3September2024|12:02PMEDT

OilAnalyst

PotentiallyNegativeEffectsofAIonOilProductionCostsandOilPrices

n

n

n

n

_

n

Energy?rmsnowmentionAImorefrequentlyonearningscallsanduseAImorethanthemedian?rmintheeconomy.ThedebateontheimpactofAIonenergyandmetalshasmostlyfocusedonthedemandsidegiventheexpectedboosttopowerdemand.However,wehighlightthatAImayeventuallyweighonoilpricesoverthenextdecadebyboostingoilsupplythroughtwomainchannels.

First,AIcouldpotentiallyreducecostsviaimprovedlogisticsandresource

allocation.Weestimatethatc.30%ofthecostsofanewshalewellcould

potentiallybereducedbyAI,resultinginac.$5/bblfallinthemarginalincentiveprice,assuminga25%productivitygainobservedforearlyAIadopters.

Second,AIcouldpotentiallyincreasetheultimatelyrecoverableresourcebase.AhypotheticalAI-catalyzed10-20%jumpintheverylowrecoveryfactorsofUS

shalecouldincreaseoilreservesby8-20%(10-30bnbarrels).

Incontrast,weestimatearelativelymodestpotentialAIboosttooildemandis

viahigherincomesof+0.7mb/doverthenext10years,whichcouldraise

long-datedoilpricesby$2/bbl.ThisAIimpactonoildemandismodestcomparedtotheAIdemandimpacttopowerandnaturalgas,andpalesincomparisonto

thenegativeestimatedeffectsonoildemandfromEVs(c.-8mb/d)andlowernaturalgasprices(-2mb/d)overthenext10years.

Takentogether,webelievethatAIwouldlikelybeamodestnetnegativetooilpricesinthemedium-to-longtermasthenegativeimpactfromthecostcurve(c.-$5/bbl)-oil’slong-termanchor-wouldlikelyoutweighthedemandboost

(c.+$2/bbl).This?ndingreinforcesthenegativelong-runoilpriceimplicationsfromourworkon:(1)2024TopProjects,(2)China’s3S’s,and(3)theLNGbearcycle.

CallumBruce,CFA

+1(212)902-3053|callum.bruce@GoldmanSachs&Co.LLC

DaanStruyven

+1(212)357-4172|

daan.struyven@

GoldmanSachs&Co.LLC

YuliaZhestkovaGrigsby

+1(646)446-3905|yulia.grigsby@GoldmanSachs&Co.LLC

Investorsshouldconsiderthisreportasonlyasinglefactorinmakingtheirinvestmentdecision.ForRegAC

certi?cationandotherimportantdisclosures,seetheDisclosureAppendix,orgoto/research/hedge.html.

GoldmanSachsOilAnalyst

Afterinitiallylagging,USenergy?rmsareincreasinglymentioningAIintheirearningscalls

ProportionofS&P500?rmsmentioningAIduringquarterlyearningscalls

Lowerlong-runoilproductioncostslikelyoutweighthe

income-relatedAIboosttooildemandoverthenextdecade

BrentcrudeoilpriceimpactsofdifferentAIchannels(USD/bbl)

ProportionofS&P500firmsmentioning"AI"duringquarterlyearningscalls70%

60%

50%Energy

40%

S&P500

30%20%10%

0%

2015201620172018201920202021202220232024

3

2

1

0

-1

-2

-3

-4

-5

-6

20292034

AIincomeboostLowercost-curve

Brentcrudeoilpriceimpact

Source:Companyreports,GoldmanSachsGlobalInvestmentResearch

WethinkthelowercostcurveeffectsofAIarelikelytobemorefrontloadedthanthedemand-sideeffects

Source:GoldmanSachsGlobalInvestmentResearch

PotentiallyNegativeEffectsofAIonOilProductionCostsandOilPrices

TheenergysectorhadinitiallylaggedinearningscalltranscriptmentionsofAIaswellasinbusinesssurveydataonthecurrentandexpectedoperationaluseofAIover2023.

Theearningscallgapdisappearedin2024astheimplicationsofarti?cialintelligence(AI)forpowerdemandandelectri?cationcameintofocusforcommoditymarkets.Thiswastrueforbothenergy(powerandnaturalgas)andmetals(copper,aluminium,andthe

batterymetals).

WeexpecttheriseofAItoreinforcetheelectri?cation-drivenin?ectioninglobalpowerdemandbeyondthatseeninhistoricalGDP-consumptionpatterns(Exhibit1),

_

fundamentallydivergingfromdecadesofstagnation.Assuch,theenergysectorisnowoneofthemostintensesectorsoftheAIdebate(Exhibit2).

Whiletheimpactsareclearlybullishforpowerdemandandprices,webelievetheboosttonaturalgasdemandismodestinthecontextofthecomingbearishLNGsupplywave.However,whataretheimplicationsofAIforcrudeoilprices?

InthisOilAnalystweanswerthesequestionsbyseparatingtheimpactsonoilintoitscon?ictingimpactsonthedemandside(positivefordemandandprices)andsupplyside(positiveforsupply,demandforprices).

3September20242

GoldmanSachsOilAnalyst

Exhibit1:OurEquityAnalystsForecastThatUSPowerConsumptionGrowthWillOutpaceGDPGrowthThrough2030fortheFirstTimeinThreeDecades

AveragegrowthratesofUSpowerconsumptionandrealGDP,bybusinesscycle,%

Exhibit2:Afterinitiallylagging,USenergy?rmsareincreasingmentioningAIintheirearningscalls

ProportionofS&P500?rmsmentioningAIduringquarterlyearningscalls

4.0%

3.5%

3.0%

2.5%

2.0%

1.5%

1.0%

0.5%

0.0%

PowerConsumptionGDP

1983-19911992-20012002-20092010-20202021-20232022-2030

ProportionofS&P500firmsmentioning"AI"duringquarterlyearningscalls70%

60%

50%Energy

40%

S&P500

30%20%10%

0%

2015201620172018201920202021202220232024

EachUSbusinesscyclestartswiththeexpansionandendswiththerecession,exceptforthecurrentcycle,whichusesrealizeddatafrom2021to2023andGSforecastsafterwards(2030istheendofthepowerconsumptiongrowthforecasts).

Source:EIA,HaverAnalytics,GoldmanSachsGlobalInvestmentResearch

Source:Companyreports,GoldmanSachsGlobalInvestmentResearch

WehighlightthatAIcouldpotentiallyhavesomewhatnegativeimpactsonoilpricesviathesupply-sidethroughtwomainchannels.

LowerLong-RunCosts

First,AIcouldlowertheindustrycostcurve-itslong-runanchor-viareducedcostsfromfurtherimprovinglogisticsandresourceallocation.

Afterinitiallylaggingthemediansector,energy?rmsareincreasinglyexpectingtoleverageAIintheirfutureoperations1,bringingintofocusthesupply-sideofthe

equation.

Forexample,NaborsIndustrieswasabletoreducedrillingtimesofsomeNorthDakota(Bakken)rigsby30%withitspartnershipwithCorvaLLCfocusedonpredictiveand

_

automateddrilling.Elsewhere,BPwasabletosigni?cantlycutoverheadsforsoftwaredevelopersduetoAI-enhanceddeveloperproductivitytools.

Consistentwiththis,newsurveydatafromtheDallasFedEnergySurveysuggestc.20-40%ofUSoilandgas?rmsarecurrentlyplanningtoleverageAItooptimisesupplychainsandautomateprocessesaswellasimprovedrillingandcompletionactivities(Exhibit3).

Drillingandcompletiontimeshavealreadyfallenc.20%and40%respectivelysince2019,withroomtofallfurther,especiallyforcompletiontimes(Exhibit4).This

mechanicallymeanslessdayratesspentonexpensiverigs.Optimisationofotherprocessescouldsimilarlyreduceothercostoverheads.

1ThiscanbeseeninUSCensusWeeklyBusinessSurveyDatawithinitsSupplementalAIsurvey.

Speci?callythequestion,“DoyouthinkthisbusinesswillbeusingAIinproducinggoodsorservicesinthenext6weeks”fortheMining/Oil&Gasextractionsectorversusthenationalaverage.

3September20243

GoldmanSachsOilAnalyst

Exhibit3:AIisexpectedtobefocusedonreducingcosts,optimisingsupplychains,andavoidingdowntime

DallasFedAIsurveyresponsesto:Howisyour?rmusingorplanningtouseAI?(Oilandgas?rmrespondents)

Exhibit4:Fullcyclewelltimesstillhaveroomtofall,althoughwillfacephysicallimits

AveragecycletimefromspudtoproductionforUSPermianShalewell

70

60

50

40

30

20

10

0

HowisyourfirmusingorplanningtouseAI?

Supply-chainoptimization

Drillingandcompletion

Accounting

Process

automation

Predictivemaintenance

Geologyor

reservoir

engineering

Businessanalysisorpredictiveanalytics

Increasedproductivity/reducedcostsIncreasedrevenueOther

Days

100

90

80

70

60

50

40

30

20

10

0

AverageBuildingTimeforaPermianWell

ProductionLaunchCompletion

AwaitingCompletionDrilling

20192024

Days

100

90

80

70

60

50

40

30

20

10

0

PredictivemaintenancecanalsoincreaseproductivityandreducecostsSource:DallasFed,GoldmanSachsGlobalInvestmentResearch

Source:EIA,Enverus,Platts,GoldmanSachsGlobalInvestmentResearch

Weestimatethatc.30%ofthecostsassociatedwithanewshalewellcouldpotentially

bereducedbyAI(Exhibit5),withtheothercostslargelydeterminedbyphysicalcommodityrequirements(cement,sand,?uids,etc)whichcannotbesigni?cantlystreamlined.

Assuminga25%productivitygain,asestimatedforearlyadoptersbyourGlobal

Economicscolleaguesusingacademicliteratureandcorporateanecdotes(Exhibit6),

thiswouldresultinapotential-7%fallintotalwellcosts.2Applyingthisinturntoour

c.$70/bblestimate(onaBrentcrudeoilbasis)ofUSshalemarginalcostsamountstoa$5/bblfallinthecostofoil’smarginalsupplier,andinthelong-runanchorofoilprices.3

Thisoptimisationprocesswilleventuallyreachitsphysicalconstraintsasdrillingspeedsreachtheirnaturallimit.Theremayalsobesomeoffsettingimpactsfrompotential

tightnessinUSpowermarkets.

_

2Oureconomists’conservativelyassumea15%cumulativeproductivityuplift(ratherthan25%)intheirbaselineestimates.The25%utilisedherebetteralignswithcorporateanecdotesintheenergyspace.

3Admittedly,fromourdiscussions,smaller?rmswhichmaybetterde?nethemarginalcostmaybelesslikelytoexploitAI.But,equally,theymayhaveevenmoretogainfromdoingso.

3September20244

GoldmanSachsOilAnalyst

Exhibit5:Weestimatec.30%ofUSshalecostscouldbereducedbyimplementingAI

USPermianshalewellcostbreakdownbypotentialAIimpact($m)

Exhibit6:AIhasgeneratedac.25%improvementinlabourproductivityonaverage

EstimatedeffectofgenerativeAIadoptiononlabourproductivity

8

7

6

5

4

3

2

1

0

AIunaffectedAIaffected

25%efficiencygaininAIaffectedcategories

30%

25%

20%

15%

10%

5%

0%

MedianAverageMedianAverage

AcademicStudiesCompanyAnecdotes

EffectofGenerativeAIAdoptiononLabourProductivity:Estimates

Source:Woodmac,GoldmanSachsGlobalInvestmentResearchSource:GoldmanSachsGlobalInvestmentResearch

APotentialBoosttoRecoverableResources

Second,byacceleratingthelearning-by-doingprocess,AIcouldincreasetheamountofpro?tablyrecoverableresources,ashifttotherightoftheindustrycost-curve.

TheDallasFedsurveyandcorporateanecdotessuggestthatwhiletheprimary

expectedbene?tsarefocusedonreducingcosts,therearealsosomeproductionand

revenueenhancingbene?tsofAI(Exhibit3).Predictivemaintenancehasthepotentialtoreduceproductiondowntime.Meanwhile,improvedgeological/reservoirengineering

hasthepotentialtoboostproductionandreserves.

Forexample,VitalEnergyincreasedcompany-levelproductionby2-3%usingautomated

processesforsubmersiblepumpsleveragingAI.ContangoResourcesincreased

operationalperformancebyautonomouslyoptimisingarti?cial(naturalgas)lift.Baker

HugheshaspartneredwithC3.aitouseAI-poweredtoolstooptimiseproduction,whileSchlumbergeraimstohave15%ofalloilwellsautonomouslycontrolledbyAIinthe

_

next3-5years.Elsewhere,ShearFracaimstouseAItofrackwellsmoreef?ciently,

enhancingrecoveredmolecules.Lastly,ADNOCrecentlyannouncedthatithadraised

outputatitsoffshoreSARB?eldby25%(c.30kb/d)usingAI-basedtechnology.

WeinvestigateahypotheticalimprovementintheverylowrecoveryfactorsofUSshale,whicharecurrentlyjustafractionofthatobservedinconventionalresources(Exhibit7).AhypotheticalAI-catalyzed10-20%jumpinwellproductivityandhigherrecoveryratescouldincreaseremainingresourcesby8-20%(10-30bnbarrels),basedonthecurrent

industrycostcurveandourlong-termBrentcrudeoilpriceassumption(Exhibit8).

Althoughthisisdif?culttoquantifyatthegloballevel,thisimpactchannelofAIwould

increasetheultimatelyrecoverableresourcebase,delayingfurtherthepeakofUSshalesupply,andfurtherslowingapotentialdrawdownofelevatedOPEC+sparecapacity.

3September20245

GoldmanSachsOilAnalyst

Exhibit7:USShalehasmuchlowerresourcerecoveryfactorsthanconventionaloilwells,leavingroomforsigni?cantimprovement

Cumulativedensityfunctionofglobaloilwellsbyresourcerecoveryfactor

Exhibit8:Improvedrecoverycouldunlock10-30bnbarrelsofoilresourcesatour$70-80/bblLTBrentpriceassumption

USL48oilresourcesbyBrentandrecoveryfactorimprovementassumption

100%90%80%70%60%50%40%30%20%10%

0%

03691215182124273033363942454851545760636669Recoveryfactor

ShaleCountResourceweightedMedian

200

180

160

140

120

100

80

60

40

20

0

60708090

USCostCurve(Brentbasis)10%improvementinrecovery20%improvementinrecovery

Coversc.1400?elsrepresenting1.2trnbarrelsofrecoveredreservesSource:Woodmac,GoldmanSachsGlobalInvestmentResearch

Source:Woodmac,GoldmanSachsGlobalInvestmentResearch

AModestBoosttoLong-RunOilDemand

We?ndthatthedemand-sideimpactsarerelativelymodestforoil,incontrasttopowerandnaturalgasmarkets.

Asoilisscarcelyusedasasourceofenergyforpowergenerationtoday,theprimarychannelthatAIwillboostoildemandisviahigherincomes.

OurGlobalEconomicscolleaguesestimatethatAIwillliftglobalGDPgrowthbyan

annualised0.2ppbytheearly2030sonanoildemand-weightedbasis,amountingtoacumulative1%boosttothelevelofglobalGDPoverthenextdecade.4UsingaslightlydecliningGDP/incomeelasticityovertime,weestimatethiswouldboostoildemandby+0.2/+0.7mb/doverthenext5/10yearsrespectively(Exhibit9).

Ratherthanassumethisimpactleadstoanever-wideningde?citandanaccelerating

_

inventorydepletion,weinsteadassumethatsuchanimpactwouldbesteadilyoffsetbyhigherOPEC+productionandlowersparecapacity.Leveragingourpriorestimates,thisdecreasedsparecapacitywouldboostlong-datedoilpricesby$1-2/bblrespectivelyover5-10years.

Weacknowledgethatthispositivedemandimpulsewillcontinuetodeepenovertime,asAIpermeatesglobally.5Nevertheless,theimpactstillpalesincomparisontothe

estimatednegativeimpactsfromEVs(c.-8mb/d)andlowernaturalgasprices(-1.5to-2mb/d)overthenextdecade(Exhibit10).

4TheOECDboostisclosertoacumulative2%overthenextdecade.

5MuchoftheexpectedGDPandoildemandupliftfallsbeyondourforecasthorizon,althoughouradoptiontimelineisonthemoreconservativeside,withmostmarketparticipantsexpectingamorefrontloaded

adoptioncurve.Nevertheless,we’dexpectthebroaderbullishmacrodemandbene?tstolagthebearishsupply-sidebene?ts.

3September20246

GoldmanSachsOilAnalyst

Exhibit9:TheGDPgrowthboostduetoAIwillboostoildemandby

0.2mb/din?veyears,0.7mb/dintenyears

AIupgradestoOil-weightedGDP(cumulative,%,lhs)andglobaloildemand(cumulative,mb/d,rhs)

Exhibit10:TheAIincomeboosttooildemandismodestcomparedtothedragsfromEVsandtheLNGbearcycle

Cumulativeoildemandimpact(2034versus2024,mb/d)

1.2%

1.0%

0.8%

0.6%

0.4%

0.2%

0.0%

20242025202620272028CumulativeWorldGDPImpact

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

202920302031203220332034

CumulativeWorldOilDemandImpact(mb/d)

2

0

-2

-4

-6

-8

-10

AIincomeboostOiltonaturalgasswitchingEVs

Cumulativedemandimpactby2034

Source:GoldmanSachsGlobalInvestmentResearchSource:GoldmanSachsGlobalInvestmentResearch

TheMoreThingsChange,TheMoreTheyStayTheSame

Takentogether,webelievethatAIwouldbeamodestnetnegativetooilpricesinthemedium-to-longtermasthenegativeimpactfromthecostcurve(c.-$5/bbl)-oil’s

long-termanchor-wouldoutweighthepositiveimpactsfromthedemandside(c.+$1-2/bbl).

WerecentlyloweredourexpectedBrentcrudeoilpricerangeby$5/bblto$70-85/bblfrom$75-90/bblpreviously.Weakerrealizedandexpectedfundamentalswarrantedpartoftherevisionviaweakertimespreads.However,alowerback-endassumptionon

continuedimpressiveef?ciencygainsinUSshalealsoplayedakeyrole.Assuch,AI

proliferationmerelyrepresentsacontinuationoftheenormousproductivityimprovementsobservedinthisindustryoverthelastfewdecades.

Long-cycleunderinvestmentremainsrelativelylow.Nevertheless,thepotential

_

unlockingofmorerecoverableresources(inUSshaleorevenelsewhere)couldfurtherdelayitspotentiallybullishconsequencesbeyondthecurrentpostponementdueto

elevatedsparecapacity.Thisfurtherreinforcesthebearishmediumtermoilprice

implicationsfromourworkon(1)2024TopProjects,(2)China’s3S’s,and(3)theLNGbearcycle.

3September20247

GoldmanSachsOilAnalyst

Exhibit12:Lowupstreamcapitalexpendituresisanecessarybutinsuf?cientconditionforhigherlong-runoilprices

Measuresofoilmarkettightness(percentiles)

3

2

1

0

-1

-2

-3

-4

-5

-6

20292034

AIincomeboostLowercost-curve

Brentcrudeoilpriceimpact

WethinkthelowercostcurveeffectsofAIarelikelytobemorefrontloadedthanthedemand-sideeffects

Exhibit11:OnbalancewethinkpotentialAIimpactsoncrudeoilpriceswillbenegative

PotentialAIimpactsonBrentcrudeoilprices(USD/bbl)

Source:GoldmanSachsGlobalInvestmentResearch,IEA,Platts,OPEC,EIA,GGIS/IMF,WorldBank

Source:GoldmanSachsGlobalInvestmentResearch

_

3September20248

GoldmanSachsOilAnalyst

DisclosureAppendix

RegAC

I,CallumBruce,CFA,herebycertifythatalloftheviewsexpressedinthisreportaccuratelyre?ectmypersonalviews,whichhavenotbeenin?uencedbyconsiderationsofthe?rm’sbusinessorclientrelationships.

Unlessotherwisestated,theindividualslistedonthecoverpageofthisreportareanalystsinGoldmanSachs’GlobalInvestmentResearchdivision.

Disclosures

Regulatorydisclosures

DisclosuresrequiredbyUnitedStateslawsandregulations

Seecompany-speci?cregulatorydisclosuresaboveforanyofthefollowingdisclosuresrequiredastocompaniesreferredtointhisreport:managerorco-managerinapendingtransaction;1%orotherownership;compensationforcertainservices;typesofclientrelationships;managed/co-managedpublicofferingsinpriorperiods;directorships;forequitysecurities,marketmakingand/orspecialistrole.GoldmanSachstradesormaytradeasa

principalindebtsecurities(orinrelatedderivatives)ofissuersdiscussedinthisreport.

Thefollowingareadditionalrequireddisclosures:Ownershipandmaterialcon?ictsofinterest:GoldmanSachspolicyprohibitsitsanalysts,professionalsreportingtoanalystsandmembersoftheirhouseholdsfromowningsecuritiesofanycompanyintheanalyst’sareaofcoverage.

Analystcompensation:Analystsarepaidinpartbasedonthepro?tabilityofGoldmanSachs,whichincludesinvestmentbankingrevenues.Analyst

asof?cerordirector:GoldmanSachspolicygenerallyprohibitsitsanalysts,personsreportingtoanalystsormembersoftheirhouseholdsfromservingasanof?cer,directororadvisorofanycompanyintheanalyst’sareaofcoverage.Non-U.S.Analysts:Non-U.S.analystsmaynotbe

associatedpersonsofGoldmanSachs&Co.LLCandthereforemaynotbesubjecttoFINRARule2241orFINRARule2242restrictionsoncommunicationswithsubjectcompany,publicappearancesandtradingsecuritiesheldbytheanalysts.

AdditionaldisclosuresrequiredunderthelawsandregulationsofjurisdictionsotherthantheUnitedStates

Thefollowingdisclosuresarethoserequiredbythejurisdictionindicated,excepttotheextentalreadymadeabovepursuanttoUnitedStateslawsandregulations.Australia:GoldmanSachsAustraliaPtyLtdanditsaf?liatesarenotauthoriseddeposit-takinginstitutions(asthattermisde?nedinthe

BankingAct1959(Cth))inAustraliaanddonotprovidebankingservices,norcarryonabankingbusiness,inAustralia.Thisresearch,andanyaccesstoit,isintendedonlyfor“wholesaleclients”withinthemeaningoftheAustralianCorporationsAct,unlessotherwiseagreedbyGoldmanSachs.In

producingresearchreports,membersofGlobalInvestmentResearchofGoldmanSachsAustraliamayattendsitevisitsandothermeetingshostedbythecompaniesandotherentitieswhicharethesubjectofitsresearchreports.InsomeinstancesthecostsofsuchsitevisitsormeetingsmaybemetinpartorinwholebytheissuersconcernedifGoldmanSachsAustraliaconsidersitisappropriateandreasonableinthespeci?ccircumstancesrelatingtothesitevisitormeeting.Totheextentthatthecontentsofthisdocumentcontainsany?nancialproductadvice,itisgeneraladviceonlyandhas

beenpreparedbyGoldmanSachswithouttakingintoaccountaclient’sobjectives,?nancialsituationorneeds.Aclientshould,beforeactingonany

suchadvice,considertheappropriatenessoftheadvicehavingregardtotheclient’sownobjectives,?nancialsituationandneeds.Acopyofcertain

GoldmanSachsAustraliaandNewZealanddisclosureofinterestsandacopyofGoldmanSachs’AustralianSell-SideResearchIndependencePolicy

Statementareavailableat:

/disclosures/australia-new-zealand/index.html

.Brazil:DisclosureinformationinrelationtoCVMResolutionn.20isavailableat

/worldwide/brazil/area/gir/index.html

.Whereapplicable,theBrazil-registeredanalystprimarilyresponsibleforthecontentofthisresearchreport,asde?nedinArticle20ofCVMResolutionn.20,isthe?rstauthornamedatthebeginningofthisreport,unlessindicatedotherwiseattheendofthetext.Canada:Thisinformationisbeingprovidedtoyouforinformationpurposesonlyandisnot,andundernocircumstancesshouldbeconstruedas,anadvertisement,offeringorsolicitationbyGoldmanSachs&Co.LLCforpurchasersof

securitiesinCanadatotradeinanyCanadiansecurity.GoldmanSachs&Co.LLCisnotregisteredasadealerinanyjurisdictioninCanadaunder

applicableCanadiansecuritieslawsandgenerallyisnotpermittedtotradeinCanadiansecuritiesandmaybeprohibitedfromsellingcertainsecuri