwane obusebenzayo, njengesabelo sabo bonke abalishumi elinesihlanu.
10Decisions

Who decides, and what the evidence serves.

Artificial Intelligence and Autonomous Systems prepares evidence. The decisions below are taken by the authorities named, under their own rules, and never by the platform.

Who decidesThe questionEvidence this page serves
Grid operators and energy regulatorsHow to plan generation and connections for data centre loadLoad analyses; connection queues
Water authorities and plannersWhere data centre water use is acceptableWater footprint evidence
Organisations deploying AIHow to reduce the footprint of their AI useEstimation methods; efficiency evidence

Outputs this topic produces

  • Energy and water footprint estimation methods for AI workloads
  • Regional data centre load analyses
  • Efficiency trend summaries
  • Evidence notes for grid and water planning
11Methods and evidence

How compute and energy is modelled, measured and tested.

The model, its assumptions, its data and its uncertainty, in the form a scientist or engineer would check. Structures follow published methods. The values drawn are illustrative, not findings.

System model

The couplings between the systems in compute and energy: signs of influence, a reinforcing and a balancing loop, and cross-links where stress jumps.

++−+Ubukrelekrele bokwenziwaIdathaCommunitiesEconomyR reinforcingB balancing
Causal loop diagram from the page’s systems; signs show the direction of influence. Illustrative structure for a system dynamics or network model, not estimated couplings.
FAIR guiding principles · GO FAIR and FORCE11W3C PROV · World Wide Web ConsortiumGuidance on consistent treatment of uncertainties · IPCCCEOS Analysis Ready Data · Committee on Earth Observation Satellites
12Core questions

What researchers ask about compute.

  • How concentrated are chip manufacturing, cloud capacity and frontier model training, and what follows from that?
  • How much energy and water do training and inference use, and how is that changing?
  • How much do efficiency gains offset growth in demand?
  • Who has access to compute for research and public-interest uses?
13Mechanisms

Concentration, efficiency and demand.

Advanced chip fabrication is concentrated in very few facilities, and a small number of cloud providers supply much of the compute for AI. That concentration creates shared exposure to disruption, from natural hazards to export controls, and shapes who can build large models.

Hardware and algorithms become more efficient over time, but cheaper computation tends to increase total use, an effect known as the rebound or Jevons effect. Data centres draw large, steady electricity loads that can strain local grids and, where they use evaporative cooling, water supplies. Inference, the everyday use of models, can come to outweigh training in total energy use.

14Methods and data

Measuring compute and its footprint.

  1. 01

    Energy accounting

    Estimating energy per training run and per query, using high-performance, cloud and edge computing. Providers disclose little, so estimates carry wide uncertainty.

  2. 02

    Supply-chain mapping

    Tracing chips, materials and facilities to identify concentration and single points of failure.

  3. 03

    Scenario analysis

    Projecting demand under different assumptions about efficiency and adoption, with probability and statistics. Projections are scenarios, not forecasts.

  4. 04

    Access studies

    Assessing which researchers and institutions can obtain compute, and on what terms.

15Iindlela zokusebenza

How it is studied, and what the methods cannot say.

  1. High-performance, cloud and edge computingCompute access for public research · Energy measurement of workloadsApplied here
  2. Artificial intelligence and machine learningDistribution shift analysis · Fine-tuning for a specific use · Interpretability methodsApplied here
  3. Probability and statisticsUncertainty on benchmark scores · Calibration assessment · Audit samplingApplied here
16Where this page’s references

Where this page’s references lead.

Amaqumrhu namaxwebhu eli phepha elifundwa ngokuthelekiswa nawo. Imeko neenguqulelo ziyatshintsha; ikhonkco ngalinye liya kwiqumrhu elikhuphayo.

17Imibuzo

Imibuzo neempendulo.

How much energy does AI use?
Estimates vary widely. Data centres use a growing share of electricity, and AI is a fast-growing part of that, according to the IEA, but precise figures depend on assumptions and disclosure.
What is power usage effectiveness?
The ratio of a data centre’s total energy use to the energy used by its IT equipment; values closer to one mean less overhead for cooling and power conversion.
Why does AI use water?
Mainly for cooling data centres, especially with evaporative systems, and indirectly through the water used to generate electricity.
Will efficiency gains reduce AI’s energy use?
Not necessarily. Efficiency lowers the cost per task, which can increase total demand. The net effect depends on adoption and use.
Why does water matter for data centres?
Many use evaporative cooling, which consumes water. In water-stressed areas, this competes with other uses, linking AI to local water risk.
Does the platform allocate compute?
No. Compute allocations for Consortium projects are a separate matter of Foundry and the annual programme, and open only when their terms exist.
Can energy figures be trusted?
Published estimates vary with assumptions about hardware, utilisation and location. The platform reports ranges with their sources and assumptions.
IingxeloNikela

Faka ubungqina obuvela kweli phepha.

Reports is the Consortium’s quarterly publication. What you know about this subject can become evidence, a case or a correction in the next edition, credited to you and published under CC BY 4.0 where it is your own work.

GCRIArtificial Intelligence and Autonomous SystemsExponential TechnologyLabs Evidence
Ubukrelekrele Bokwenziwa Neenkqubo Ezizimeleyo

Bring a method, a dataset or a question.

Information and enquiries are open today. An enquiry receives a reference and names the next step; applications, payments, compute allocations and registrations open only when their team, terms and review capacity exist.