利用人工智能加速气候行动报告（英文版）--波士顿 VIP专享VIP免费

下载本文档

阅读 803
格式 pdf
大小 2.87 MB
约52页
2023-12-30
收藏
点赞(0)
海报
举报

Accelerating Climate

Action with AI

November 2023

By Amane Dannouni, Stefan A. Deutscher, Ghita Dezzaz, Adam Elman, Antonia Gawel,

Marsden Hanna, Andrew Hyland, Amjad Kharij, Hamid Maher, David Patterson,

Edmond Rhys Jones, Juliet Rothenberg, Hamza Tber, Maud Texier, and Ali Ziat

Boston Consulting Group partners with leaders

in business and society to tackle their most

important challenges and capture their greatest

opportunities. BCG was the pioneer in business

strategy when it was founded in 1963. Today,

we work closely with clients to embrace a

transformational approach aimed at beneting all

stakeholders—empowering organizations to grow,

build sustainable competitive advantage, and

drive positive societal impact.

Our diverse, global teams bring deep industry and

functional expertise and a range of perspectives

that question the status quo and spark change.

BCG delivers solutions through leading-edge

management consulting, technology and design,

and corporate and digital ventures. We work in a

uniquely collaborative model across the rm and

throughout all levels of the client organization,

fueled by the goal of helping our clients thrive and

enabling them to make the world a better place.

Contents

01 Foreword

02 Executive Summary

05 The Climate Action Imperative

and the Promise of AI

09 How AI Can Help

Accelerate Climate Action

22 Navigating AI’s Potential Risks

28 AI for Climate:

A Summary of Critical Policy

Outcomes

41 Endnotes

43 About the Authors

45 Acknowledgements

47 References

Contents

/52

下载本文档

文本预览下载提示常见问题

AcceleratingClimateActionwithAINovember2023ByAmaneDannouni,StefanA.Deutscher,GhitaDezzaz,AdamElman,AntoniaGawel,MarsdenHanna,AndrewHyland,AmjadKharij,HamidMaher,DavidPatterson,EdmondRhysJones,JulietRothenberg,HamzaTber,MaudTexier,andAliZiatBostonConsultingGrouppartnerswithleadersinbusinessandsocietytotackletheirmostimportantchallengesandcapturetheirgreatestopportunities.BCGwasthepioneerinbusinessstrategywhenitwasfoundedin1963.Today,weworkcloselywithclientstoembraceatransformationalapproachaimedatbenefitingallstakeholders—empoweringorganizationstogrow,buildsustainablecompetitiveadvantage,anddrivepositivesocietalimpact.Ourdiverse,globalteamsbringdeepindustryandfunctionalexpertiseandarangeofperspectivesthatquestionthestatusquoandsparkchange.BCGdeliverssolutionsthroughleading-edgemanagementconsulting,technologyanddesign,andcorporateanddigitalventures.Weworkinauniquelycollaborativemodelacrossthefirmandthroughoutalllevelsoftheclientorganization,fueledbythegoalofhelpingourclientsthriveandenablingthemtomaketheworldabetterplace.Contents01Foreword28AIforClimate:02ExecutiveSummaryASummaryofCriticalPolicyOutcomes05TheClimateActionImperative41EndnotesandthePromiseofAI43AbouttheAuthors45Acknowledgements09HowAICanHelpAccelerateClimateAction22NavigatingAI’sPotentialRisks47ReferencesForewordThisreportaimstoprovidepolicymakers,corporatedecisionmakers,andclimateleaderswithaclearandconciseunderstandingoftherolethatartificialintelligence(AI)canplayinclimateaction.Morespecifically,itsgoalsaretohighlightAI’ssignificantThisworkdrawsoninterviewswitharangeofclimatepotentialtohelpaddressourenvironmentalchallenges,changeandAIexperts,buildsonpreviousresearchfromtoshedlightonclimate-relevantAIrisks,andtoofferorganizationsincludingClimateChangeAIandtheAIforpolicymakersastreamlinedframeworkfordesirablethePlanetAlliance,andleveragesBCG’sanalysisandclientpolicyoutcomes.experienceaswellasGoogle’stechnicalandoperationalexpertise—anditsexperienceindevelopingsolutions.Throughoutthereport,weshareexamplesofsuccessfulearlyapplicationsofAIforclimateandofinstancesinwhichpolicymakershavealreadytakentheinitiativetoenable,promote,orguidetheuseofAIforclimateactionacrosssectors.1ACCELERATINGCLIMATEACTIONWITHAIExecutiveSummaryAcceleratingclimateactionisimperative,asweareWhileAIisonlyjuststartingtobeappliedtoclimateonapathtofallshortoftheParisAgreement’sgoalchallenges,leading-edgeorganizationsandusecasestokeepwarmingunder1.5°Celsius.arealreadydeliveringresults—anddemonstratingthepromiseofAIforclimate—alongthreedimensions.•TheUnitedNationsIntergovernmentalPanelonClimateChange(IPCC)estimatesthat,basedonactiontodate,•Information.AI-curatedinformationsourcesareaidingtheworldwilllikelyseewarmingof2.8°Cwithcata-nationsinshapingtheirclimatestrategy—andinre-strophicconsequences.spondingtoemergenciessuchaswildfires.•TheIPCCforecaststhatinordertomeetthe1.5°Cgoal,•Prediction.AI’spredictivepowerishelpingsavelivesbytheworldwillneedtoreduceemissions—fromthebase-offeringadvancewarningoffloods.lineof2010levels—by43%by2030.•Optimization.AIapplicationsareenablingorganiza-ByscalingcurrentlyprovenapplicationsandtionstounderstandandreducetheirScope1,2,and3technology,artificialintelligence(AI)hasthecarbonfootprints.1potentialtounlockinsightsthatcouldhelpmitigate5%to10%ofglobalgreenhousegas(GHG)emissionsAIalsoposesrisksthatmustbeconsideredandby2030—andsignificantlybolsterclimate-relatedmanagedthoughtfullytoensureitsusehasanetadaptationandresilienceinitiatives.positiveimpactonclimate.•87%ofexecutivesviewAIashavingthepotentialto•Energy-RelatedGHGEmissions.A2022paperinaddressclimateissues.NatureClimateChangeestimatesthatcloudandhyperscaledatacentersareresponsiblefor0.1%–0.2%•AI’spositiveimpactwillbemultipliedshoulditcontrib-ofglobalGHGemissionsandthatroughly25%ofdatautetoscientificbreakthroughsthatopennewpathwayscenterworkloadsarerelatedtomachinelearning(ML).forclimateaction.Yet,newerandmorecomplexAImodelsmayrequiremoreenergy.Atpresent,robustforecastsforAI’sfutureAIcancontributetoclimateactionbyreducingenergyrequirementsremainelusivegivenuncertainemissions,guidingadaptationstounavoidableadoptionratesandthebroadspectrumofpotentialclimatechangeimpacts,andprovidingfoundationaltechnicaladvancementswiththepotentialtodecreasecapabilitiesthatenableclimateaction.AI’senergyintensity.Nonetheless,AIprovidersarealreadystrivingtoenhanceenergyefficiencyand•Mitigation.Helpingwithboththereductionandremov-integratecleanenergysources.alofemissions—andwiththeunderlyingmeasurementneededtosizethechallengeandtrackprogress•WaterUse.Water-basedcoolingremainsthemostenergy-efficientoptionfordatacenters,anditsoverall•AdaptationandResilience.Aidingcountries,regions,impactonwaterconsumptionislow.In2016intheUS,cities,citizens,andbusinessesinforecastingclimate-datacenterswereestimatedtohaveusedlessthanrelatedhazards,developingplanstoaddressthem,and0.02%ofthecountry’swaterconsumptionforcooling.respondinginrealtimetocrisesNevertheless,insomecases,water-basedcoolingcanputpressureonlocalwaterresources.Datacenteroper-•FoundationalCapabilities.Enablingclimate-relatedatorshavebeguntoaddressthisissuebyprovidingmoremodeling,researchintoclimateeconomics,andnewdisclosure,exploringnewcoolingtechniques,andinvest-approachestoclimateeducationandsupportingbreak-inginreplenishmentinitiatives.throughsinfundamentalresearchBOSTONCONSULTINGGROUPGOOGLE2•Waste.WhiledatacenterscurrentlyaccountforonlyaPolicymakershaveacriticaloversightroletoplayinsmallfractionoftheworld’se-wastechallenge,thereismaximizingthebenefitsfromAI-drivenclimateanopportunityfortechfirmstobuildonearlycircularityactionwhileminimizingitsrisks.Criticalpolicysuccessesandtakeamorethoughtfulapproachembrac-outcomestopursueincludethefollowing:ingmorerecyclingandreuse.•EnablingAIforclimateprogressbyencouragingdata•OtherPotentialRisks.AIapplicationsshouldbesharing,ensuringaffordabletechnologyaccess,buildingsustainableandequitablebyintention.AIcanbeappliedawareness,andinvestingintalenttobothclimate-friendlyandclimate-unfriendlyappli-cations,cannarroworwidendisparitiesbetweenthe•AcceleratingthedeploymentofAIforclimatebydefiningGlobalNorthandtheGlobalSouth,andcanbetrainedpublicandprivatesectorpriorities,deliveringonpublicondatasetsthatreflecttheworld’sdiversity.Leaderssectorusecases,andencouragingprivatesectoractionandmodelbuildersneedtobemindfulintheirdesignchoices.•PromotingenvironmentallyandsociallyresponsibledeploymentofAI3ACCELERATINGCLIMATEACTIONWITHAIHowWeDefineArtificialIntelligenceAccordingtotheMassachusettsInstituteofTechnology,•deliveringimprovedpredictions(predictiveusecases),AIisdefinedastheabilityofcomputerstoimitatehumanandcognitivefunctionssuchaslearningandproblem-solving,usingmathandlogictosimulatetheprocessofreasoning•suggestingoptimizationmovesandrecommendationsthathelpshumanslearnfromnewinformationandmaketoreachtargets(prescriptiveusecases).decisions.ThesegoalscanbeattainedbyapplyingwiderangeofForthepurposesofthisreport,weareusingabroadertechniquesincludingthoseinthetablebelow--allofwhichdefinitionofAIthatcomprisesasetofmathematicalandweincludeinthisreport’sdefinitionofAI.computersciencetechniquesaimedatanalyzingdatatohelpunderstandandnavigatereal-worldphenomenaApplyingAItoreal-worldproblemsiscommonpracticethrough:today.Thetechnologyhasprovenitsabilitytohelppublicandprivateorganizationshaveabetterunderstandingof•providingbetterinformation(descriptiveusecases),theircontext,providebetterservices,andimprovetheiroperationalperformance.TechnologyGeneralExampleClimate-RelatedExampleAdvancedAnalyticsSupermarketInventoryEnergyConsumptionManagement.AdvancedanalyticsOptimization.AdvancedanalyticsTheuseofadvancedmathematicalandcanidentifybestsellersanddemandcanoptimizeabuilding’scarbonstatisticaltechniquestodevelopinsightsdynamics,enablingmoreefficientfootprintbyadjustingheating,fromstructuredandunstructureddata.shelvingandrestockingstrategies,cooling,andlightingsystemsintherebyreducingwasteandensuringresponsetoreal-timedatafrompopularitemsarealwaysinstock.sensorsandweatherforecasts.MachineLearningCreditCardFraudDetection.PredictingWildfires.MachineMachinelearninghelpsbanksandlearningmodelscananalyzeweatherTrainingcomputerstolearnandmakecreditcardcompaniesdetectdata,satelliteimagery,andterrainpredictionsfromdata.Historicaldataunusualtransactions,enablingtheminformationtopredictthelikelihoodconstitutestheinputs,whilepredictionstoalertcardholdersandminimizeofwildfires,helpingauthoritiestakebasedonneworunseendataarethefraudlosses.preventivemeasuresandoptimizeoutputs.resourceallocation.DeepLearningMedicalImageEvaluation.ExtremeWeatherPrediction.AppliedtotheanalysisofmedicalDeeplearningcananalyzevastAspecializedformofmachinelearningimagessuchasX-raysandMRIs,amountsofhistoricalandreal-timethatusesartificialneuralnetworkstodeeplearninghelpsdoctorsdiagnosemeteorologicalandsatellitedata,generatehierarchicalinsightsfromdiseasesandotherabnormalitiesleadingtomoreaccurateforecastsdiversedatasets,suchasimages,text,ormoreaccurately,enablingmoreforhurricanes,tornadoes,andaudio.Thesemodelsareabletorecognizetimelyandeffectivetreatments.typhoons.patternsorfeatureswithinthedata,forexample,byidentifyingobjectsinimages.LargeLanguageModelsCustomerServiceChatbots.GreenTechnologyInnovation.LargelanguagemodelsenableLargelanguagemodelscanAdvancedAImodelstrainedonvastcompaniestoautomatetheprocessaccelerateinnovationbydigestingamountsoftextdata—andabletoofansweringcustomerquestionsresearchpapersandpatentgeneratehuman-liketextasoutput,suchandhelpingtroubleshootissues,applicationsandrapidlysurfacingasforGenerativeAIusecases.enhancingtheefficiencyof,andideasandidentifyingknowledgegaps.satisfactionwith,customerserviceoperations.BOSTONCONSULTINGGROUPGOOGLE4TheClimateActionImperativeandthePromiseofAIDespitesignificantprogressoverthelastseveralyearsEveniftheworldsucceedsinlimitingwarmingto1.5°C,inmobilizingtheglobalcommunitytointensifyitstherewillstillbeadverseimpacts.Alreadytodayat1.1°C,climateactionsandambitions,theworldisnotontheIPCCreportsthatover3billionpeopleliveinareastracktomeettheParisAgreement’stargettolimittem-highlyvulnerabletoclimateimpacts.Wearealreadyperatureriseto1.5°C.Thistargetwasselectedbecauseseeingtheimpactonweather,agriculture,watersecurity,scientistsbelievethatabovethatlevel,theeffectswouldandmigration.Ifweovershootthetarget,thepicturebecatastrophicandpotentiallyirreversible.Atpresent—becomesincreasinglydire:seaswillrisefurther,droughtsbasedonupdatednationalpledgessinceCOP26in2021—willbeworse,andextremeweathereventswillbemoretheUnitedNationsEnvironmentProgrammecurrentlycommon.estimatesthatweareonapathtowarmingby2.8°C.25ACCELERATINGCLIMATEACTIONWITHAIIna1.5°Cworld,theIPCCforecaststhat48%oftheworld’sClimateActionHasanAnalyticalChallenge—populationwillbeexposedtodeadlyheatlevelsformoreandAICanHelpthan20daysayear.Ina3°to4°Cworld,thatnumberincreasesto74%.Ifwestayonourcurrenttrajectory,theLeadersincreasinglyunderstandtheurgency.Sofar,194WorldBankestimatesanadditional143millionpeople—partiestotheParisAgreementhavedevelopedNationallymorethanthecombinedpopulationsoftheUnitedKing-DeterminedContributions(NDCs)—eachrepresentingdom,Morocco,andMalaysia—couldbedisplaced.3And,detailedcommitmentsforhowtheircountrywillhelptheabsentsignificantinvestmentsinresilience,majorglobalworldmeettheParisAgreement’s1.5°Cgoal—upfrom75cities—forexample,Tokyo,Osaka,Mumbai,Bangkok,partiesinFebruary2021.4NewYork,London,andLagos—willfindthemselvespartlyunderwater.Butavoidingthemostcatastrophicimpactsofwarmingrequiresmorethanpoliticalwill.Toachieverealprogress,WeurgentlyneednewtoolstoacceleratethereductionweneedtodevelopamuchricheranalyticalunderstandingandremovalofGHGemissions—andtohelpcitizens,ofacomplexsystemcomprisingmanyvariablesandfeed-cities,regions,countries,andbusinessesmakeplanstobackloops.(SeeExhibit1.)adapttotheinevitableimpactsofwarming.AIoffersmuchpromise.Exhibit1-Climateisaninterlinked,multi-parametersystemCoreclimatecharacteristicsHumanactivities,suchasIcecapChangesinWaterEmissionshavevaryingfossilfuelburningandlandmeltingprecipitationtempera-impactsoncoreusechanges,createsigniﬁcantclimatecharacteristics,volumesofgreenhousegas.ClimatechangeprocessesCloudstureandchangesintheseprocessescanworsenHumanactivitiesCarbonSalinitythegreenhousecyclegaseﬀect.IncreaseinOceanimpermeabledisturbancecirculationupheavalsurfaces(enhanced)GreenhouseAverageGulfStreamUrbanizationtemperaturemodiﬁcationeﬀectLanduseriseAbruptchangesGlobalclimatewarmingchangeDeforestationEuropecoolingCO2SealevelriseN2OFluctuationsinCH4climatecharacteristicsdrivemajorCyclonesimpacts—natural,physical,andTransportGreenhouseFoodLossofsocioeconomic—atgasemissionsHeattraditionalbothlocalandglobalwaveslifestylesscales.DiseaseDroughtsspreadHeatingFossilfuelDisastersBiodiversityburninglossesAgricultureCasualtiesIndustryFaminesEconomiclossesMajorimpactsSource:PhilippeRekacewicz,EmmanuelleBournay,UNEP/GRID-Arendal;BCGanalysis.BOSTONCONSULTINGGROUPGOOGLE6Developingmodelsisessentialtounderstandingtherela-EstimatingAI’sPotentialContributiontionshipsamongvariables—andtoanticipatingthelikelyimpactofdifferentstrategiesandchoices.ButmodelingBasedonourresearchandexperience,thethreebroadthesecomplexinterconnectionsonalocalandglobalscaleareasinwhichAIcanaccelerateclimateprogressaretheisahugechallenge.Itrequiresassemblingmassive,longi-following:tudinal,andreal-timeglobaldatasets.Informationisneed-edonclimate(forexample,temperatures,oceanprocess-•Mitigation.Helpingwithboththereductionandremov-es,andmeteorologicalphenomena)andonhumanalofemissions—andwiththeunderlyingmeasurementactivities(forexample,emissions,andlandusechanges).neededtosizethechallengeandtrackprogressAndnotallthenecessarydataisevenavailable.•AdaptationandResilience.Aidingcitizens,countries,Butunderstandingthecomplexsystemsthatdriveregions,cities,andbusinessestoprepareforandclimate-relevantoutcomesisexactlythekindofchallengerespondtotheinevitableimpactsofawarmingplanetatwhichAIexcels.Byamalgamatingandprocessingmassivedatasets,AIcanrevealelusivepatternsand•FoundationalCapabilities.Enablingclimateactionvaluableinsights,facilitatescenariodevelopmentandviaimprovementsinclimatemodeling,climateeco-prediction,acceleratetheevaluationofmultiplecoursesnomics,andclimateeducation,aswellasacceleratingofaction,enableoperationaloptimizations,andhelpbreakthroughinnovationsthatwillopennewhorizonsmonitorprogresstowardpredefinedgoals.forclimateactionBusinessleadersagree.Ina2022BCGsurveyofseniorexecutiveswithleadershiprolesrelatedtoclimateorAI(seeAIisEssentialforSolvingtheClimateCrisis),87%viewedAIasahelpfulunlockforclimateissues.TheysawsupportingemissionsreductionasthetopclimateusecaseforAIintheirorganizations,butexpressedinterestinotherapplicationsaswell.(SeeExhibit2.)Exhibit2-LeadersbelieveAIcanplayaroleinclimateaction,especiallyinhelpingtoreduceemissionsInwhichareasofclimate-relatedadvancedanalyticsandAIdoyouseethegreatestbusinessvalueforyourorganization?(%)Reducingemissions61%Measuringemissions57%87%Predictinghazards44%Managingvulnerabilities42%ofrespondentssaythat37%AIisahelpfultooltoRemovingemissions28%Facilitatingclimateresearch,addressclimatechangeclimateeconomics,andeducationMitigationAdaptation&ResilienceFoundationalCapabilitiesSource:BCGClimateAIsurvey2022.Allrespondentshavedecision-makingauthorityoverclimateorAItopicsattheirorganizations.Respondentswerepermittedtogivemorethanoneanswer.7ACCELERATINGCLIMATEACTIONWITHAIuserid:246491,docid:149874,date:2023-12-27,sgpjbg.comRegardingemissionsreductionpotential,a2021BCGAndAIoffersmanyfoundationalcapabilitiesthatsup-study(seeReduceCarbonandCostswiththePowerofAI)portbothshort-termandlong-termclimateaction.ForestimatesthatcurrentlyprovenAI-enabledusecasescouldexample,itcansupporttoday’sclimateresearchwithreduceemissionsby5%to10%by2030.Ifthatpotentialishigher-fidelityclimatechangesimulations.Butitalsohasfullyrealized,AI-drivenapplicationswouldberesponsiblethepotentialtoacceleratebreakthroughinnovationsinforachievingroughlybetween10%and20%oftheIPCC’sdomainssuchasphysics,chemistry,biology,andmaterial2030interimemissions-reductiontargetfortheworldtosciencethatcould“bendthecurve”onclimateprogress.achievenetzeroby2050.5Similarly,aMicrosoft/PwCstudylookingatfoursectors(agriculture,energy,transport,andAllofourestimatesarebasedonthecurrentstateofAIwater)estimatesthatAIhasthepotentialtoreduceglobaltechnology—andthusspeakprimarilytoAI’spotentialinGHGemissionsby4%.6Further,respondentsinaCapgemi-currentlyprovenapplications.Today,weareintheearlynisurveyofcompaniesthathadleveragedAIforclimatestagesoftheadoptioncurve.TransformingpotentialtoactionreportedthattheireffortstodatehadachievedGHGachievementwillrequirethatallorganizationsfullyem-reductionsofbetween11.3%and14.3%dependingonthebraceAIasanessentialenableroftheirclimateactions.sector—andtheseexecutivesbelievethatAIcouldreduceAnditisimportanttonotethatourassessmentdoesnotoverallGHGemissionsby15.9%inthenextthreetofiveencompassmajorAI-drivendisruptionsandbreak-years.7throughs—forexample,newmaterialsforbatteries,newdrought-resistantcrops,novelcarbonremovaltechnolo-Onadaptationandresilience,AIcanhelpcitiesforecastgies,andscalableapproachestonuclearfusion—thattheirclimatevulnerability,developestimatesofthecostofcouldunlockmassivepositiveimpact.inaction,andmodeltheimpactofdifferentclimateinter-ventions.TheseinsightscanaidtheminidentifyingtheThepromiseofAIisreal.Whilewearealreadyseeingactionswiththegreatestbenefit,generatingprivate-sectorbenefits,weneedtoaccelerateitscontributiontoenthusiasmforfundinginvestableprojects,andsecuringplanet-savingclimateimpact.Thenextchapteroffersapublicandphilanthropicsupportforessential,butdeeperdiveintotheprimaryknownclimate-relatedusenon-bankable,adaptations.Italsocanhelpguidereal-timecasesforAI—andhighlightssomeexamplesofhowanddecision-makinginagriculture—forexample,increasingwhereAIisalreadymakingapositivedifference.cropproductionthroughintelligentirrigationsystems—orinfast-movingcrisessuchaswildfires.BOSTONCONSULTINGGROUPGOOGLE8HowAICanHelpAccelerateClimateActionAIhasdemonstratedthepotentialtoenableandAI’sRoleinEmissionsMitigationcatalyzeclimateprogressinthreebroadareas:takingemissionsmitigationtothenextlevel,shap-Gettingsmarteronreducingandremovingemissionsisingstrategiesforadaptationandresilience,andsupportingessential.AndAIisalreadydeliveringsignificantwinsthatbothclimateresearchandreinforcingtechnologies.Someneedtobescaled.ItscontributionsfallintotwobroadAIapplicationsareinearlystages,somearebeingtested,areas:measurementandmonitoring,andreductionandandothersarealreadybeingscaled.Butallwillneedtoberemoval.8embracedmorebroadlyifwearetofulfillthepromiseofAItolimitwarmingtolessthan1.5°C.MeasurementandMonitoringWithoutreliable,clean,andindependentlyverifiabledata,Exhibit3summarizesthemostpromisingofthecurrentlyeffectiveclimateactionisdifficult.Countriesandcompa-knownAIusecasesforclimate.Therestofthischapterwillniesneedtoknowtheirbaselinesandtracktheirprogress,offermoredetailoneach,alongthewayhighlightinginspir-bothatthemacrolevel(“WhatareourtotalGHGemis-ingexamplesofhowAIishelpingunlockandacceleratesions?”)andthemicrolevel(“Whichaspectsofouropera-climateprogress.tionsandbroadersupplychainarethebigdrivers?Areoureffortsatreductionorremovaldeliveringtheexpectedresults?”).9ACCELERATINGCLIMATEACTIONWITHAIExhibit3-KeyAIapplicationstoaccelerateclimateprogressMitigationAdaptationandResilienceMeasurementReductionHazardVulnerability&Monitoring&RemovalPredictionManagementMacro-levelmeasurementEnablingemissionsreductionBuildingearlywarningsystemsRespondingtocrisese.g.,calculatingcarbonfootprinte.g.,integratingrenewableenergye.g.,predictingnear-terme.g.,monitoringdroughtandatthecountrylevelintosmartgrids,optimizingextremeeventssuchasﬂooding,wildﬁrespreadtransportationofgoodsdrought,andcyclonesMicro-levelmeasurementSupportingnature-based&Projectinglong-termtrendsBuildingresilientinfrastructuree.g.,calculatingcarbontechnologicalremovale.g.,modelinglocalizedsea-level&protectingbiodiversitye.g.,intelligentirrigation,footprintsofindividualproductse.g.,assessingnaturalcarbonriseanddroughtfrequencystocksmonitoringofendangeredspeciesFoundationalCapabilitiesClimatemodelinge.g.,monitoringdroughtandwildﬁrespreadClimateeconomicse.g.,developingcostofinactionassessmentsEducation&behavioralchangee.g.,developingrecommendationsforclimate-friendlyconsumptionInnovation&breakthroughse.g.,supportingresearchonfusionSource:BCG,AIforthePlanetAlliance.EffectivemeasurementandmonitoringsolutionsleverageSolutionsareemergingformicro-levelmeasurementasAItoprocessandanalyzedatafrommultiplesourcessuchwell.Google’sEnvironmentalInsightsExplorer(EIE)usesassatellitedata,weatherdata,sensors,andotherheavymachinelearningtooffercityplannersannualestimatesdatasets—whichcan,forexample,helpanorganizationofemissionsfrombuildingsandtransportation,treedevelopabaselineforitsScope1,2,and3emissions.AIcanopystatus,andemissionsreductionopportunitiescanalsodeliverinsights,revealingpatternsinemissionssuchasthepotentialforexpandedrooftopsolar.Houston,andsuggestingthebestwaystoprioritizeabatementTexas,usedEIEtoperformasolarassessmentandinformefforts.thedevelopmentofits5millionMWhSolarEnergyTargetPlan.Similarly,CO2AI,anovelSaaSplatform,enablesInthedomainofmacro-levelmeasurement,Climatebusinessleaders—togetherwiththeirvaluechainTRACEhasbeenanearlymover.Thisnonprofitoffersfreepartners—todevelopanaccurateestimateoftheiremissionsdataformorethan80,000individualsourcesorganizations’Scope1,2,and3emissionsdowntotheandfacilitiesaroundtheglobe,providingadatafoundationproductlevel.Italsohelpsthemtomodelandevaluatetohelporganizationsgetstartedwithmitigationplans.Itsemissionsreductionopportunities.(Seethesidebardatacould,forexample,assistcountriesseekingtotransi-CO2AI:HelpingBusinessEcosystemsReducetheirtionawayfromcoalandotherfossil-fuelbasedelectricityCarbonFootprints.)generationbypinpointingthelargestemittersandreveal-ingthemixofpowersourcesbyregion.(SeethesidebarClimateTRACE:ProvidingTimely,IndependentEmissionsData—forFree.)BOSTONCONSULTINGGROUPGOOGLE10ClimateTRACE:ProvidingTimely,IndependentEmissionsData—forFreeMakingrealprogressonclimaterequirestimelyandaccu-SupportedbyGoogle.org,amongothers,ClimateTRACEratedataonemissionstoinformgovernmentpolicyandusesAIandmachinelearningtocalculateGHGemissionsbusinessaction.Buthistorically,emissionsdatahasbeenonaglobalscale,withthegoalofmovingtowardreal-timebasedonself-reporting,calculatedusingvaryingalgo-precision.Toachievethis,itsmodelanalyzesmorethan59rithms,andsubmittedyearsafterthefact.Climateterabytesofdatafromover300satellitesandmorethanTRACE—aglobalcoalitionofnonprofits,techstartups,and11,000sensorstocreatehighlygranularemissionsdataforresearchers—offersapowerful,free,andindependentover80,000sourcesglobally.Thatnumberisexpectedtoalternative:thefirstcomprehensivesource-levelglobalgrowtomorethan70millionsourcesbytheendof2023.inventoryofGHGemissions.Applicationarea:Macro-LevelMeasurementClimateTRACEtracksglobalemissionsSource:ClimateTRACE.Usedwithpermission.ACCELERATINGCLIMATEACTIONWITHAI11CO2AI:HelpingBusinessEcosystemsReducetheirCarbonFootprintsInordertomakerealprogressondecarbonization,organi-Inoneexample,aglobalhealthcarecompanyseekingtozationsneedamoregranularandactionableviewoftheirreduceitsScope3emissionsby20%by2030embracedcarbonfootprints,bothacrosstheirScope1,2,and3emis-CO2AI.Theplatformenabledittoincorporate50timessionsandatthelevelofindividualproductareas.Untilmorefactorsintoitscalculationsandtodevelopannow,thatkindofsinglesourceoftruthhasnotbeenavail-activitybasedemissionsbaselinethatwas20%moreabletohelpoperationsleadersunderstandemissionshotprecise.AndCO2AI’ssimulationandroadmappingtoolsspotsandexplorepotentialsolutions.enabledittoidentifydecarbonizationopportunitiesthatwoulddeliver120%ofitsemissionreductiontarget.CO2AI,aninnovativeSaaSplatform,helpsorganizationsseamlesslymapemissionsacrosstheirvaluechainsandApplicationarea:Micro-LevelMeasurementleveragethoseinsightstodriveclimateaction.AIplaysacentralroleinbothassemblingemissionsdataandmatch-ingittoactivitiesandproducts—andinsimulatingsolu-tionsandbuildingdecarbonizationroadmaps.MeasuringandmanagingemissionswithCO2AISource:CO2AI.Usedwithpermission.BOSTONCONSULTINGGROUPGOOGLE12ReductionandRemovalIntherealmofagriculture,theintegrationofAItoolswithAIhasthepotentialtoaidorganizationsinreducingandtechnologiessuchasdronescanhelpfarmersmonitorremovingemissionsintwoways:enablingemissionstheircropsinrealtimeforbetterfieldmanagement,thusreductionandsupportingnature-basedandtechnology-enhancingagriculturalproductivitywhileminimizingGHGbasedcarbonremoval.emissions.Moreover,AI-drivenprecisionfarminghelpsempowerfarmerstomakewell-informed,data-drivenEnablingEmissionsReduction.AIcancontributetothedecisionsregardingfarmingpractices,cropselection,creationofmoreefficientandcleanerenergysystemsinirrigation,fertilizing,pestmanagement,andharvesting.multipleways.Itcan,byconsolidatinginformationfromThisapproachstreamlinesresourceutilizationand,ifdonedozensofdifferentorganizationsandgridcomponents,purposefully,canminimizetheenvironmentalimpactprovideinsightsonhowtooptimizeelectricgridopera-associatedwithfarmingpractices.Forexample,Alphabet’stions—andsupportinformeddecision-makingongridprojectMineralisusingrobotics,AI,andcomputervisionplanning.Itcanalsohelpspeedtransitionfromfossilfuelstocreateamoresustainablefoodproductionsystem.Itistoalternativeenergysourcesthroughbettersupplyanddevelopingperception-poweredsolutionswithpartnersdemandforecaststhatreducetheneedforbatterystorageacrosstheagriculturevaluechain—fromgroceryretailersandstandbypowerandenablemoreefficientreal-timeandenterprisefarmstoequipmentmanufacturersandbalancingofelectricgrids.cropprotectioncompanies—todevelopabetterunder-standingofthecomplexinteractionsofplants,theirgrow-Forexample,Tapestry,anAlphabetproject,iscreatingaingenvironments,andfarmmanagementpractices.12singlevirtualizedviewoftheelectricitysystemwiththegoalofloweringemissions,minimizingoutages,shorteningAnotherinterestingusecaseinvolvesusingAItoreduceinterconnectionqueues,andintegratingmorerenewablescontrails.Contrails,thewhitecloudsthatsometimesformintothegrid.AIisattheheartofitscomputationalandbehindairplaneswhentheyfly,areresponsibleforaboutsimulationtools.Relatedly,onthesubjectofrenewables,35%oftheaviationsectors’globalwarmingimpact.AIFrance’sEngiehaspartneredwithGoogleCloudtodevelopsolutionsdevelopedbyGoogleResearchinpartnershipandpilotanAI-poweredtoolthatcanprovidegridopera-withBreakthroughEnergyhaveenabledairlinepilotsintorswithmoreaccuratereal-timeforecastsofwindenergytrialstudiestoreducecontrailsbyupto54%.13(Seethesupplies.9sidebarTheContrailsImpactTaskForce:AddressingAvia-tion’sOtherContributiontoWarming.)InAfricaandIndia,HuskPowerSystemsprovides“pay-as-you-go”100%renewablepowertooff-gridandweak-gridSupportingNature-BasedandTechnology-Basedcommunitiesthatis30%cheaperthanthealternative:Removal.AccordingtotheIPCC,limitingwarmingtodieselgeneration.HuskestimatesthatitsAImodelen-1.5°Cby2100willrequireanextensivedeploymentofCO2ablesittopredictuserdemandwith80%accuracyacrossremovalmeasures,ofwhichtherearetwobroadtypes:itsmicrogrids,therebyimprovingcapacityutilization,re-nature-basedremovalinwhichcarbonisremovedbyandducingcosts,deliveringlowerprices,andguidingcapitalstoredinnaturalsinkssuchasforests,algae,andwetlands,investmentinadditionalcapacity.andtechnology-basedremovalviaapproachessuchasdirectaircapture(DAC).14AIcanplayasupportingroleinMoreover,AI-driveninsightscanalsoenablepeopleandbothtypesofremoval.organizationstomakesmarterdecisionsthatdecreaseemissions.Forinstance,asaresultofusingAItoimproveInnature-basedremoval,AI-basedsolutionscanhelpdemandforecasting,manufacturerscanavoidbothover-quantifyandverifythelevelofcarbonsequestrationproductionandtheemissionsthoseunsoldgoodswouldachievedinanecosystem,enablingpublicandprivateproduce.Similarly,AI-optimizedtransportationcanreducesectorleaderstomakeinformeddecisionsregardingtheemissionsbyidentifyinganddirectingdriverstothemostdeploymentofnaturalsolutions,includinglandmanage-efficientroutes.AsofSeptember2023,GoogleMaps’mentandreforestationefforts.Oneactorinthisspaceisfuel-efficientroutingfeaturewasestimatedtohavehelpedAlboClimate.Usingmyriadremotesensingimageryandpreventmorethan2.4millionmetrictonsofCO2eemis-proprietaryAIalgorithms,AlboClimatemonitorsandsionssinceitslaunchinOctober2021—equivalenttoquantifiesenvironmentalmetricssuchasabove-andtakingapproximately500,000fuel-basedcarsofftheroadbelow-groundcarbonsequestrationandland-usedynamicsforanentireyear.10Andbetween2011and2022,Google’sinforestryandagriculturalecosystemsandprovidestrans-Nestthermostatsareestimatedtohavehelpedcustomersparentandreliabledatatovariousstakeholdersofthecumulativelysave113billionkWhofenergy—theroughnature-basedmarkets.equivalentofdoublePortugal’sannualelectricityuse—byproposingthermostatadjustmentsbasedoncustomerbehavior,suchasautomaticallyadjustingtemperatureswhencustomersareawayfromhome.1113ACCELERATINGCLIMATEACTIONWITHAITheContrailsImpactTaskForce:AddressingAviation’sOtherContributiontoWarmingAviationisoneofthehardestsectorstodecarbonize.Thetrialreducedcontrailsby54%across70liveAccordingtotheIPCC,contrails—thethin,whitelinesyouAmericanAirlinesflights.Aspartoftheinitiative,Americansometimesseebehindflyingaircraft—accountforroughlyAirlinesintegratedcontrail-likelyzonesintothetablets35%ofaviation’sglobalwarmingimpact.theirpilotsused,enablingthemtomakein-flightaltitudeadjustmentstoavoidcreatingcontrails,justastheydotoContrailsarecreatedincertainatmosphericconditionsavoidturbulence.thatenablewaterdropletstocondenseandfreezearoundthesootparticlesfromjetengineexhaust.SomedissipateInOctober2023,anewpartnershipwasannouncedwithquickly,whileothersformintopersistentcontrail-cirrusEUROCONTROL’sairtrafficcontrolcenterthatmanagescloudsthatcanlastforhours,trappingheatthatwouldtheairspaceoverBelgium,theNetherlands,Luxembourg,otherwiseescapeintospace.andnorthwestGermany—oneofthebusiestairspacesintheworld.Withthispartnership,EUROCONTROLwillbeTomitigatethis,GoogleResearchteamedupwithAmeri-abletoprovideaircraftflyingthroughitsairspacewithcanAirlinesandBreakthroughEnergytocombineAIandinformationabouthowtoavoidmakingcontrails.15hugeamountsofdatatopredictwherecontrailswillformandhowplanescanavoidmakingthem.Applicationarea:EnablingEmissionsReductionContrailsdetectedovertheUnitedStatesSource:HowAIishelpingairlinesmitigatetheclimateimpactofcontrails,GoogleBlog,August2023.Usedwithpermission.BOSTONCONSULTINGGROUPGOOGLE14Inthesphereoftechnology-basedremoval,advancedProjectingLong-TermTrends.TheimpactsofclimatetechnologiessuchasDACcanfilterandcaptureCO2fromchangevaryacrossregionsandcitiesduetothecomplextheairasitpassesthroughadevice.ThecapturedCO2interplayoflocalgeography,weatherpatterns,oceancur-canthenbestoredundergroundin,forexample,salinerents,andothervariables.AIcanplayapivotalroleinaquifers—orpreparedforindustrialapplications.Thisdevelopingpowerfulclimatemodelsthatcananticipatetechnologycurrentlyfacesquestionsaboutitsenergyadverseimpactssuchasrisingsealevelsanddrought—efficiency,whichmayhinderscalability,butongoingre-andassessingtheirimplicationsforlocalcommunitiesonsearchanddevelopmenteffortsmayresolvethesechal-factorssuchaseconomicdevelopment,infrastructure,andlenges.Anotherapproachisbioenergywithcarboncaptureagriculturalandfishingoutput.Theseinsightsenabletheandstorage,whichgeneratesenergyfrombiomasssuchdevelopmentofthoughtfulresiliencestrategiestomitigateaswoodandagriculturalwasteandcapturestheresultingtheeffectsofclimatechange.CO2forundergroundstorageorindustrialuse.AIcanplayaroleinassessingoptimalcaptureandstoragelocations,VulnerabilityManagementmonitoringforpotentialleaks,andoptimizingthePredictionisundeniablyacrucialaspectofpreparedness;industrialprocessesandmaterialsusedincarboncapture.however,itshouldbecomplementedbyproactiveeffortsatthelocalleveltofortifycommunitiesagainstbothsuddenAIasanEnablerofAdaptationandResilienceandprotractedcrises.AIisshowingsignificantpromiseinhelpingmanagecrisesandbuildforwardresilienceforEvenifmitigationeffortsenableustoachievetheParisphysicalinfrastructureandlivingcreatures.Agreementgoaloflimitingwarmingto1.5°C,communitieswillneedtoadaptandbuildresiliencetotheconsequencesRespondingtoCrises.Whenacrisisstrikes,itisincredi-ofawarmingplanet.AI-drivensolutionscanplayacriticalblyvaluabletohavetoolsthathelptoensuretherightroleintwospecificareas:predictinghazardsandmanag-resourcesareallocatedtotherighttasksandintherightingvulnerability.locations.AIcanhelpcrisismanagersbycombiningdatafromscatteredsourcesandofferingaconsolidated,HazardPredictionreal-timeviewoffactsontheground.OneexampleisSomeclimateimpacts,suchassea-levelrise,areslowGoogle’sWildfireBoundariestrackerthatusessatellitemoving.Others,suchasflooding,arefast.TheincidenceimageryandAImodelstodetectthelocationofwildfireofextremeweatherevents—forexample,heatwaves,heavyboundariesinrealtimeandthendisplaytheminSearchprecipitation,droughts,andseverestorms—isincreasing.andMapstosupportbothrespondersandresidentsinACDPanalysisestimatesthatS&P500companiesfaceamakinginformeddecisions.17potential$40billionto$50billionimpactfromphysicalclimaterisksby2026.16Inlightofthat,governmentandAnotherexampleisinhelpinggovernmentsandNGOsbusinessleadersneedtodoubledownonpreparation—toprepareandprovidesupportandsheltertopeopleandAIcanmakeasignificantdifferenceintwoareas:displacedbyextremeweatherevents.(Seethesidebarenablingearlywarningandprojectinglong-termtrends.HelpingUNHCRGetAheadofForcedDisplacementinSomalia.)BuildingEarly-WarningSystems.AIcansavelivesandminimizepropertydamagebypredictingdevastatingBuildingResilientInfrastructureandProtectingweathereventsandgivinggovernmentsandpeopletimeBiodiversity.AIcanalsosupportriskassessmentandtoprepare.Riverinefloodsofferanexample.Extremerainsremediationplanningforcriticalinfrastructure,helpingtaketimetoflowdownriverandthattimecanbeputtolocalitiesmodelvulnerabilities,prioritizeresilience-build-goodusetakingactiontomitigatetheirimpactoncommu-inginvestments,andpressure-testplans.InIndia,Googlenitiesalongthebanks.FloodHub,aninitiativeofGoogleishelpingthegovernmentaddressfoodandwatersecurityResearch,enablesgovernments,aidorganizations,andchallengesbyinterpretingsatellitedatatotrackfarmindividualstotaketimelyactionandprepareforriverineboundaries,measureforestandwoodlandacreage,andfloodsvialocallyrelevantflooddataandforecastsuptoidentifyimprovementstoirrigationstructurestopreparesevendaysinadvance.(SeethesidebarProvidingEarlyfordroughts.18WarningofRisingRiverLevelswithFloodHub.)15ACCELERATINGCLIMATEACTIONWITHAIProvidingEarlyWarningofRisingRiverLevelswithFloodHubAnnually,floodscausethousandsoffatalitiesworldwide,Real-timefloodforecastsandvisualizationsareavailableondisruptthelivesofmillions,andleadtosignificanttheFloodHubplatformand,inmanycases,alsoonSearchfinancialcosts.TheyareoneofthedeadliestnaturalandMaps.Asof2023,FloodHubcoversmorethan80disasters,andclimatechangeisincreasingtheircountries,providingforecastsuptosevendaysinadvance.Itfrequencyandseverity.offersalertsforgeographiesacrossAfrica,Europe,SouthandCentralAmerica,andtheAsia-Pacificregionthatcom-Betterpredictionofimpendingfloodinghasthepotentialbinedarehometo460millionpeople.InOctober2023,tosavelivesandmitigatetheextentofpropertydamage.FloodHubexpandedtotheUSandCanada,coveringmoreFloodHub—poweredbyAImodelsdevelopedbyGooglethan800riverbankswhereover12millionpeoplelive.TheResearch—aimstopredictwhenandwhereriverineflood-goalistobringfloodforecastingtoeverycountryandtoingwilloccurinordertoprovidetimelywarningstogovern-includemoretypesoffloodsthroughongoingcollaborationsments,organizations,andthepeoplelikelytobeaffected,withgovernments,communities,academics,andorganiza-empoweringthemtoactbeforethefloodstrikes.tionssuchastheWorldMeteorologicalOrganization.19Applicationarea:EarlyWarningSystemsFloodHub’sheatmapoffloodriskSource:Google.Usedwithpermission.BOSTONCONSULTINGGROUPGOOGLE16HelpingUNHCRGetAheadofForcedDisplacementinSomaliaMillionsofpeopleinSomaliafacedisplacementasaItpartneredwithOmdena,aglobal,crowdsourcedcommu-consequenceofclimate-drivennaturaldisasters,resourcenityofAIexperts,todevelopAIsolutionsthatcouldpredictshortages,andviolentconflict.In2022alone,climate-displacementamonthinadvance.Omdenadevelopeddrivenevents—drought,floods,wildfires,andstorms—machinelearningmodelsthathelpedpredictareasfordisplaced1.2millionSomalis,representingnearlytwo-interventionbasedonidentifiedconflict“hotzones”com-thirdsofallinternaldisplacementsfortheyear.Onpointbinedwithdroughtandagriculturalproductionmetrics.tosupportthoserefugees,UNHCR,theUnitedNations’TheinsightsfromthesemodelsareenablingUNHCRtorefugeeagency,wantedabetterwaytoforecastwhereoptimizetheassignmentsofitspersonnelanddeploymentandwhentodeployitsresources.ofresources.Applicationarea:AdaptationandResilience17ACCELERATINGCLIMATEACTIONWITHAIMoreover,AIcanhelppeopleunderstandandmitigatetheEducationandBehavioralChange.AIcanhelpinflu-impactsofawarmingplanetonagriculture,fisheries,andenceclimate-friendlybehavior.Afterall,peoplemaywishthebroadernaturalworld.Inagriculture,itcancontroltomakeclimate-friendlychoicesbutlacktheinformationintelligentirrigationsystems,identifyearlysignsofcroptoknow,forexample,therelativecarbonfootprintoftwodiseases,predictfutureyieldsbasedoncurrenttrends,andpairsofjeans.Byprovidingrelevantinformation,AIcanpromoteknowledgetransferacrossbiomes.Itcanalsohelpconsumersmakeenvironmentallyconsciouschoices.helpprotectwilderplaces.GlobalForestWatch,forexam-Forexample,GoogleMapsusesAItosuggestfuel-efficientple,usesAIandsatelliteimagerytocreateareal-timetoolroutesthathavefewerhills,lesstraffic,andconstanttomonitorandcombatdeforestation.speedswiththesameorsimilarETA.Thefeatureisavail-ableintheUS,Canada,Europe,andEgyptandwillbeAI’sFoundationalCapabilitiestoSupportrollingoutinIndiaandIndonesiaduring2023.InIndiaandClimateActionIndonesia,thefeaturewillbeexpandedtotwo-wheelers,helpingevenmorepeopletotravelmoresustainably.20AIcanalsohelpturbochargefoundationalcapabilitiesthatareessentialtoshapinganeffectiveresponsetotheInnovationandBreakthroughs.AIcanalsoaccelerateclimatecrisis.breakthroughinnovationindomainsthatcouldopennewfrontiersinthebattleagainstclimatechange.Forinstance,ClimateModeling.AIcanstrengthenclimatemodelsbyinthefieldofmaterialscience,AIhasalreadyaidedthefillingindatagaps,enablingtheincorporationofadditionaldiscoveryofanewfamilyofsolid-statematerialsthatvariables,andnavigatingdatasetstoolargeforhumanconductlithium.Thesesolidelectrolyteswillbehelpfulinanalysis.Itcanalsoyieldmoreaccurateestimatesbythedevelopmentofsolid-statebatteriesthatofferlongermodelingmultidimensionalcomplexsystemsandtherangesandincreasedsafetyforelectricvehicles.Inthefeedbackloopsamong,forexample,climateandsocioeco-fieldofcleanenergy,GoogleDeepMindhaspioneeredanomicvariables.ThegreateraccuracyandprecisionofthedeepreinforcementlearningsystemthathelpsresearchersAImodelsinturnenhancestheimpactofAI-supportedbettercontrolnuclearfusionplasma,openingnewavenuesclimateapplications.toadvancefusionresearchandthuscleanenergyalterna-tives.(SeethesidebarAcceleratingFusionSciencethroughClimateEconomics.AIcanimproveestimatesoftheBetterPlasmaControl.)financialimplicationsofclimate-relatedimpactsandre-sponsemeasures,enablingpolicymakersandprivate-AIisclearlypoisedtoplayapivotalroleinshapingsectorleaderstobetterunderstandwhichinvestmentspositiveclimateoutcomes.But,alongsideitsbenefits,todayarelikelytoyieldthegreatestbenefit.Forexample,itiscrucialtoalsoacknowledgethepotentialrisksassoci-governmentleadersinLagos,Nigeria—oneoftheAfricanatedwithitsimplementation.Thenextchapterofferscontinent’smostvulnerablecities—leveragedAItomodelsomeperspective.thepotentialfinancialandsocioeconomicimpactsofrisingsealevelsandevaluatethoseinvestmentswiththehighestimpactonadaptationandresilience.(SeethesidebarHelpingLagosShapeItsStrategyforClimate-relatedAdaptationandResilience.)AnotherexampleisSprout,aninsurtechstartupthatprovidescoffeefarmerswithcropinsurancebasedonweatherfluctuations.(SeethesidebarSprout:HelpingSmallholderCoffeeFarmersNavigateClimateChange.)BOSTONCONSULTINGGROUPGOOGLE18HelpingLagosShapeItsStrategyforClimate-RelatedAdaptationandResilienceLagosisoneofAfrica’smostpopulousurbanareas.ItisTheseAI-poweredtoolsthenenabledLagos’sleaderstoalsohometothecontinent’sfourth-busiestportandcon-simulatetherelativecontributionofdifferentstrategiestributesaround30%ofNigeria’sGDP.Atanaverageeleva-undermultiplescenarios.Italsohelpedestimatethecosttionofonly1.5metersabovesealevel—andwithmuchofofinaction,which,atabout$30billion,represents12timesthecityatorbelowsealevel—itisalsooneofthecitiesthebudgetoftheLagosmetropolitanarea.mostendangeredbyrisingsealevels.Byutilizingthesetools,policymakershavecuratedapriori-ButLagoshassetitselfonajourneytobuildamoreresil-tizedportfolioofprojectsspanningthreebroadareas.Oneientfuture,leveragingdataandanalyticstobuildarobustoftheseareasconcentratesonenhancinginfrastructureadaptationandresilienceplanthataddressestheimpactresilience,exemplifiedbyinitiativessuchastheconstruc-ofextremeweatheronitskeysystems.Theplanisguidingtionofan18kmembankmentandseawallsdesignedtothecitytomobilizefinancingandimplementeffectivesafeguardmorethan2.7millionindividuals,includinggovernance,legislation,andmonitoringcapabilities.700,000fromvulnerablepopulations.AnotherrevolvesaroundbolsteringcommunityresilienceandsafeguardingAI-enabledtoolshaveplayedacentralroleinguidingvulnerablegroups,withinitiativessuchastheestablish-Lagos’sjourney—first,inriskassessmentandtheninmentofinfectiousdiseasesurveillancesystems.Andtheshapingandprioritizingstrategies.Lagos’sriskassessmentthirdfactorcentersonproactivelyaddressingrisksandwasbasedmainlyonfourkeyindicators:floodintensity,enhancingcrisisresponsecapabilities,suchastheestab-capitalcosttodamagedinfrastructure,impactofextremelishmentofwell-definedpost-disasterprocedures.eventsonGDP,andthenumberofcitizensaffected.Thedataiscalibratedinincrementsofonly500squareThecityhasactivelyengagedwithadiversearrayofstake-meters—smallenoughtoshowfine-grainedpatternsofholdersfromthepublic,private,andsocialsectors.Cur-climatevulnerability.Overall,themodelestimated165rently,itisintheprocessofimplementingitsClimatesquarekilometerswouldlikelybeinundated,700,000AdaptationandResiliencePlanwhilesecuringfundingresidentswouldneedtoberelocated,andtherewouldfromvarioussourcesincludingtheprivatesector,publicbe$5billionindamagetotransport,communications,sector,NGOs,andothers.andpowerinfrastructure.Theheatmapbelowshowsthemodel’spredictionofinundationrisk,withtheredareasApplicationarea:ClimateModeling,ClimateEconomics,themostvulnerable.andAdaptationandResilienceInundationriskheatmapforLagosSource:BCGClimateImpactAIPlatform.ACCELERATINGCLIMATEACTIONWITHAI19Sprout:HelpingSmallholderCoffeeFarmersNavigateClimateChangeCoffeeproductionfacesasignificantthreatfromtheFirst,viatheSproutmobileapp,itofferslocallycustomizedextremetemperaturesandunpredictablerainfallthatareagronomyadvicetofarmersonhowbesttonavigateunex-aresultofclimatechange.Smallfarmsareparticularlypectedweathertrends.Thisreducesintra-seasonalrisks,vulnerable.Andmanyareseeingdeclinesinyieldsoftherebyrequiringlessinsurancecover.Second,viaanupto15%asaconsequence—yetfewhaveaccesstoorinnovativeindexinsuranceoffering,Sprouthelpsalleviatecanaffordcropinsurance.downsideimpacts.Premiumsarepaidbythefarmers’directcustomers—andultimatelybydeveloped-worldSprout,aninsurtechstartup,offersaninnovative,AI-drivencoffeeconsumers.Indexinsurancedoesn’trequirefarmerssolutionthathelpsfarmersnavigaterisksandwithstandtofileclaimstoreceivecompensation.Instead,itprovidesfullorpartialcropfailuresresultingfromextremeclimateautomaticpayoutsshouldanindexvariableforthelocalconditions.Sproutbringsitsownproprietarydataoncoffeearea(forexample,rainfall)fallbelowatargetvalue.SproutfarmingtogetherwithdatafromsatellitesandotherispilotingitsClimateSmartCoffeeTMprogramwithsupportsourcestodotwothingsthatsupportsmallholderfarmers.fromUSAIDDevelopmentInnovationVenturesinKenya.Sproutaspirestoofferprotectiontoover1millionfarmersworldwideby2030.Applicationarea:ClimateEconomicsSprout’sAI-enabledcropinsuranceofferingSource:Sprout.Usedwithpermission.BOSTONCONSULTINGGROUPGOOGLE20AcceleratingFusionSciencethroughPlasmaControlSimulationNuclearfusionhasthepotentialtobeasourceofabun-AcollaborationofGoogleDeepMindandtheSwissPlasmadant,cleanenergy.ItisthereactionthatpowersthestarsCenteratEPFL,aSwissresearchuniversity,hasleveragedoftheuniverse.ButhugebreakthroughswillberequiredAItocreatethefirstdeepreinforcementlearningsystemforittobecomecosteffectiveandscalable.Acriticalforfusionresearch.ItsimulatesEPFL’sVariableConfigura-aspectoffusionresearchinvolveslearninghowtocontroltionTokamak(TCV)andhassuccessfullymodeledwaystoandsustainahydrogen“plasma”thatishotterthanthestabilizeandsculptplasmathathavesubsequentlybeencoreofthesun.Onetoolresearchersuseisatokamak,avalidatedintheactualTCV—openingnewavenuestodoughnut-shapedvacuumthataimstocontaintheadvancenuclearfusionresearch.plasmabymakingthousandsofadjustmentspersecondtoasetofpowerfulmagneticcoils.ThemagnetsseektoApplicationarea:InnovationandBreakthroughskeeptheplasmafromtouchingthevesselwalls,whichwould,ataminimum,dissipateheat,orworse,damagethetokamak.Sincetheworld’stokamaksareinhighdemand,onewaytoadvanceandacceleratefusionresearchisthroughsimulation.21ACCELERATINGCLIMATEACTIONWITHAINavigatingAI’sPotentialRisksAsgovernmentsandbusinessesincreasinglyrelyonIssue1:AI’sEnergy-RelatedGHGEmissionsAItohelpmitigateemissions,buildresilience,andadapttoachangingclimate,acriticalquestionIn2022,globaldatacenterelectricityconsumptionac-arises:whatrisksareassociatedwiththeriseofAI?Under-countedfor1.0%to1.3%ofglobalfinalelectricityde-standingandnavigatingtheserisksisessentialifwearetomand.21Further,a2022paperinNatureClimateChangescaleAIresponsiblyandmanageitsenvironmentalfoot-estimatesthatcloudandhyperscaledatacentersareprint.GHGemissions,water,andwastemanagementareresponsiblefor0.1%to0.2%ofglobalGHGemissionsandthreekeyareasthatwillbeaddressedinthischapter.thatroughly25%oftheirworkloadsarerelatedtomachinelearning.22Acriticalquestionishow—withAIatthestartofanewinnovationandadoptioncurve—datacenterelectricityuseandrelatedGHGemissionswillevolvegoingforward.Thereisacriticalneedformoredeepresearchonthistopic,butatthisstage,wecanmakesomeimportantobservations.BOSTONCONSULTINGGROUPGOOGLEHistorically,datacenterenergyconsumptionhasgrownSohowmightthesefactorsevolve?A2022paperbymuchmoreslowlythandemandforcomputingpower.researchersfromUCBerkeleyandGoogle,whichstudiedBetween2015and2018,forexample,theIEAreportsthattheenergyrequirementsformachinelearningtraining,datacenterelectricityusewasflatdespiteadoublingofidentifiesfourbestpracticeswiththepotentialtoreducecomputedemandandtriplingofinternettraffic.Further-energyuseandemissions.25Let’stakeeachinturn:more,apaperinSciencereportsthatbetween2010and2018—aperiodduringwhichdatacentercomputede-•EnergySourceCarbonIntensity.Indatacentermandincreasedsixfoldandinternettraffictenfold—dataemissionsprofiles,suchasinrealestate,locationcenterelectricityusegrewjust6%asaresultofashiftmatters.In2022,Norway’selectricgridhadanaveragefrominefficienton-premisesdatacenterstothehighlycarbonintensityof29gCO2e/kWh,comparedwithanenergy-optimizedcloud.23Andwestillhaveoptimizationaverageof102inBrazil,367intheUS,489inSingapore,potentialinthefuture.In2022,theaverageannualpowerand709inSouthAfrica.26Andevenwithinacountry,theusageeffectivenessforGoogle’sglobalfleetofdatacentersshareofcarbon-freeenergycanvarysignificantlyfromwas1.10,comparedwiththeindustryaverageof1.55.24regiontoregion—meaningthatmanyorganizationsoperateinjurisdictionswithouteasyaccesstocleanLookingforward,Exhibit4illustratesthekeydriversofAI’senergy.AndwhiletradeassociationssuchastheCleanelectricityuseandGHGemissionsthroughoutitslifecycle.EnergyBuyersAssociation,RE-Source,andtheAsiaAtthelevelofthegrid,emissionswillbedeterminedbyCleanEnergyCoalitionareworkingtoincreaseaccess,thepowersourcesselectedtopoweradatacenterandthechangesarestillneeded.efficiencyofthedistributionsystem.ComputemachinesanddatacenterdesignchoicesalsocanhaveapowerfulOtherchoicesmattertoo.Informationandcommunica-influence.Thechoiceofmoreenergy-efficientserverstionstechnology(ICT)companieshaveledtheworldinoptimizedforAImodelscanmakeasignificantdifference.adoptingrenewableenergypowerpurchaseagreementsAImodeldeveloperscanbemoremindfuloftheenergy(PPAs),withAmazon,Microsoft,Meta,andGoogleintensityoftheirprogrammingchoices.Andacriticalhavingsignedrenewableenergyagreementstotalingunknownishowquicklyandhowmuchend-userdemandalmost50GWofcleanenergygenerationcapacityforAI-enabledproductsandservicesincreasescomputethrough2022,equaltothegenerationcapacityofdemand.Sweden.27Andmanytechleadersaregoingfurther.Google,forexample,hasatargettorunoncarbon-freeenergyineverygridwhereitoperatesby2030,isalreadyprocuringenergyfromgeothermalpowerplantsandbattery-basedbackuppowersystems,andispilotingtheuseofAIalgorithmstobetterpredictwindproductionandfacilitateitsintegrationintothegrid.28Exhibit4-CriticalfactorsdeterminingAI’semissionsfootprintAI-relatedPowerEnergyDatacenterDevelopersEnd-userenergyﬂowplantsdistributionoperationsbuildingdemandforAImodelsAI-enabledAI-relatedEnergysourceDataemissionscarbonintensitycenterModelproductsinfrastructuredevelopmentAIadoption&coding&usageSource:BCGanalysis.23ACCELERATINGCLIMATEACTIONWITHAI•DataCenterInfrastructure.Datacenterenergyneeds•AdoptionandUsage.AIuseisforecasttogrow.Inanaredrivenbyserverandinfrastructurechoices.ServersIBMGlobalsurvey,35%ofcompaniesreportedalreadyspecificallydesignedformachinelearningcanrunAIusingAIintheirbusinessin2021,withanadditionalmodelsfasterwhileusinglesspower.Butserversaren’t42%statingthattheyarepilotingAIapplicationsforlat-theonlysourceofenergydemand.Additionalpoweriseradoption.33SomeofthegrowthinAIadoptionmightneededtooperatethefacility—forexample,forcoolingreplaceexistingworkloadsintoday’sdatacenters,butandlighting.Andspecificdatacenterequipmentchoicesgiventheexpansionofusecases,aggregatedemandwillmatter.AstudybyresearchersatBerkeleyandGooglegrow.Atpresent,however,thereisnorecent,peer-findsthatGoogleowned,designed,andoperatedcloudreviewedresearchthatforecastsfutureAIworkloadsanddatacenters,forexample,canbe~1.4–2.0timesmoreenergyneeds.energyefficientthantraditionaldatacenters,andhard-warespecificallydesignedtosupportAIandmachineAnyAImodelhastwophasestoitslifecycle:traininglearningcanbe~2–5Xmoreefficientthanoff-the-shelfandinference.Traininggivesthemodelitssmartsandsystems.29Google’smachine-learningoptimizedTensorhappenssporadically,whileinferenceistheproductionProcessingUnits(TPUs)areanexample:TPUversion4phaseinwhichthemodelisusedbycustomers.Infer-hasproventobeoneofthefastest,mostefficient,andenceworkloadsarethereforedrivenbycustomeradop-mostsustainablemachinelearninginfrastructurehubstionandusage.Andtoday,inferenceprocessingaccountsintheworldwiththepotentialtogenerate93%fewerfor80%–90%ofAIandmachinelearningworkloads.34emissionscomparedwithunoptimizedserversusingP100GPUs.30ThesefactorsaresynthesizedinExhibit5,whichmakesacriticalpoint:whilethereisuncertaintyonfutureAIHowever,newprocessorsrequirenewmanufacturingemissions,ourdecisionsmatter.Wearenotsuggestingpracticesandthereforeattentionmustbepaidtotheirthattheseareequallyimportant,onlymakingthepointembodiedcarbon.Futurehardwaredesignshouldlookthatalreadytodaytherearetechnology,architecture,andatoptimizingfulllifecycle,insteadofjustoperational,locationoptionsthatcansignificantlymitigateAI’sGHGGHGemissions.Andrelatedly,moreeffortisneededimpact.fromthesemiconductorindustrytounderstandandreduceembodiedemissions.And,beyondtheseactivemeasures,thereareotherfactorsthatwillinfluencethespeedandscaleofAIdeployment•ModelDevelopmentandCoding.State-of-the-art(forexample,theeconomicsofdatacenterdevelopment).approachestodevelopingAImodelsarevariedandDatacenterscantakeyearstobuildandrequiresignificantevolving,butonethingisclear:thedesireformorecapitalinvestment.preciseandaccuratemodeloutputshasbeenleadingtomorecomplexmodelsthatrelyonlargersetsofIssue2:AI’sWaterUsetrainingdataandrequiremoreprocessingpower.31ThesemorecomplexmodelsmayleadtohigherDatacentersgenerateinsight,buttheyalsogenerateheat.energyconsumption,allotherfactorsbeingequal.Thisheatmustbedissipatedtoprotecttheservers,com-Nonetheless,itisimportanttonotethatAImodelmunicationequipment,andstoragedevicestheycontain.designisanevolvingfield,andnewreleasesandForwarehouse-scaledatacenters,water-basedcoolingisversionsofcomplexmodelsconsistentlydemonstratethemostcommonapproach.Andamongwater-coolingimprovedenergyefficiencywhilemaintainingmodelsolutions,evaporativecooling—inwhichwaterabsorbsperformance.Indeed,ongoingimprovementsinambientheatthroughevaporation—isfrequentlybothsoftwareandalgorithmicoptimizationhavethethecheapestandthemostenergyefficient.Googlefoundpotentialtosignificantlyenhanceefficiencyandthatitswater-cooleddatacentersuseabout10%lessdecreasecomputationalrequirements.Oneexampleenergythanitsair-cooleddatacenters.35Butwateristhedevelopmentofevolvedsparsemodelsfordeepcoolinghasthepotentialtoexacerbatepressureonneuralnetworks,whichcanreducecomputationsbywaterresourcesinspecificlocations.AVirginiaTechapproximately5to10timeswhilemaintainingthestudyestimatesthatone-fifthofUSdatacenterssamelevelofoutputqualitycomparedwithdenser(primarilylocatedinWesternstates)drawtheirwaterbaselinemodels.32frommoderatelytohighlystressedwatersheds.36BOSTONCONSULTINGGROUPGOOGLE24Exhibit5-CriticalchoicescanshapeAI’semissionsfootprintFactorandBaselineMetricReducedemissionsBaselineIncreasedemissionsAIadoption&usage–100%+100%Baseline:CurrentAIcomputeconsumptionUnknownModeldevelopment&coding–80–Baseline:Transformer(2017)90%Datacenterinfrastructure(servers)Evolved(2019)andBaseline:UnoptimizedsystemsenergyusePrimer(2021)models(P100from2017)–93%–23%Datacenterinfrastructure(non-ITequipment)Baseline:TraditionaldatacenterenergyneedsML-oriented1ML-oriented1(TPUv4,2021)(TPUv2,2019)–30–50%ClouddatacentersEnergysourcecarbonintensity3–94%–77%–22%–16%+12%+60%Baseline:Globalgridemissionintensityin2022NorwayBrazilSingaporeSouthAfricaFLAP-DUScountries4Note:Thefigureisbasedonthemostrecentdataavailableonactualemissionsprofilesfordifferentchoices—andisnotintendedtosuggestthatthesefactorsareequallyimportant.1Patterson,D.etal.,2021.CarbonEmissionsandLargeNeuralNetworkTraining.2Patterson,D.etal.,2022.Thecarbonfootprintofmachinelearningtrainingwillplateau,thenshrink.3EnergyInstituteStatisticalReviewofWorldEnergy.4EuropeDataCentersmainlocations:Frankfurt,London,Amsterdam,Paris,andDublin.Inthedatacentersector,wateruseisnotwidelyreport-Techplayersarealsostartingtoexplorenew,morewa-ed—andactualvolumeswillvarywidelybasedontheter-efficientapproaches.Inadatacenter,temperature,center’ssize,location,localweatherconditions,andtheairflow,andrelativehumidity(RH)arethethreecriticaluseofitsinfrastructure.A2016studyfromtheUSfactorstomanage.BasedonguidancefromtheAmericanDepartmentofEnergyestimatesdatacenterwaterSocietyofHeating,Refrigerating,andAir-Conditioningconsumptionat1.7billionliters/day,ofwhich0.3billionEngineers,Metahasexperimentedwithshiftingthelow-liters/dayisusedonsiteforcooling,or0.02%oftotalUSer-boundforRHinitsdatacenterfrom20%to13%.Thewaterconsumptionof1,218billionliters/day.37nine-monthpilotyieldedwatersavingsof40%.41Clearly,datacenteroperatorsneedtobemindfulofandDatacenteroperatorsarealsotakingstepstoreplenishmanageatradeoffbetweenenergyuseandwateruse.theirwaterconsumption.Google,forexample,hasatargetGooglebegandisclosingwateruseforeachofitsownedtoreplenish120%ofthefreshwatervolumeitconsumesUSdatacenterlocationsin2021andforglobalownedonaverageby2030throughinitiativesincludingwetlanddatacenterlocationsin2022.AGoogleanalysisoftherestoration,rainwaterharvesting,andlandconservation.2021datafindsthatitsembraceofwatercoolingacrossitsdatacentersreduceditscarbonemissionsbyroughly300kilotons—theemissionsequivalentofabout64,000passengervehicles.38The5.2billiongallonsofwaterrequiredtodothatin2022wascomparabletothewaterneededtoirrigate34.8ofthemorethan11,000golfcoursesintheUnitedStatesortheannualwaterconsumptionof69,800averageAmericanhomes.39,4025ACCELERATINGCLIMATEACTIONWITHAIIssue3:WasteOtherPotentialRisksTheUnitedNationsestimatesthatglobally53.6millionTheriseofAIalsobringsasetofsocietalandethicalrisksmetrictonsofelectronicwastewasgeneratedin2019.Thisthatmustbemanagedvigilantly.Itisimportanttohaveafigurerepresentsa21%increaseinjustfiveyears.Andthesetofclearprinciplesthatguidethedevelopmentandusevolumeisprojectedtoincreaseto74.7millionmetrictonsofAI.Forexample,Googlehasoutlinedthefollowingsevenby2030—representinganeardoublingofe-wasteoverakeyprinciplesforitsdevelopmentofAIapplications:4420-yearperiod.1.BesociallybeneficialWhilethereiscurrentlyalackofspecificdataregardinge-wastegeneratedbydatacenters,theyclearlyonlyac-2.Avoidcreatingorreinforcingunfairbiascountforafractionofthebroadere-wastechallenge.None-theless,thereisaclearimperativefortechfirmstotakea3.Bebuiltandtestedforsafetysmarter,morecircularapproachtowaste.4.BeaccountabletopeopleCirculareconomyprinciplesemphasizetheopportunitytomaximizethelifespanofproductsandmaterialsthrough5.Incorporateprivacydesignprinciplespracticessuchasreuseandrecycling.Fortechcompanies,thisentailsdesigningproductsanddatacenterinfrastruc-6.Upholdhighstandardsofscientificexcellenceturewithaneyetowardlongevityandeaseofupgradingorrepurposing.Moreover,itinvolvesestablishingefficient7.Bemadeavailableforusesthataccordwiththesesystemsforrecyclingandrefurbishingelectronicequip-principlesment,ensuringthatvaluablematerialsarereclaimedandthathazardoussubstancesaredisposedofresponsibly.ApplicationSelectionandOptimizationMetrics.Thereareanynumberofclimate-unfriendlyapplicationstoMicrosoft’sCircularCenters,forexample,focusonfindingwhichAIcouldbeapplied,forexample,oilandgasexplora-productiveusesfordecommissionedequipment,includingtion.Google,forinstance,haspledgednottodevelopnewhomesforolderequipment,suchasinschools.TheycustomizedAI/machinelearningsolutionstofacilitatebreakserversdownintocomponentsthatcanbereusedbyupstreamextractionforoilandgas.Inaddition,algorithmsothers,andreturnotheritemstosuppliersforrecyclingcouldbedesignedtooptimizeforfinancialorotherout-andreclamation.42Thefirstofthesecentersopenedincomesoverenvironmentalones.Forexample,travelweb-2020inAmsterdamandhasbeenabletochannel83%ofsitealgorithmscouldsteercustomerstothecheapeste-wasteintoreuse—and17%intorecycling.Basedonthatflightsregardlessofcarbonemissionsinsteadofguidinginitialsuccess,fiveotherCircularCenterswereestablishedthemtowardaninformedcompromisebetweenpriceandin2022.emissions.In2016,Googleannounceda“ZeroWastetoLandfill”goalItisalsoimportanttoguardagainstunintendedconse-foritsdatacenteroperations,whichitdefinesasmorequences.WhileAIcanhelpoptimizeresourceuseandthan90%ofwastedivertedfromlandfill.43In2022,38%ofcurtailemissions,therearescenariosinwhichitcouldGoogleownedandoperateddatacentershadachievedinfluenceconsumerbehaviorsthatleadtounintendedZeroWastetoLandfill.increasesinemissions.Forinstance,AI-poweredautono-mousvehiclesandsmarttransportationsystemscanOtherfirms—forexample,IronMountainthroughits2021optimizeroutesandreducefuelconsumption,buttheiracquisitionofITRenew,whichspecializedinrefurbishingconveniencecouldleadtoincreasedvehicleuse,whichandrepurposinguseddatacenterequipmentfromhyper-couldpotentiallyincreaseoverallemissionsifthevehiclesscaleoperators—arepursuingcircularityasabusinessarenotpredominantlyelectricorpoweredbyrenewableopportunity.energysources.Lookingbeyondtheriskofnegativeenvironmentalimpact,wemustguardagainstunethicalapplications,forexample,thespreadingofmisinformationanddisinformation.BOSTONCONSULTINGGROUPGOOGLE26EquityandBias.Clearly,AImodelsneedtobetrainedonPrivacyandSecurity.Thelargedatasetsthattraindiversedatasetsthatreflecttheworld’srangeofpeopletoAIalgorithmswillattimesincludepersonaldata.Itisensurebothfairnessandtheaccuracyofmodeloutput.essentialtoensurethatAIapplicationsarealignedwithAndbeyondthat,itisessentialthatwetakepositivestepsestablishedprivacystandardsandregulationstoprotecttoensurethatthegrowthofAIdoesnotexacerbateregion-individuals.aldisparities.Today,themajorityofadvancedclimatemodelingandAIdevelopmentoccursintheGlobalNorth.GiventhepromiseofAItoaddresstheclimatecrisis,Thisdividearisesfromseveralfactors,includingtheGlobalpolicymakerswillwanttoencourageitsuse—butalsoNorth’smoreextensivetechnologicalinfrastructurethatmitigateitspotentialrisks.Thenextchapteroffersasum-allowsforricherdatacollection—forinstance,throughmaryofcriticalpolicyoutcomes.satellitesanddrones—anditsgreatercomputingpower.Additionally,AIexpertiseandresourcestendtobemorereadilyavailableintheseaffluentregionswhencomparedwiththeGlobalSouth.However,theconcentrationofAIdevelopmentandapplicationincertainregionshassignificantrepercussions.ClimateAImodelsprimarilytrainedondatafromtheGlobalNorthmayinadvertentlyneglectvitalinformationabouttheGlobalSouth,withitsdistinctclimatepatterns,vulnerabilities,andemissionssources.AGoogleResearchteaminGhana,forexample,isfocusedonleadingmanysustainabilityinitiativesofparticularinteresttoAfricaincollaborationwithlocaluniversitiesandresearchcenters.45Additionally,historicaldataoftenmirrorsemissionsandclimateimpactsinmoreindustrializedregions,furtherskewingdatarepresentation.WhensuchbiasedAImodelsinformclimateassessmentsorpolicymakingfortheGlobalSouth,theyriskyieldingflawedandinaccurateoutcomes.Excludingspecificregionsorhistoricalperiodscanresultininaccuratepredictionsandevaluationsofcarbonemissions,environmentalconsequences,andclimatetrendsintheseunderservedareas.Thisunderminesclimatescienceaccuracyandobstructseffectivedecision-makingandplanningforclimateadaptationandmitigationintheGlobalSouth.27ACCELERATINGCLIMATEACTIONWITHAIAIforClimateASummaryofCriticalPolicyOutcomesPolicymakersplayacentralrolebothinharnessingthePolicycanmakeadifferenceinensuringandacceleratingpotentialofAIforclimateactionandinensuringitsthefollowingthreecriticaloutcomes:sustainableandequitableuse.•EnablingtheuseofAIforclimateactionbybuildingInthischapter,weshareasetofsuggestionsforpolicymak-awarenessandensuringequalaccesstodata,tech-ersthatsynthesizestheconvergencesandcomplementari-nologyinfrastructure,andtalentaroundtheglobetiesacrossmorethan30expertinterviewsandacompre-hensivereviewoftheliterature(SeetheReferencessection•Deployingpublic-sectorsolutionsonpriorityusecasesformoredetail).Wealsodrawonbest-practiceapproacheswhilecatalyzingprivatesectoractionthroughtherightfromrelatedpolicydomainssuchasenergy,transport,andincentivesbuildings.•PromotingtheresponsibleuseofAIinclimateaction,takingintoaccountitspotentialenvironmentalaswellassocialimpactsBOSTONCONSULTINGGROUPGOOGLEExhibit6offersamenuofpossiblepolicymovesinsupportEncourageDataCollectionandSharingofthesedesirableoutcomes.Therestofthechapterpro-AIimpactstartswithgooddata.Withoutit,algorithmsvidesexamplesofhowpolicymakers—aswellasbusinesscannotgenerateaccurateandeffectiveinsightsorrecom-andsocial-sectorleaders—havealreadytakenstepstocon-mendations.Actionablewildfirealerts,forexample,cannottributetotheseoutcomes.Whilepolicyprioritiesmustbebedevelopedwithoutaccesstohigh-quality,real-timetailoredtothespecificcircumstancesandcapabilitiesofsatellitedata—andgoodagriculturaladviceisimpossibleeachcountryandregion,webelieveallleadersshouldifdataisavailableonlyforalimitednumberofcropsoradoptclearAIprinciplestoensuretheresponsibledevel-geographies.Moreover,toyieldusefulinsights,manyAIopmentandapplicationofthetechnology.applicationsforclimatewillneedtotapintomultipledatasources.LondonTransport,forexample,leveragesdataonWeexpandonthethinkingintheremainderofthechapter,emissionssources,roadtraffic,airquality,andpopulationandhopeourframingprovidesinspirationandastartingdensitytomonitorairpollutionchallengesandidentifypoint.highlyexposedlocations.Makingthedataavailableisnotenough;italsoneedstobeaccessibleinstandardformatsEnableAIforClimatethatallowdatafromdifferentsourcestobemergedsafelyandefficiently.IfwearetobesuccessfulinmaximizingAI’scontributiontoclimateaction,onecriticalareaforpolicyfocusisontheItisthereforeessentialthatpolicymakerstakestepstosupplyside,ensuringthatAI’scriticalinputs—high-qualitymakedataaccessible—ataminimumintheclimatedata,technologyinfrastructure,andtalent—areavailablesphere,giventheurgentneedforrapidandeffectiveactionwhereverneeded.Wediscusseachinturn.toreduceemissionsandbuildresilienceglobally—whilealsoprotectingtradesecretsandintellectualproperty.Exhibit6-AIforClimate:AsummaryofcriticalpolicyoutcomesPolicyEnableAIforclimateDeployAIforclimateGuideAIforclimateoutcomesEnsuretheavailabilityofdata,DriveAIsolutionsforclimateinPromoteenvironmentallyandtechnologyinfrastructure,andtalentpublicsector—andcatalyzesociallyresponsibleuseofAIprivatesectoractionEncouragedatacollectionandsharingDeﬁneanddeliveronpublicAddressenvironmentalimpactsPromotetheprincipleofsectorprioritiesofAIoperationsclimate-relateddataasacommongoodShapeandexecuteAIforclimateEnhancetransparencyandstrategiesanddemonstrationcasesstreamlinetheadoptionofPolicyEnsuretechnologyaccessEncourageprivatesectoradoptionsustainableAIpracticesmovesandaﬀordabilityCreateincentivestoaccelerateuseofPromotesociallyresponsibleEncourageandinvestinessentialAIforclimateuseofAIforclimatetechnologyinfrastructuretosupportAIEncouragefairness,inclusiveness,Cultivateawarenessandbuildexpertisesafety,anddataprivacyInvestinknowledgeandtalenttodriveAIforclimatesolutionsSource:BCGanalysis.29ACCELERATINGCLIMATEACTIONWITHAILeaderscould:EncouragecollectionandsharingofExamples:climate-relateddataandtoolsacrosspublicandprivateorganizations.TheUKClimateChangeStatisticsPortalgivesopenaccesstoclimate-relateddataandstatistics(weather,emissionsofGHG,statusofsurfacewater,renewableenergyshare,etc.).Thisdataiscompiledfromvariousgovernmentdepartments,agencies,andpublicbodies.TheUSClimateResilienceToolkit,developedbyacollaborationofmultiplefederalagenciesandorganizationsledbytheNationalOceanicandAtmo-sphericAdministration,providesessentialclimateinformation,projections,andtoolstohelporganizationsenhancetheirresiliencetoclimate-relatedchallenges.SignSmart,alsoknownastheNationalGreenhouseGasInventorySystem,isanapplicationsystemdevelopedbytheIndonesiangovernment.Itsobjec-tiveistoprovidevalid,accurate,andup-to-datedataandinformationaboutGHGemissions,whilealsoenhancingtheeffectivenessofdataprocessingandGHGestimationatthenational,provincial,anddistrict/citylevels.DataCommonsisaninitiativeledbyGoogledesignedtocentralizeandstreamlinepubliclyavailabledatafromdiversesources.Itprovidesawiderangeofclimateandsustainability-relateddata,coveringareassuchasemissions,naturaldisasters,andwaste.Accessingthisdataismadesimplethroughaninterfacethatsupportsnaturallanguagesearches.TheAI-drivenqueryfunctionretrievesresultsdirectlyfromDataCommons,providinglinkstotheoriginalsourcesofinformationanddata.DefinestandardprocessesandExamples:protocolsforclimate-relevantdatagatheringandsharingtoensuredataTheEuropeanSpaceAgency’sClimateChangeInitiative(CCI)aimstodeliv-isrobust,trustworthy,safe,anderaconsistent,satellite-deriveddatasetofEssentialClimateVariables(forrespectfulofprivacy.example,greenhousegases,sealevel,glacierstatus,etc.)toaidclimatemodelersandresearchersoverthelongterm.Toguaranteeconsistencyamongthevariousprojectswithintheprogram,ithasreleaseddatastan-dardsoutliningtheminimalrequirementsforclimatedataproducers.TheWorldMeteorologicalOrganizationhaspublishedtheTechnicalRegula-tions,aninternationalframeworkfordatastandardizationandinteropera-bilityinthefieldsofmeteorology,hydrology,climatology,andrelatedenvi-ronmentaldisciplines.Thesestandardsenablethecontinuousoperationofglobalsystems,ensuring24/7observations,dataexchange,management,forecasting,andthedeliveryofauthoritativescientificassessmentsandstandardizedservices.Createdatacatalogs46acrosspriorityExamples:sectorsforallclimate-relevantdatacategories(forexample,weather,TheWorldMeteorologicalOrganizationhasnotonlypublishedasetofdatawater,agriculture,energyuse,andstandardsbutalsoaCatalogueforClimateData,acuratedlistingofglobal,socioeconomicfactors).regional,andnationaldatasetsofclimate-relateddatathatmeetitsstan-dardsfordataqualityandstewardship.ClimateChangeAI,anonprofit,haspublishedtheCCAIDatasetWishlistthatenumeratescurrentlyunavailabledatathatcouldaccelerateAI-drivenclimateprogress.Itclassifiesdesirabledatabytopicaswellasbyitscurrentstateofavailability(forexample,publicdataneedingstructure,privatedataneedingrelease,scattereddataneedingcollation,andscarcedataneedingcollection).BOSTONCONSULTINGGROUPGOOGLE30EnsureTechnologyAccessandAffordabilityThiscriticaltechnologyinfrastructure,whilewidelyavail-Withoutdevicescollectingdatafromtherealworld(forableinthedevelopedworld,isnotcurrentlyaccessibleinexample,sensors,drones,andsatellites)—andwithoutmanyless-developedregions,particularlytheGlobalSouth.connectivity(forexample,fiberand5G)andcomputingAffordabilityisalsoaconcern.Forexample,customersininfrastructure(forexample,clouddatacenters)—AIalgo-less-developedcountriespay6%ofpercapitagrossnation-rithmsarepowerless.Consideranapplicationsuchasalincomeformobilebroadbandservice,whilethoseinAI-drivensmartirrigation.Fieldsensorsneedbroadbandtohigh-incomecountriespayjust0.4%.47senddatatoserversandstoragedevicesinthecloud.Theserversthenprocessthedata,comparingitwithinsightsfromatrainingset,andformulaterecommendationsthatneedtobecommunicatedbacktotheirrigationsysteminthefield.Leaderscould:Examples:Buildprivate-publicpartnershipsJapan’sMinistryofEconomy,Trade,andIndustryispartneringwithSakuratoensureaffordableaccessandInternettoexpandJapan’scomputationalcapacity.ItisprovidinghalfofthedeploylocallyorregionallycriticalneededinvestmenttobuildanAIsupercomputer,aimingtoaccelerateAIAItechnologyinfrastructuredevelopmentandimplementationinJapan.(forexample,clouddatacenters,satellites).TheEuropeanHigh-PerformanceComputingJointUndertaking—a€7billioninitiativeover2021–2027tobuildpetascaleandpre-exascalesupercomput-inginfrastructureinEurope—aspirestoaccelerateEuropeaninnovationbyexpandingaccesstostate-of-the-arttools.TheUSCommunityInfrastructureforResearchinComputerandInforma-tionScienceandEngineering(CIRC)programseekstoincreaseresearcheraccesstocriticaltechnologyinfrastructurebyfundingthedevelopmentandimprovementoftop-tierresearchinfrastructure.EmpowertheprivatesectortoExamples:buildand/orexpandAItechnologyinfrastructure.Norway’sdatacenterstrategy,introducedin2018,featurespublicinvest-mentsinfiberinfrastructureaswellastaxincentives(suchaspropertytaxreductions)aimedatattractingdatacenteroperatorsandensuringtheaccessibilityofcomputinginfrastructure.TheSouthAfricangovernmentaimstolaunchaNationalCloudandDataPolicy.Thisinitiative—currentlyunderconsultation—seekstoprovidepolicycertaintyforinvestmentsindatacentersandcloudservicesandtoreinforceSouthAfrica’sleadingpositioninAfrica.ThedraftpolicyincludesprovisionsforestablishingaSpecialEconomicZoneforDigitalandICT,aswellaspoliciesaddressingdataprotection,datalocalization,andcross-borderdatatransfers.SupportR&DforAItechnologyExamples:infrastructureacrossthepublicsector,privatesector,andacademia.TheUSNetworkingandInformationTechnologyResearchandDevelop-ment(NITRD)programisafederallyfundedR&DinitiativefocusedonadvancedIT,coveringcomputing,networking,andsoftwarenationwide.Its25memberagenciesinvestabout$9.6billionannuallyinvariousR&Dprograms,includinghigh-capabilitycomputingsystems,advancedcommuni-cationnetworksandsystems,andmore.TheQuantumTechnologiesFlagshipisa10-year,€1billionEUinitiativethataimstoadvanceEurope’sleadershipinquantumtechnologiesbybridgingresearchwithpracticalapplications.Itbringstogetherresearchinstitutions,industrypartners,andpublicfunders.31ACCELERATINGCLIMATEACTIONWITHAICultivateAwarenessandBuildExpertiseTranslatingthatsupportintoclimateprogresswillrequire,PeopleneedtoremainatthecenterofpolicyandclimateinadditiontomoreAIcomputinginfrastructure,moreaction.TheymustbeawareofAI’spotentialandsupporttalent.Withouttechnicalexpertssuchasdatascientists—itsuseforclimateaction—andpeoplewithAIskillswillbeanddomainexpertssuchasclimatologistsandclimateneededtohelpaddresstheclimatechallenge.economists—itwillbeimpossibletodevelop,deploy,andgovernclimate-relatedAIapplications.InarecentBCGItisimportanttobuildawarenessamongallstakehold-survey,78%ofbusinessleaderswithresponsibilityforers—policymakers,corporatedecision-makers,civilser-climate,AI,orbothcitedinsufficientaccesstoqualifiedvants,andthebroaderpublic—ofAI’spotentialcontribu-talentasabarriertousingAItoaddresstheirclimate-relat-tionstoclimatesolutionsaswellasofitsrisks.Withedchallenges.Andcriticaltalentisnotonlyinshortsupply,awarenessofthechallenges,ofkeyachievements,andofbutalsoheavilyconcentratedinthedevelopedworld.bestpracticecomessupportforAI-enabledclimateaction.AccordingtotheOECD’sAIPolicyObservatory,NorthAmericaishometo30%oftheworld’sdatascientistsandmachinelearningexperts,whilesub-SaharanAfricahostsunder2%.Leaderscould:EstablishAIandclimatetrainingandExamplesofAI-relatedtrainings:literacyprogramsforpolicymakersandthebroaderpublicsectorGovernmentAICampus,createdbyGoogle.organdtheRockefellerworkforce.Foundation,isanonlinecareer-developmentinitiativeforgovernmentstaffthatpreparesthemtoleadintheageofAI.AI4Gov—aEuropeanUnion-fundedmastersdegreeprogramofferedbyfourleadinguniversities,focusesonAI’spublicsectorapplication.ItispartofabroaderEuropeaninitiativetocreateAI-relatedmastersprogramstobuildskillsinareassuchasAIethicsandAI’sapplicationtohealthcare.TheUNESCO-developedArtificialIntelligenceandDigitalTransformationCompetencyFrameworkprovidesguidanceontheessentialAIanddigitalcompetenciesforcivilservants.Thisinitiativerespondstoasignificantde-mandforeffortstoenhancethedigitalskillsofgovernmentofficials,particu-larlyinAfrica.48UNESCOconductedworkshopsinAfricaandIndiatogainadeeperunderstandingofthechallengesbeforeformulatingitssolutions.49Examplesofclimate-relatedtrainings:TheUNClimateChangeLearningPartnership(UNCC:Learn)offersarangeofintroductoryandadvancedonlinecoursesforpolicymakerstolearnhowtoaddressclimatechangeandapplyanintegratedapproachtoclimateactionthroughoutthevariousstagesofthepolicycycle.ClimateChangeandEnergy:PolicymakingfortheLongTermisanexecutiveeducationprogramforpolicymakersdevelopedbyHarvard’sKennedySchoolofGovernment.Itseekstoequipthemwiththeknowledge,analyti-caltools,andframeworksneededtocomprehendclimatescienceandeco-nomics—andtocraftpoliciesandadaptationstrategies.BOSTONCONSULTINGGROUPGOOGLE32SupportthecreationandexpansionExampleofAIforClimatetrainings:ofAIandclimate-relatedupskillingprogramsforcorporates(forexample,TheClimateChangeAIsummerschoolaimstoequipindividualswhohaveclimatechangetrainingforAIexperts,AIabackgroundinAIand/orclimatechangewiththeknowledgeandskillsintroductionforclimateexperts).necessarytoaddresssignificantclimatechallengesusingAI.ExamplesofAI-relatedtrainings:Quebec’sMinistryofEmploymentandSocialSolidarityhasgranted$23.4millioninfundingtoSCALEAItoupskillover25,000professionals,execu-tives,andmanagersinAIbetween2019and2023.TheEU-fundedprojectArtificialIntelligenceSkillsAlliance(ARISA)unites20partnersforafour-yearperiod(2022–2025)tocreatetheEuropeanstrat-egyforAIskillsdevelopment,includingup/reskillingcurriculaandlearningprograms.ExamplesofClimateChangetrainings:Climate-KIC,aknowledgeandinnovationcommunitysupportedbytheEuropeanInstituteofInnovationandTechnology,provideseducationpro-grams,attheintersectionofzero-carbon,climateresiliency,andinnovation,inEuropeandonlineforpostgraduatesandprofessionals.ThenationalbusinesssupportagencyofIreland,Skillnet,offerstheClimateReadyAcademytohelpIrishbusinessesdevelopclimate-relatedskills,includingsector-specificprogramssuchastheEnergyLeadersProgramme.BuildAIandclimatemoduleswithinExamplesofclimate-relatedcurricula:educationcurricula(forexample,early-stageinitiationforK-12,cross-skillsSinceSeptember2020,Italianstudentsineverygradehavespent33hoursforAIstudentsorclimatestudents,etc.).eachyearlearningaboutclimatechangeandsustainability,bringingItalytotheforefrontofenvironmentaleducationworldwide.IntheUS,theNationalOceanicandAtmosphericAdministration’sCollec-tionofClimateandEnergyEducationalResources(CLEAN)providesopenaccesstoover700validated,ready-to-useteachingmaterialsandguidelinessuitableforsecondarythroughhighereducationclassrooms.ExamplesofAI-relatedcurricula:CountriessuchasFinland,theUK,Japan,andSingaporehaveintroducedcomputationalthinkingandprogrammingtopedagogicalcoursestoincreasestudents’exposuretocodingandcomputingatearlystagesofeducation.50Morocco’sMinistryofHigherEducation,ScientificResearch,andInnovationhasinitiatedthe“Code212”project,designedtohelpstudentsacquireskillsincoding,programming,bigdata,AI,andrelatedfields.AcoreobjectiveistoestablishCode212centersinallnationaluniversities,therebyenhancingstudents’digitalcompetenciesalongsidetheirspecializedstudies.33ACCELERATINGCLIMATEACTIONWITHAIDeployAIforClimateForGermany,anindustrialgiantwheretheClimateActionActmandatescarbonneutralityby2045,theprioritiesforWhileenablingAIisnecessarytohelpsocietymitigateandAImightbeacceleratingthedecarbonizationofindustriesbuildresiliencetoclimateimpacts,policymakerscanplayanddrivingenergyefficiency.51,52Bycontrast,Bangla-animportantroleinacceleratingthedeploymentofAIdesh—givenitssignificantvulnerabilitytosea-levelrisetechnologiesinboththepublicandprivatesectors.Weandextremeweather—mightprioritizeacceleratingitsdiscusseachinturn.NationalAdaptationPlananddevelopingearlywarningsystemsforcommunities.DefineandDeliveronPublicSectorPrioritiesThedozenAIusecasesforclimateoutlinedinExhibit3Inadditiontoidentifyingtheirhigh-priorityusecases,areallimportant,butthatdoesn’tmeanthatthey’repolicymakersmustalsoexpeditetheapplicationofAItoalwaysequallyimportant.Everycountryandregionisclimatechallenges.ThepublicsectorcanserveasapivotaldifferentandeachconfrontsitsowndistinctsetofclimatecatalystforacceleratingAI-supportedclimateactionandchallengesgivenitsgeography,industrymix,humansetacompellingexamplefortheentireeconomy.Thisisresources,andrelativewealth.Therefore,everycountryparticularlyimportantsince,while87%oforganizationsorregionneedstodevelopitsownclimateprioritiesacknowledgetheroleofAIinclimateaction,only40%canandactionplan.envisionpracticalapplicationswithintheirownoperations.Leaderscould:Example:IntegrateAIsolutionsandexpertiseAI4PublicPolicy—aninitiativeoftheEU’sR&DprogramHorizon2020—isintogovernmentstrategicplanningcreatinganopencloudplatformforautomated,scalable,transparent,andonclimatepriorities(forexample,citizen-centricpolicymanagement.Lisbon,forexample,isusingittomapNationallyDeterminedContributions,itscurrentinventoryof,andexpansionpotentialfor,solarpanels—withtheNationalAdaptationPlans,andsec-goalofdevelopingrenewablessupplyforecaststoinformbuildingcodesandtor-specifictransitionplans).incentivebudgets.DefineprioritysectorsorusecasesExamples:53forAItosupportclimateactionatlocalandregionallevels(forexample,Denmark’sAIstrategy(2019)identifiesthreeclimate-relatedpriorityareas:aspartofNationallyDeterminedenergyefficiency,precisionagriculture,andtrafficoptimization.Contributions,NationalAdaptationPlans,orNationalAIstrategies).TheNetherlands’StrategicPlanforAI(2019)includescommitmentstoleverageAIinagricultureandinacceleratingtheenergytransition.TheUK’sAIRoadmapadvocatesforAIusetoaddressclimatechange,particularlyintheenergysector.ThegovernmentoftheRepublicofthePhilippinesisadvocatingusingAItotackleclimatechangeadaptationchallengesanddisasterriskreduction—inparticular,throughprogramsledbytheDepartmentofScienceandTechnology,suchasAIforaBetterNormal.ImplementprioritypublicsectorAIExamples:solutionsforclimate.TheUSNationalAIInitiativeActof2020formalizestheNationalOceanicandAtmosphericAdministration’s(NOAA’s)roleincoordinatingAIapplica-tionsforclimate,ocean,Earth,andspacesciences.NOAAhascreatedtheNOAACenterforArtificialIntelligence(NCAI),whichcollaboratesacrossscientificfieldstopromoteresponsibleandequitableAIuseforenvironmen-talresearch.SingaporeisusingAItopredictfloodsandtestflood-resilientinfrastructure.Forexample,apartnershipbetweentheislandnation’sNationalResearchFoundationandtheHydroinformaticsInstitute(H2i)hasdevelopedVirtualWater,asurfacewatersimulationtoolboxthatcanpredictandsimulatefloodsresultingfromheavyrainfallevents.BOSTONCONSULTINGGROUPGOOGLE34EncouragePrivateSectorAdoptionToaddressthesechallenges,policymakerscouldactivatePolicymakerscouldalsoplayacatalyticroleinacceleratingthefollowingfivecorrespondingkeylevers:privatesectorAIadoptionbyaddressingmajorchallengeshinderingtheat-scaleimplementationofAIforclimate.•Createpublic/privatepartnershipstodriveAIadop-Globally,weseethefollowingfivekeyhurdles:tioninkeysectorsandapplications•UnclearGoals.Intheabsenceofclearregional,•Removeregulatorybarrierstoandprovideregulatorynational,orsector-specificobjectivesforclimatesupportandclarityforAIadoptionaction,AI-driveninnovationmaybecomefragmented,andresourceallocationinefficient.Theestablishment•AccelerateinfrastructuremodernizationandofpriorityinnovationdomainsforclimateactioncanbedigitalizationtoenableAIuseasignificantunlock.•SupportAIresearchandinnovationinclimate-•RegulatoryLimitations.Insomecases,particularlyrelevantdomainsbyinvolvingbothacademiaandtheinheavilyregulatedsectorssuchasenergyandtrans-privatesectorport,policymakerscouldadoptathoughtful,risk-basedgovernanceframeworkforAIthatensuressufficient•OptimizeincentivestodriveAIadoptionatscaleforprotectionsandsafeguardswhiletakingcarenottostifleclimateactionprioritieswhenexistingmechanismsareinnovation.And,conversely,thelackofclearregulatoryinsufficient(forexample,publicfunding,taxcredits,etc.)frameworksinotherareasmightalsoimpedeinvest-ment.Nonetheless,theseleversneedtobetailoredtothedistinctchallengesofeachjurisdictionandsector.•InfrastructureChallenges.Manyhigh-emissionThesechallengesareshapedbybothglobalandcontext-sectorsrelyonlegacyinfrastructurethatcannotreadilydependentfactorssuchasmarketintricacies,complexity,supportAItechnologies.Forexample,outdatedroadnet-regulatoryframeworks,andinnovationdynamics.There-worksmaystruggletoaccommodatesmarttrafficman-fore,policymakerscouldadoptastrategicapproachbyagementsystemspoweredbyAI—orlimitedpenetrationidentifyingthemosteffectivemeasurestoaddresstheseofsmartmeterscouldhindertheat-scaledeploymentofchallengesanddeterminingtheappropriatelevelofAI-basedenergyefficiencysolutions.intervention,whetherthroughadvocacy,incentives,orbindingmeasures.Theycouldalsodrawinsightsfrom•InnovationCosts.DevelopingandimplementingAIbestpracticeswithineachsectorandjurisdiction,aswellsolutionsoftendemandssignificantfinancialresourcesasacrosssectorsandjurisdictions,inordertolearnfromandtime.Givingaccesstospecificinnovationplatforms,oneanother.suchashigh-performingcomputinginfrastructure—ortotargetedinnovationfunding—canremovesignificantForexample,considertheimplementationofAIingridbarrierstoinnovationfortheinnovationecosystemplanningandmanagement.AIsolutiondevelopersmaybroadly,andparticularlyforsmallandmedium-sizedencounteraconstraintthatnecessitatesholdingspecificenterprises.licensestotestordeveloptheirsolutions.Insuchcircum-stances,policymakersmightconsiderissuingregulatory•DeploymentCosts.Inmanyareas,theremaybesig-waiversorestablishingcontrolledtestingenvironments.nificantentryoradoptioncosts,includingcapacityandSimilarly,inregionswithlimitedsmartmeteradoption,capabilityneeds,thatposechallengesforcompanies,policymakerscouldtakeactionbyallocatingfundingfortheirstakeholders,andcustomers.Forexample,theinfrastructuremodernizationorofferingincentivestoadoptionandimplementationatscaleofprecisionagri-encourageelectricityproviderstoreplaceagingmeters.culturecouldbeheldbackbecauseoffarmerconcernsThesemeasureswouldfacilitatetheintegrationofAIregardingtheexpenseofinstallingtheInternetofThingsapplicationsdesignedtoimprovetheenergyefficiencyofhardwareneededtosupportit.buildings.Tounderscoresectoraldistinctions,inthefollowing,weprovideexamplestoillustratehowpolicymakershavebegunimplementingpoliciesinsector-specificcontexts.35ACCELERATINGCLIMATEACTIONWITHAILeaderscould:Createpublic/privatepartnershipstoExamples:driveAIadoptionwithinandacrosskeysectorsandapplications.Buildings:DOITSMARTer—apublic-privatepartnershipbetweentheTech-nicalUniversityofCluj-Napoca,ØsfoldUniversityCollege,theNorwegiancompanyNxTech,Romania’sAlbaIuliamunicipality,andtheNGOCenterfortheStudyofDemocracy—seekstodevelopAI-drivenenergyefficiencysolutionsforpublicbuildingsinRomaniaandNorway.Energy:TheUSDepartmentofEnergy’sPrincetonPlasmaPhysicsLabora-toryispartneringwiththeRenaissanceFusionstartuponAI-driveneffortstoacceleratethedevelopmentofcarbon-freefusionenergy.Transport:InProjectGreenLight,Googleiscollaboratingwith12citiesincludingManchester,RiodeJaneiro,Jakarta,andAbuDhabitoreducestop-and-starteventsthroughAI-supportedtrafficlightmanagement.Earlyindicatorsshowapotentialforuptoa30%reductioninstops,whichcouldreduceemissionsatintersectionsupto10%.Thiscouldhaveasignificantimpact,ascarsatcityintersectionsgenerate29timesmorepollutionthandocarsontheopenroad.RemoveregulatorybarrierstoandExamples:provideregulatorysupportandclarityforAIadoption.Electricity:Spain,Brazil,andAustraliahavelaunchedregulatory“sandboxes”fortheirelectricitysectorsthatcanofferwaiversforclimate-relatedpilots.InAustralia,forexample,gridparticipantscanbegrantedexemptionsfromregisteringasanetworkserviceprovider.Agriculture:Keyagro-foodstakeholdershaveco-signedasetofnon-bindingguidelinesentitledtheEUCodeofConductonAgriculturalDataSharing.Similarly,TheAmericanFarmBureauFederationhasworkedwithstake-holderstoestablishthePrivacyandSecurityPrinciplesforFarmData.Accelerateinfrastructuremoderniza-Examples:tionanddigitalizationtoenableAIuse.Buildings:In2022,theUKgovernmentimposednewbindingrequirementsthatenergysuppliersinstallsmartmetersinhomesandsmallbusinesses.Thetarget,currentlyundernegotiation,aimsfor80%smartmetercoverageinhomesand73%insmallbusinessesby2025.Energy:TheModernizationFundisadedicatedfundingprogramtosupport10lower-incomeEUMemberStatesinmodernizingtheirenergysystemsandimprovingenergyefficiency(forexample,modernizationofenergynetworks,districtheating,etc.).Similarly,adoptedinlate2022,theDigitaliz-ingtheenergysector–EUactionplan,whichischaracterizedbycybersecurity,efficiency,andsustainability,aimstocultivateacompetitivemarketplacefordigitalenergyservicesanddigitalenergyinfrastructure.SupportAIresearchandinnovationExamples:inclimate-relevantdomainsbyinvolvingbothacademiaandtheEnergy:InMay2023,theUSDepartmentofEnergyannounceda$40mil-privatesector.lioninvestmentin15projectsfocusedondevelopinghigh-performance,energy-efficientcoolingsolutionsfordatacenters.Agriculture:TheEU’sHorizon2020programforresearchandinnovationhasallocatedmorethan€200milliontosupportthedeploymentofprecisionfarming.BOSTONCONSULTINGGROUPGOOGLE36Industry:TheUK’sProjectBluebirdiscreatingadigitaltwinofUKairspacethat,byidentifyingwaystomakeairtrafficcontrolmoreefficient,canhelptheaviationindustryachieveitsgoalofnetzeroby2050.Urbanplanning:VirtualSingapore—acollaborationbetweentheNationalResearchFoundation,theLandAuthority,andtheGovernmentTechnologyAgency—usestopographicanddynamicdatatocreatea3Ddigitalreplicaofthenationstatethatistheofficialplatformforsimulationandvirtualtestingofurbanplanningsolutions.OptimizeincentivestodriveAIExamples:adoptionatscaleforclimateactionprioritieswhenexistingmechanismsAgriculture:AustraliahaslaunchedtheFarmsoftheFutureAgTechGrantareinsufficient(forexample,publicProgramtoencouragefarmerstoadoptprecisionagriculturetoboostpro-funding,taxcredits,etc.).ductivityandimproveresourcemanagement,includingwaterefficiencyanddroughtreadiness.Buildings:TheUSFederal-StateBuyCleanPartnershipisacommitmentbythefederalgovernmentand13statestopurchaselower-carbonmaterialsforfederalandstate-fundedprojectswiththegoalofreducingemissionsandsendingaunifieddemandsignaltothemarket.Industry:Singapore’sEnergyEfficiencyFundoffersfivedifferentgrantpro-gramsthatreimbursebusinessesupto70%ofinvestmentsthatimprovetheenergyefficiencyofindustrialfacilities.Similarly,Japan’sMinistryofEconomy,Trade,andIndustryisofferingsubsidiestopromotetheadoptionofenergy-efficienttechnologiesintheindustrialandcommercialsectors.GuideAIforClimateAIdevelopersrecognizethenecessityofminimizingAI’senvironmentalfootprint,especiallyinthecontextofPolicymakersplayacrucialroleinshapingtheevolutionofexpandingtheuseofAI—andarealreadyworkingtomakeAI—notjustinacceleratingitspositivecontributions,butalgorithmsanddatacentersmoreenergyefficientandtoalsoinminimizingitspotentialnegativeimpacts.Withincreasetheircommitmenttorenewables.Forinstance,regardtoAIforclimate,clearprinciplesandguidelinesareintermsofcleanenergyutilization,severalorganizationsessentialintwoareas:first,maximizingtheenvironmentalsuchastheCleanEnergyBuyersAssociation(CEBA)infriendlinessofAIapplications,andsecond,ensuringthattheUS,RE-SourceintheEU,andtheAsiaCleanEnergyAIapplicationsrespecttheprivacyanddiversityofpeopleCoalition(ACEC)inAPAC(eachwithmembersincludinganddonotinadvertentlyexacerbateinequalities.Wedis-prominenttechplayers)areactivelyadvocatingforthecusseachinturn.accelerationandfacilitationofaccesstocarbon-freeenergyforcorporatebuyersinvariousregions.AddressEnvironmentalImpactsofAIOperationsAsdiscussedinthepreviouschapter,AIhasitsownenvironmentalfootprint.Andwhileithasnotyetbeencomprehensivelymeasured,itisessentialthatwemonitorandmanageit.37ACCELERATINGCLIMATEACTIONWITHAILeaderscould:PromotetransparencyofAI’sExamples:environmentalimpactandpotentialviapragmaticreportingguidelinesTheEuropeanCodeofConductforDataCentresisaninitiativeledbytheandstandards.EuropeanJointResearchCentre.Since2020,thisinitiativehasregularlyreleasedannualguidelinesandambitiousvoluntarystandardsaimedatpromotingenergyefficiencybestpracticeswithindatacenters.Theseguidelinesencompassrecommendationspertainingtoenergyconsumption,aswellasenvironmentalmeasurementandreporting.Spain’sNationalGreenAlgorithmsPlanseekstoraiseawarenessamongAIdevelopersoftheenergyconsequencesoftheirdesigndecisions—andtodrawattentiontomoresustainablechoicesthroughthedevelopmentofstandardsandtoolstomeasuretheenergyconsumptionofalgorithms.TheGermanBundestaghasamendeditsEnergyEfficiencyAct,whichincludesspecificregulationsfordatacentersandprovisionsconcerningtheincorporationofrenewableenergysourcesintodatacenterenergyportfolios,standardsforPowerUsageEffectiveness(PUE),andmandatesforthereuseofheatgeneratedbydatacenters.EncourageAItechnologyproviderstoExamples:makerobustsustainabilityandcleanenergycommitments—andtoadoptThe24/7Carbon-FreeEnergyCompactwasintroducedduringtheHigh-levelclimate-friendlydevelopmentbestDialogueonEnergyin2021.TheparticipantsintheCompact—bothMem-practices.berStatesandnon-stateentities—committedtoasetofprinciplesthatwillleadtoenergypolicies,technologies,procurementpractices,andsolutionsthattransformthebroaderenergyecosystemwiththegoalofachievingacarbon-freeenergygrid.TheClimateNeutralDataCenterPactisaself-regulatoryinitiativethatincludes100+datacenteroperatorsandtradeassociationstomakedatacentersinEuropeclimateneutralby2030.FacilitatetheadoptionofsustainableExamplesofaccesstocleanenergy:practicesforAIoperationsbysupportinginvestmenttoreduceAI’sTheEuropeanCommissionisworkingtoestablishanambitioustargettoenvironmentalimpactandremovingincreasetherenewablesshareofitsenergymixtomorethan40%by2030,barrierstocleantechnologyadoptionwhileintheUS,theBidenadministrationhassetanambitiousgoalto(forexample,reformingplanningforachievea100%carbon-freeelectricitysectorby2035.Theseoverarchingenergyinfrastructure,marketrules,etc.).frameworksinstillconfidenceamonginvestors,fosteringthegrowthofcleanenergyindustries.GovernmentsacrossEurope(includingtheUK,Norway,andEstonia)andelsewhere(suchasSouthAfrica,Brazil,andIndia)employcompetitiveauctionsforsourcingcleanenergy—anapproachthatoffersdistinctadvan-tagesovermoretraditionalfixed-pricesubsidies(forexample,feed-intariffs).Examplesofaccesstocleantech:TheUSInflationReductionActof2022offerstaxincentivesandotherfinancialrewardsforpurchasingequipmentthatreducesand/orsequesterscarbon.TheEU’sGreenDealIndustrialPlanof2023,amongotherinitiatives,pro-videsfundingtoacceleratethegrowthofEurope’scleantechindustry.BOSTONCONSULTINGGROUPGOOGLE38PromotetheSociallyResponsibleUseofAIMoreover,AI’spotentialformisinformationposesathreatforClimatetoinformedclimatediscourse,asthetechnologycanbeAsdiscussedinthepreviouschapter,AIintroducesseveralexploitedtoinfluencepublicperception.Lastly,datapriva-socialrisksthatrequiremindfulmanagementtoensurecyandsecuritymustbeapriority,asAIapplicationsentailthebenefitsofAIforclimatearebroadlyshared.thehandlingofsensitiveinformation,andbreachescouldhavesevereconsequences.Forexample,onesignificantconcernisbiaswithinAImodels,whereskewedorlimitedtrainingdatacanleadtobiasedoutcomes.Forinstance,ifclimate-relatedalgo-rithmsdonotconsiderthespecificcontextandneedsofthecommunitiesmostaffected,biasedmodeloutputwillholdbackprogress.Additionally,thereareregionaldispari-ties.TheGlobalSouthfacesmoreclimatechallenges,whileAIexpertiseandinfrastructureremainpredominant-lyconcentratedintheGlobalNorth.ThisdisparitycanlimitaccesstoAI-drivensolutionswheretheyareneededmost.Leaderscould:Examples:EngagewidelywithallstakeholdersUNESCO’sRecommendationontheEthicsofAIfeatures10principlesthattodevelopandcommunicateprinci-defineahuman-rights-centeredapproachtoAIdevelopment,includingplesforsociallyresponsibleAI,do-no-harm,dataprotection,humanoversight,andnon-discrimination.addressingsuchfactorsasfairness,Similarly,theOECDAIPrinciples,adoptedin2019,definefivekeyprinciplesinclusiveness,safety,anddataforanAIthatisinnovativeandtrustworthyandrespectshumanrightsandprivacy.democraticvalues.IntheUS,theBlueprintforanAIBillofRightsidentifiesfiveprinciplesthatshouldguidethedesign,use,anddeploymentofautomatedsystems,aswellasahandbooktomovefromprincipletopractice.TheEuropeanHigh-LevelExpertGrouponAIhasdevelopedEthicsGuide-linesforTrustworthyArtificialIntelligence.CreateassessmentframeworksExample:tomeasurethealignmentofAIapplicationswithsociallyTheAIVerifyFoundation,aninitiativeofSingapore’sInfocommMediaresponsibleprinciples.DevelopmentAuthority,offersaframeworkandtoolstoassessAIapplica-tionsalongkeyethicaldimensionssuchassafety,security,andfairness.39ACCELERATINGCLIMATEACTIONWITHAIKeepingwarmingtobelowtheParisAgreement’s1.5°CAsweargueinthisreport,evenatthecurrentstateofthetargetwillrequireboldandresponsibleactiononthetechnology,AIhasthepotentialtomakeasignificantnetpartofpolicymakers,businessleaders,andtechnologists.positivecontributiontoaddressingtheclimatechallenge—Itmayalsorequirethemtotakedifferent,moreagile,andtherearealreadymanyinspiringlighthouseexamplescollaborative,andtransparentapproaches.Agilebecauseofitsapplication.However,timeisoftheessence!oftheevolvingimpactsofclimatechangeandtherapidpaceofprogressinAIandclimatescience;andcollabora-tiveandtransparentbecausenosinglestakeholderneces-sarilyhasaccesstoallthedataandexpertiseneededtoaddressitsclimatechallenges—andinsightsoftenbloomwhendifferentperspectivescometogether.BOSTONCONSULTINGGROUPGOOGLE40EndnotesExecutiveSummary1Scope1emissionsrefertodirectemissionsfromsourcestheorganizationownsorcontrols(forexample,itsownvehiclesandproductionequipment).Scope2comprisesindirectemissionsfrompurchasedenergy(forexample,electricity).Scope3includesindirectemissionsacrossthevaluechain,whetherfromsuppliers(excludingpurchasedenergy)orproductusagebycustomers.TheClimateActionImperativeandthePromiseofAI2EmissionsGapReport2022,UNEP,2022.3Groundswell:PreparingforInternalClimateMigration,TheWorldBank,2018.4NDCSynthesisReport,UNFCCC(UNClimateChange),2022.5TheIPCCWorkingGroupIIIreport,ClimateChange2022:Mitigationofclimatechange,estimatesthatfortheworldtoachievenetzeroby2050,emissionswouldneedtobereducedby43%by2030.6HowAIcanenableasustainablefuture,PwC/Microsoft,2019.7Howartificialintelligencecanpoweryourclimateactionstrategy,Capgemini,2020.HowAICanHelpAccelerateClimateAction8Weincludeavoidedemissionsinouranalyses.Forexample,how—byidentifyingnewtransitionpathwaysfordevel-opingcountries—theycangrowtheireconomieswithfeweremissions,thereby“avoiding”emissionsthatwouldhavebeengeneratedhadtheysimplypursueda“businessasusual”approachtogrowth.9Machinelearningcanboostthevalueofwindenergy,GoogleBlog,February2019.10Newwayswe’rehelpingreducetransportationandenergyemissions,GoogleBlog,October2023.11GoogleEnvironmentalReport2023,p.6.12Ibid.,p.28.13HowAIishelpingairlinesmitigatetheclimateimpactofcontrails,GoogleBlog,August2023.14Rogelj,J.etal.,MitigationPathwaysCompatiblewith1.5°CintheContextofSustainableDevelopment,inGlobalWarmingof1.5°C.AnIPCCSpecialReportontheimpactsofglobalwarmingof1.5°Cabovepre-industriallevelsandrelatedglobalGHGemissionpathways,inthecontextofstrengtheningtheglobalresponsetothethreatofclimatechange,sustainabledevelopment,andeffortstoeradicatepoverty,2018.15TCFDInsightsSeriesS&P500,CarbonDisclosureProject,September2022.CarbonDisclosureProject(CDP),anonprofitcharitableorganization,operatesaworldwidedisclosureplatformthatempowersinvestors,corporations,municipalities,andgovernmentalregionstoeffectivelyoverseetheirecologicalfootprints.16ZvikaBen-HaimandOmerNevo,Real-timetrackingofwildfireboundariesusingsatelliteimagery,GoogleResearch,2023.17HowAIisimprovingagriculturesustainabilityinIndia,GoogleBlog,January2023.18Newwayswe’rehelpingreducetransportationandenergyemissions,GoogleBlog,October2023.19Ibid.,GoogleBlog,October2023.20Howwe’reusingAItocombatfloods,wildfiresandextremeheat,GoogleBlog,October2023.NavigatingAI'sPotentialRisks21IEAanalysis(accessedinSeptember2023)basedonMasanetetal.(2020),Malmodin(2020),Hintemann&Hinterholzer(2022)andreportedenergyusedatafromlargedatacenteroperators.22Kaack,L.H.etal.,Aligningartificialintelligencewithclimatechangemitigation,NatureClimateChange,12,518–527,2022.23Masanet,E.etal.,Recalibratingglobaldatacenterenergy-useestimates,Science,2020.24GoogleEnvironmentalReport2023,p.38.25Patterson,D.etal.,TheCarbonFootprintofMachineLearningTrainingWillPlateau,thenShrink,2022.26ElectricGridCarbonintensitybasedonEmber’sYearlyElectricityData;Ember’sEuropeanElectricityReview;Ener-gyInstituteStatisticalReviewofWorldEnergy,2022.27BasedonIEApublication,Renewableenergysection,accessedinSeptember2022.41ACCELERATINGCLIMATEACTIONWITHAI28GoogleDeepmindBlogpost,Machinelearningcanboostthevalueofwindenergy,February2019.29Patterson,D.etal.,CarbonEmissionsandLargeNeuralNetworkTraining,2021.30Patterson,D.etal.,TheCarbonFootprintofMachineLearningTrainingWillPlateau,thenShrink,2022.31ThecomputingpowerneededtotrainAIisnowrisingseventimesfasterthaneverbefore;MITTechnologyReview,2019;WhatIsaTransformerModel?NVIDIA,2022.32So,DavidR.etal.,Primer:SearchingforEfficientTransformersforLanguageModeling,2021.33IBMGlobalAdoptionIndex,2022.34NVIDIAestimatesthat80-90%oftheAI/MLworkloadisforinference.AmazonWebServicesestimatesthat90%ofAI/MLdemandinthecloudisforinference.35GoogleEnvironmentalReport2023.p.50.36Siddik,M.etal.,TheenvironmentalfootprintofdatacentersintheUnitedStates,EnvironmentalResearchLetters,16064017,2021.37Totalannualdatacenterswaterconsumptionestimatedto626billionliters,ofwhich95–100billionlitersusedon-siteforcooling;Shehabi,Arman,Smith,Sarah,Sartor,Dale,Brown,Richard,Herrlin,Magnus,Koomey,Jonathan,Masanet,Eric,Horner,Nathaniel,Azevedo,Inês,&Lintner,William.UnitedStatesDataCenterEnergyUsageReport.UnitedStates.Figure25.38Ourcommitmenttoclimateconsciousdatacentercooling.GoogleBlog.November,2022.39GoogleEnvironmentalReport2023,p.95.40BasedonUSGSestimationsofwateruseintheUnitedStatesin2015:USAEnvironmentalProtectionAgency,Watersense,statisticsandfacts.Basedonanaveragehouseholdsizeof2.5personswitheachAmericanusinganaverageof82gallonsofwateradayathome.41Meta,PublicWaterReporting:ExpandingtheOperatingEnvelope.42AzureMicrosoft,LearnhowMicrosoftCircularCentersarescalingcloudsupplychainsustainability,2022.43GoogleEnvironmentalReport2023,p.57.44GoogleAIPrinciples.45Accra.AIforClimate:ASummaryofCriticalPolicyOutcomes46Adatacatalogfunctionsasacomprehensivedatainventory.Itmaintainsarecordofallaccessibledata,helpingus-ersswiftlylocatespecificinformation.47TheAffordabilityofICTServices2022,ITUPolicyBrief,2022.48ArtificialIntelligenceneedsassessmentsurveyinAfrica,UNESCO,2021.49StrengtheningAIGovernance:UNESCOPilotsitsCapacityBuildingFrameworkonDigitalTransformationinAfricaandIndia,UNESCONews,July2023.50Seow,P.etal.,EducationalPolicyandImplementationofComputationalThinkingandProgramming:CaseStudyofSingapore,achapterinComputationalThinkingEducation,Kong,SC.andAbelson,H.,editors,Springer,Singapore.51Decarbonization:AWin-WinfortheEconomyandEcology,GermanyWorks.52DigitalDecarbonizationGermany,AcceleratingGermany’sclimateprotectionwithefficientdigitalsolutions,Imple-mentConsultingstudycommissionedbyGoogle,2023.53Gailhofer,P.etal.,TheroleofArtificialIntelligenceintheEuropeanGreenDeal,astudyforthespecialcommitteeonArtificialIntelligenceinaDigitalAge(AIDA),PolicyDepartmentforEconomic,ScientificandQualityofLifePolicies,EuropeanParliament,Luxembourg,2021.54“Concentrationdynamicsofcoarseandfineparticulatematteratandaroundsignalisedtrafficintersections,”Envi-ronmentalScience:Processes&Impacts,2016.BOSTONCONSULTINGGROUPGOOGLE42AbouttheAuthorsBCGHamidMaherisamanagingdirectorandpartnerinAmaneDannouniisamanagingdirectorandpartnerinBCG’sCasablancaoffice.HeheadsBCGXinAfricaandBCG’sCasablancaofficeandleadsthefirm’sdigitalandisafoundingandsteeringcommitteememberofthetechnologypracticeinAfrica.OneofhisareasoffocusisAIforthePlanetAlliance—andleadsBCG’sengagementthenexusbetweentechnologyandsustainabledevelop-asthealliance’sknowledgepartner.Hecanbereachedatmentatthecompany,country,andregionallevels.Hecanmaher.hamid@bcg.com.bereachedatdannouni.amane@bcg.com.EdmondRhysJonesisapartnerandassociatedirectorStefanA.DeutscherisapartneranddirectorinBCG’sinBCG’sLondonoffice.Heco-leadsBCG’sCenterforBerlinoffice.HeisacorememberofBCG’sTechnologyClimateandSustainabilityPolicyandRegulationandisaAdvantageandTechnology,Media&TelecommunicationscorememberoftheClimateandSustainabilitypractice.practices.HeisBCG’sglobaltopicleaderforITinfrastruc-Hecanbereachedatrhysjones.edmond@bcg.com.tureanddatacenteroperation.Hecanbereachedatdeutscher.stefan@bcg.com.HamzaTberisanassociatedirectorinBCG’sAliZiatisaprincipalspecializingindatascienceCasablancaoffice.PriortojoiningBCG,hewasseniorbasedinBCG’sCasablancaoffice.HeholdsaPh.D.programofficer,UnitedNations,ExecutiveOfficeoftheinAIandMachineLearningandhasbeeninvolvedSecretaryGeneral,ClimateActionteam.Hecanbeinthedevelopmentofseveralclimate-relatedAItools.reachedattber.hamza@bcg.com.Hecanbereachedatziat.ali@bcg.com.AmjadKharijisaprojectleaderinBCG’sCasablancaGhitaDezzazisanassociateinBCG’sCasablancaoffice.office.Hecanbereachedatkharij.amjad@bcg.com.Shecanbereachedatdezzaz.ghita@bcg.com.43ACCELERATINGCLIMATEACTIONWITHAIGoogleAdamElmanistheheadofSustainabilityforGoogleAndrewHylandisaseniormanagerinGoogle’sGovern-Europe,MiddleEast,andAfrica,andleadsthecompany’smentAffairsandPublicPolicyteamworkingonAI.sustainabilityworkacrosstheregion.MarsdenHannaistheheadofSustainabilityandDavidPattersonisadistinguishedsoftwareengineeratClimatePolicyatGoogle,whereheleadsthecompany’sGoogle,workingondomain-specificcomputerarchitec-engagementwithgovernmentsonsustainability,climate,turesformachinelearning.andenergypolicyissues.MaudTexieristheglobaldirectorofCleanEnergyJulietRothenbergisthegroupproductmanagerofCli-andDecarbonizationDevelopmentatGoogle.ShemateAIatGoogleResearch,leadingproductmanagementleadsateamresponsiblefordevelopingandscalingforClimateAIinitiativeswithinGoogleResearch.24/7carbon-freeenergyforGoogle’sglobalinfrastructureworldwide.AntoniaGawelistheglobaldirectorofSustainabilityandPartnershipsatGoogle,responsibleforfurtheringGoogle’sglobalclimateandsustainabilitygoalsthroughpartner-ship.ForFurtherContactIfyouwouldliketodiscussthisreport,pleasecontacttheauthors.BOSTONCONSULTINGGROUPGOOGLE44AcknowledgementsTheauthorswouldliketoexpresstheirgratitudetothemanyleadingtopicexpertsonAI,climate,technology,datacenters,andpolicywhograciouslysharedtheirtimeandperspectiveswithus.Withouttheircontributions,thisreportwouldnothavebeenpossible.Youcanfindthefulllistofintervieweesbelow.IntervieweeCompanyTitleJacquesAmselemAlboClimateCEOandCo-FounderJenBennettGoogleTechnicalDirector,Sustainability,KateBrandtGoogleOfficeoftheCTOWilliamBrentHuskPowerSystemsChiefSustainabilityOfficerHelenElizabethBurdettWorldEconomicForumChiefMarketingOfficerMarkCaineGoogleHeadofTechnologyforEarthFrançoisCandelonBCGSeniorLead,Energy&ClimateCharlotteDegotCO2AIGlobalDirector,BCGHendersonInstituteValElbertBCGCEO&FounderCharlotteGastineauBCGManagingDirectorandPartner,Telecom&SarahGoodmanBCGMediaGrowthStrategy,DigitalTransformationAlonHarrisGoogleAssociateGemmaJenningsGoogleDeepMindPartnerandAssociateDirector,VilmaKazaGooglePublicSectorJustinKeebleGoogleProgramManager,Research&SearchAshleyKing-BischofSproutProductManagerShivamKishoreUNEnvironmentProgramEUSustainabilityandClimatePolicyLeadNaomieLecardAlboClimateManagingDirectorforGlobalSustainabilityBrian(Juhyuk)LeeGoogle.orgFounder,CEO,CPTORonitLevaviMoradGoogleSeniorAdvisor,DigitalTransformation,SustainabilityHeadofBusinessDevelopmentSustainabilityteamDirector,ProgramManagement,Research&Search45ACCELERATINGCLIMATEACTIONWITHAIIntervieweeCompanyTitleMathieuMarcotteCEIMIADirector,AIEcosystemMobilizationandTravisMcCoyGoogleSpecialProjectsNicolasMiailheTheFutureSocietyStevenMillsBCGDirector,ProductManagement,Climate&SustainabilityReinaOtsukaUNDevelopmentProgramJohnPlattGoogleFounderandPresidentGolestan(Sally)RadwanUNEnvironmentProgramKirstenRulfBCGChiefAIEthicsOfficer,LeadforAIinPublicLeonorSaitkoulovBCGSectorDuncanSmithGoogleDeepMindBenTownsendGoogleDigitalInnovationLeadforNature,Climate,JosephWegenerWorldEconomicForumandEnergyMikeWernerGoogleSimsWitherspoonGoogleDeepMindDistinguishedScientistDavidYoungBCGChiefDigitalOfficerAshley(Schoettle)ZlatinovGooglePartnerandAssociateDirector,Data&DigitalizationSeniorDataScientistExternalCommsSeniorManagerGlobalHeadofDataCenterSustainabilityProjectFellowGlobalHeadofCircularEconomyClimateActionLeadManagingDirector&SeniorPartner,GlobalTopicLeaderforTotalSocietalImpact/SustainabilityAIPolicyLead,SocialImpactBOSTONCONSULTINGGROUPGOOGLE46ReferencesTitleSponsor(s)GeneralresourcesonAIanditspotentialforclimateactionDigitalDecarbonisation:HowthedigitalsectorisAGoogle-commissionedreportfromImplementConsultingsupportingclimateactionGroup,2022ClimateAI:HowartificialintelligencecanpowerAreportfromCapgemini,2020yourclimateactionstrategyTheroleofArtificialIntelligenceintheEuropeanAreportfromtheEuropeanParliament’sspecialcommitteeGreenDealonArtificialIntelligenceinaDigitalAge,May2021EnvironmentalandethicalconsiderationsMeasuringtheenvironmentalimpactsofartificialAreportfromtheOECDincollaborationwithTheGlobalintelligencecomputeandapplications:TheAIPartnershiponArtificialIntelligence,2022footprintSustainableAI:EnvironmentalImplications,Anacademicarticle,FacebookAI,2022Challenges,andOpportunitiesThePresidioRecommendationsonResponsibleAreportfromtheWorldEconomicForumincollaborationGenerativeAIwithAICommons,2023RecommendationontheEthicsofArtificialAreportfromUNESCO,2021IntelligenceAPolicyRoadmapfor24/7Carbon-FreeEnergyAreportfromGoogle,2022AIforclimateinthecontextofpolicymakingClimateChangeandAI:RecommendationsforAreportfromtheGlobalPartnershiponAIincollaborationGovernmentActionwithClimateChangeAIandtheCentreforAI&Climate,November2021ITI’sGlobalAIPolicyRecommendationsAreportfromtheInformationTechnologyIndustryCouncil,2021BiodiversityandArtificialIntelligence:AreportfromTheGlobalPartnershiponArtificialIntelligence,Opportunities&RecommendationsforAction2022FinalreportonbridgingdatagapsAtechnicaldocumentfromtheNetworkforGreeningtheFinancialSystem,2022DataadaptationatdifferentandtemporalscalesAtechnicalpaperfromtheAdaptationCommitteeoftheUnitedNationsFrameworkConventiononClimateChange,theKyotoProtocol,andtheParisAgreement,2020AIExcellence:EnablingconditionsforAI’sAreportfromtheEuropeanCommission,layingouttheEU’sdevelopmentanduptakecoordinatedplanonartificialintelligence,2021ArtificialIntelligenceinthePublicSectorAreportfromtheGovTechGlobalPartnershipincollaborationwiththeWorldBank,2020BeijingConsensusonArtificialIntelligenceandAreportfromUNESCO,2019EducationArtificialIntelligenceandDigitalTransformation:AreportfromUNESCO,November2022CompetenciesforCivilServantsStakeholdersforaCohesiveandSustainableAbriefingpaperfromtheWorldEconomicForum,JanuaryWorld:TheRoleofLighthouseProjects2020ArtificialIntelligenceandInternationalAffairs:AreportfromChathamHouse,2018DisruptionAnticipated47ACCELERATINGCLIMATEACTIONWITHAIAddCo-SponsorlogohereBostonConsultingGrouppartnerswithleadersUciamvoloraditatur?Aximvoloreribusmoluptatiinbusinessandsocietytotackletheirmostautetharioquianustfaciisreperrovitatiaimportantchallengesandcapturetheirgreatestdipsandeliasitlaborum,quassitio.Itasvolutemopportunities.BCGwasthepioneerinbusinessesnullesutfaccusperchiliatidoluptatur.Estiunt.strategywhenitwasfoundedin1963.Today,Eteiuminumetdolumeteteosexeumharchicweworkcloselywithclientstoembraceateceserrumnateminranisquiadisimi,omniatransformationalapproachaimedatbenefitingallverormolorerionsedquiaeseveliquiatiusstakeholders—empoweringorganizationstogrow,sundaeporeiumetetillesciatibeaturautquebuildsustainablecompetitiveadvantage,andconsequiaautassumfugitquiautexcepudit,drivepositivesocietalimpact.omniavoloratur?Expligendeliaecturmagnam,queexpedignistexetvoluptaquam,officibernamOurdiverse,globalteamsbringdeepindustryandatquidemveliusnus.functionalexpertiseandarangeofperspectivesthatquestionthestatusquoandsparkchange.NemfaccaboresthillamendiadoluptaeBCGdeliverssolutionsthroughleading-edgeconseruptateinimvolesequidmolumquam,managementconsulting,technologyanddesign,consequeconsedipithillabo.Imaioevelenditiumandcorporateanddigitalventures.Weworkinaharibus,conreicturautemost,vendamamellaniauniquelycollaborativemodelacrossthefirmandestrundemcorepudaderroremporrumquat.throughoutalllevelsoftheclientorganization,fueledbythegoalofhelpingourclientsthriveandenablingthemtomaketheworldabetterplace.Forinformationorpermissiontoreprint,pleasecontactBCGatpermissions@bcg.com.TofindthelatestBCGcontentandregistertoreceivee-alertsonthistopicorothers,pleasevisitbcg.com.FollowBostonConsultingGrouponFacebookandX(formerlyknownasTwitter).©BostonConsultingGroup2023.Allrightsreserved.11/23

1、当您付费下载文档后，您只拥有了使用权限，并不意味着购买了版权，文档只能用于自身使用，不得用于其他商业用途（如 [转卖]进行直接盈利或[编辑后售卖]进行间接盈利）。
2、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。
3、如文档内容存在违规，或者侵犯商业秘密、侵犯著作权等，请点击“违规举报”。

碎片内容

碳中和的最新文档