TheEnormousEconomicPotentialofGenerativeAI

October2025

JosephBriggsGoldman,Sachs&Co.+1212-902-2163joseph.briggs@

SeniorEconomist

Investorsshouldconsiderthisreportasonlyasinglefactorinmakingtheirinvestmentdecision.ForRegACcerti?cationandotherimportantdisclosures,seetheDisclosureAppendix,orgoto

/research/hedge.html

.

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

AnOverviewofGenerativeAI

Step1:TrainingDatatoNeural

Network

Step3:AIOutputtoHuman

Interface

Step2:NeuralNetworktoAI

Step4:Applications

Output

PreviousMLmethods:

PreviousMLmethods:

Datatrainedonspecialized

databasesforspecificpurposes

(e.g.,makestatisticalpredictions

aboutelectionresults,answer

questionsaboutbiomedical

literature,etc.)

GenerativeAI:

Datatrainedonlarge,generalized

databases(i.e.,theentireinternet);

thus1)widerrangeofusecasesand

2)moreeasilyabletospawn

complementaryinnovationswith

specializedusecases("deepening

ofAI").

Modelsgeneratestatistical

PreviousMLmethods:

-Textclassification

-Facialandimagerecognition

-Statisticalpredictionandinference

GenerativeAI:

-Answercomplextextualquestions

withhuman-likelanguageand

structure

-Createoriginalimages,graphics,andvideobasedonuserqueries

-Generateandexplaincodewhichcanbeusedforotherprogramming

anddatascienceapplications

PreviousMLmethods:

Usersmustusespecificcodeor

syntaxtomakenarrowrequests

basedonthemodel'sintended

function.

GenerativeAI:

Useoflargelanguagemodels

(LLMs)allowsforadvancednatural

languageprocessing(NLP)that

incorporatescontextinlargerswaths

oftext,enablingawidervarietyof

requestsandamoreaccessible

interfaceforhuman-AIinteraction.

predictionsbasedonrelationshipsin

trainingdata.

GenerativeAI:

Modelsseektogeneratenew

informationthatisindistinguishable

fromhumandata.Achievedviathe

introductionofasecond

"discriminative"neuralnet,which

evaluatestheoutputoftheprimary

"generative"neuralnetfor

authenticityrelativetohumanoutput.

This"adversarialneuralnetworks"

approachforcesthegenerative

networktoreviseitsoutputandlearn

toconsistently"fool"the

discriminativenetwork.

Threekeychanges:

1.Generalizedratherthanspecialized-widerusecasesandmorecomplementaryinnovations

2.Generativeratherthandescriptive-canproduceoriginalresultsindistinguishablefromhumanoutput

3.Approachableratherthantechnical-caninterfaceviacomplexandcontextualnaturallanguage

Source:GoldmanSachsGlobalInvestmentResearch.

2

3

O*NETdata:OurBaselineAssumes13TaskCategoriesUptoDifficultyLevelof4CouldBeAutomated

Source:GoldmanSachsGlobalInvestmentResearch.

4

Two-ThirdsofCurrentOccupationsCouldbePartiallyAutomatedbyAI

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

Source:GoldmanSachsGlobalInvestmentResearch.

5

One-FourthofCurrentWorkTasksCouldBeAutomatedbyAIintheUS…

ShareofIndustryEmploymentExposedtoAutomationbyAI:US

PercentPercent

50

40

30

20

10

0

46

44

32

37363533

312928282827262625

19

9

1211

6

4

OfficeandAdministrative

Support

Legal

ArchitectureandEngineering

Life,Physical,andSocial

Science

BusinessandFinancial

Operations

CommunityandSocialService

Management

SalesandRelated

ComputerandMathematical

Farming,Fishing,andForestry

ProtectiveService

HealthcarePractitionersand

Technical

EducationalInstructionand

Library

HealthcareSupport

Arts,Design,Entertainment,

Sports,andMedia

AllIndustries

PersonalCareandService

FoodPreparationandServing

Related

TransportationandMaterial

Moving

Production

ConstructionandExtraction

Installation,Maintenance,and

Repair

BuildingandGroundsCleaning

andMaintenance

1

50

40

30

20

10

0

Source:GoldmanSachsGlobalInvestmentResearch.

6

…AndinEurope

ShareofIndustryEmploymentExposedtoAutomationbyAI:EuroArea

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

PercentPercent

50

40

30

20

10

0

45

34

3129

242221

15

87

ClericalSupportWorkers

Professionals

TechniciansandAssociate

Professionals

Managers

Total

ArmedForcesOccupations

SkilledAgricultural,Forestryand

FisheryWorkers

ServiceandSalesWorkers

ElementaryOccupations

PlantandMachineOperators,

andAssemblers

CraftandRelatedTrades

Workers

4

50

40

30

20

10

0

Source:GoldmanSachsGlobalInvestmentResearch.

7

Globally,18%ofWorkCouldbeAutomatedbyAI,withLargerEffectsinDMsthanEMs

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

Source:GoldmanSachsGlobalInvestmentResearch.

8

LargerEstimatesAssumetheAutomationofPhysicalTasksLessThatSeemLikelyintheNear-Term

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

Source:GoldmanSachsGlobalInvestmentResearch.

9

ReplacementinLegalandAdministrativeFields,Productivity-EnhancementElsewhere

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

Source:GoldmanSachsGlobalInvestmentResearch.

10

StudiesofEarlyAdoptersFindThatAIAdoption

IncreasesWorkerProductivityby27-31%onAverage

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

Source:GoldmanSachsGlobalInvestmentResearch.

11

SomeRiskofFrictionalUnemployment…

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

Source:GoldmanSachsGlobalInvestmentResearch.

12

…butTechnologicalInnovationDrivesEmploymentGrowthOvertheLongRun

MillionsEmployment,byNewandPre-ExistingOccupationsMillions

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

200

175

150

125

100

75

50

25

0

OccupationsThatExistedin1940

Professionals

Managers

Clerical&Admin

Production

Construction

PersonalServices

Transportation

Technicians

Sales

CleaningServices

Health

Farming

Total

OccupationsThatDidNotExistin1940

200

175

150

125

100

75

50

25

0

Source:GoldmanSachsGlobalInvestmentResearch.

13

Historically,WorkerDisplacementfromAutomationHasBeenRoughlyOffsetbyCreationofNewRoles

PercentagepointsContributionstoTotalWageBillGrowth*Percentagepoints

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

1.5

1.0

0.5

0.0

-0.5

-1.0

-1.5

DisplacementofWorkersFromOldRolesRemploymentofWorkersinNewRoles

*Linesindicate10-yearmovingaveragesof

Greenshading=

ITautomationboom

reinstatement(green)anddisplacement(red)effects

1950196019701980199020002010

1.5

1.0

0.5

0.0

-0.5

-1.0

-1.5

Source:GoldmanSachsGlobalInvestmentResearch.

14

GenerativeAICouldBoostAggregateLaborProductivityGrowthby1.5ppintheUS

PercentagepointsEffectofAIAdoptiononAnnualLaborPercentagepoints

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

66

55

44

33

22

11

00

-1-1

ProductivityGrowth,10-YearAdoptionPeriod

ReemploymentofDisplacedWorkersLaborDisplacement

IncreasedProductivityofNon-DisplacedWorkers

2.4

1.30.8

0.3

2.92.4

0.7

0.5

1.5

MuchLessPowerfulAI

SlowerAdoption(30Years)

SlowerAdoption(20Years)

SlighltyLessPowerfulAI

NoLaborDisplacement

Baseline

SlightlyMorePowerfulAI

MoreLaborDisplacement

MuchMorePowerfulAI

Source:GoldmanSachsGlobalInvestmentResearch.

15

MorePessimisticEstimatesNotthatDifferent;DrivenbyOutlookforTimelineandNewTaskCreation

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

Source:GoldmanSachsGlobalInvestmentResearch.

16

ProductivityGrowthBoostsCouldBeSizableinOtherCountriesAsWell

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

Source:GoldmanSachsGlobalInvestmentResearch.

17

ASizableInvestmentCycle

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

Source:GoldmanSachsGlobalInvestmentResearch.

18

MarketsHaveAlreadyUpgradedAIHardware

RevenuesbyOver$300bnAnnuallySince2023

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

Source:FactSet,GoldmanSachsGlobalInvestmentResearch.

19

ModelsandComputationalDemandsAreGrowingMuchFasterThanCostorEnergyEfficiencies

Percent

Percent

AIGrowthTrends,Annual

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

500

450

400

350

300

250

200

150

100

50

0

IncreaseCostsDecreaseCosts

500

450

400

350

300

250

200

150

100

50

0

LLMQueriesNumberof

LargeScaleModels

TrainingFLOPsperModel

FLOPsperWattFLOPsperDollar

Source:EpochAI,GoldmanSachsGlobalInvestmentResearch.

20

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

LargerModelsAreStillLeadingtoImprovedPerformance

Log(Perplexity)

ModelSizevs.Performance,

Log(Perplexity)

Inverted

2012-2023

Inverted

1

1.5

2

1

1.5

2

2.5

2.5

3

3

3.5

3.5

4

4

4.5

5

4.5

5-●-5.5

5.5iiiiiiiiiiiiiir6

131415161718192021222324252627log(NumberofParameters)

Source:EpochAI,GoldmanSachsGlobalInvestmentResearch.

MMLUScoreModelSizevs.Performance,MMLUScore

1

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0

CurrentMMLU-ProLeaderboard

1

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0

202122232425262728

log(NumberofParameters)

21

PriorGPTInvestmentCyclesHaveExceededtheAIBuildoutasaShareofGDP

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

PercentofGDP

PercentofGDP

PeakHistoricalInvestmentImpulse,

EmergingTechnologiesinFrontierEconomies

5

4

3

2

1

0

5

4

3

2

1

0

UKRailroads

(1860s)

USRailroads

(1880s)

USAuto

Infrastructure(1910s)

USElectricMotor

(1920s)

USICTHardware

(1990s)

USTelecom

(1990s)

USGenerative

AI(2020s)

PercentagePoints

PercentagePoints

InvestmentinGeneralPurposeTechnologies,ShareofGDPvs.Pre-ProductivityBoomLevel

2.5

2.0

1.5

1.0

0.5

0.0

-0.5

-1.0

Startoflabor

productivityboom

ElectricMotor-ManufacturingPlants&EquipmentInformationTech-ICTEquipmentandSoftware

GenAI-Semis,Compute,DataCenters,Power

-5-4-3-2-1012345678910

2.5

2.0

1.5

1.0

0.5

0.0

-0.5

-1.0

YearsSinceProductivityBoom(t=0)

Source:BureauofEconomicAnalysis,GoldmanSachsGlobalInvestmentResearch.

22

TheExpectedPresentDiscountedValueofCapitalRevenuefromAIExceedsCapexProjections

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

2025

2026

2027

2028

2029

2030

2031

2032

2033

2034

2035

2036

2037

2038

2039

2040

2041

2042

2043

2044

2045

2046

2047

2048

2049

2050

15

10

25

20

$bn

600

500

400

300

200

100

0

PresentValueofFutureAICapitalRevenue,

GSBaselineAssumptions

600

500

400

300

200

100

0

$bn

$tn

PresentDiscountedValueof

AICapitalRevenues,AlterativeScenarios

-25

$tn

-20

19

-15

12

12

11

-10

8

7

6

5

5

-5

5

r0

High(20%)

High(24%)

High(60%)

FasterSlower

Low(10%)

Low

(8%)

Low(28%)

0

GSDiscountProductivity

Capital

Adoption

Base-Rate

Share

Rate

Source:GoldmanSachsGlobalInvestmentResearch.

23

FortheexclusiveuseofNEIL.WANG@TROWEPRICE.COM

FirstMoversHaveShownMixedPerformanceinPriorInfrastructureBuilds

HistoricalEpisode

FirstMoverReturns

FastFollowerReturns

Explanation

UKTurnpikes

Neitherabnormallyhighorlow

Neitherabnormallyhighorlow

Regulationlimitedreturns.

UKCanals

High

Low

Investmentvaluableforfirstmovers;Introductionofrailroadslimitedreturnstolaterentrants.

UKRailways

Low

High

Post-bustconsolidatorswereabletopurchaseassetsatlowvaluations,long-runvaluecreationdrovehealthyreturnsforcompaniesthatboughtrailroadsatdeepdiscounts.

USRailroads

Mixed

High

Insomecasesearlymoverswereabletocaptureoutsizedreturnsvialocalmonopoly(e.g.,PennsylvaniaRailroadbecametheworld's

largestcorporations),butothershadnegativereturns(halfofallrailroadmileagebuiltbefore1870wentbankruptorreorganized).

Laterconsolidators(e.g.NYCentral,UnionPacific)wereabletogenerateoutsizedreturnviapurchasesatlowvaluations.

USElectricity

Neitherabnormallyhighorlow

Neitherabnormallyhighorlow

Firstmoverskeptmarketshare,butregulationandorganizationasutilitieslimitedreturns.

USIT

High

High

Firstmoverswereabletograbmarketshareinmainframes(IBM),

operatingsystems(Microsoft),networking(Cisco),andcloud

computing(AWS),althoughthelate1990ssawoverexuberanceandoverinvestment.Laterentrantsalsofaredwellinmanyareasgiventransformativeeffectoftechoneconomy.

GlobalFiber-OpticCables

Low

High

Overbuildledtolowreturnsforfirstmovers,laterbuyerscapturedfinancialupsidebypurchasingassetsatlowvaluations.

USTelecom

Low

High

Overbuildledtolowreturnsforfirstmovers,laterbuyerscapturedfinancialupsidebypurchasingassetsatlowvaluations.

Note:Redshadingindicatesunderperformance,greenoutperformance,yellowneutralperformance.

Source:GoldmanSachsGlobalInvestmentResearch.

24

FirstMoversHaveShownMixedPerformanceinPriorInfrastructureBuilds

FirstMoverDominantFastFollowerDominantImplicationsForAILeadersToday

Highswitchingcosts

Lowswitchingcosts

Uncertain

AILeadersOutperform

AILeadersUnderperform

Networkeffectsestablishedbyfirst-moverproductdeterminestandards

Productmarketstandardsundetermined

Networkeffectsimproveproductquality

Productqualitydeterminedindependentlyofnetworkeffects

Scarceassetsnecessaryforproductdevelopment

Complementaryassetsabundant

Thetechonologystackisverticallyintegrated

Thetechnologicalstackisseparatelyownedandeasilyredeployed

Highcostofimitation

Lowcostofimitation

Slowtechnologicalchange

Fasttechnologicalchange

Slowmarketgrowth

Fastmarketgrowth

Patent/IPprotectionsstrong

Patent/IPprotectionsweak

Customeruncertaintyonuselow

Customeruncertaintyonusehigh

Note:Redshadingindicatesreasonforpotentialunderperformance,greenoutperformance,yellowneutralperformance.

Source:GoldmanSachsGlobalInvestmentResearch.

25

SomeSignsofVerticalIntegration

Source:AndreessenHorowitz,GoldmanSachsGlobalInvestmentResearch.

26

SoftwareProvidersBenefitedFromHighSwitchingCosts,CompaniesTodayAdoptingMultipleModels

Percentagepoints

Percentagepoints

AnnualContributionstoPrivateSector

OutputGrowthFrom:

0.6

0.5

0.4

0.3

0.2

0.1

0.0

-0.1

-0.2

Software(Total)SoftwareInvestment

AverageInvestment

ShareofTotal:27%

8182838485868788899091929394959697989900010203040506070809101112

0.6

0.5

0.4

0.3

0.2

0.1

0.0

-0.1

-0.2

Note:SoftwarecontributionsestimatedbasedonresultsfromByrneetal.(2013)andOlinerandSichel(2000).

PercentNumberofModelsDeployed,Percent

100

80

60

40

20

0

byCompanyRevenue

5+

4

3

2

1

0

100

80

60

40

20

0

$500mn-$50bn$5bn-20bn$>20bn

Source:AndreessenHorowitz,GoldmanSachsGlobalInvestmentResearch.

27

PreviousMilestoneTechnologiesHaveLedtoLaborProductivityBooms,buttheTimingIsHardtoPredict

PercentTechnologyAdoptionRatesforHouseholdsandFirmsPercent19811986199119962001200620112016

100

80

60

40

20

0

Personalcomputerinvented

Productivity

boombegins

50%AdoptionThreshold

Electricity:Manufacturing(Bottom)

Electricity:Household(Bottom)

PC:Workplace(Top)

PC:Household(Top)

Developmentofelectricmotor

100

80

60

40

20

0

1899190819171926193519441953

USLaborProductivity

Percentchange,10-yearannualrate

Percentchange,

10-yearannualrate

4.5

4.0

3.5

3.0

2.5

2.0

1.5

1.0

0.5

0.0

Personal

computer

invented:1981

Developmentofelectric

motor:

~1890

Green=Resultingproductivityboom

4.5

4

3.5

3

2.5

2

1.5

1

0.5

0

18851905192519451965198520052025

Source:GoldmanSachsGlobalInvestmentResearch.

28

Economy-wideAIAdoptionCurrentlyStandsat9.7%

Source:CensusBureau,GoldmanSachsGlobalInvestmentResearch.

29

AdoptionRatesVaryWidelyAcrossSectors

Source:CensusBureau,GoldmanSachsGlobalInvestmentResearch.

30

AdoptionRatesAreHigherAmongComputing,WebSearch,andMotionPictureCompanies

Source:CensusBureau,GoldmanSachsGlobalInvestmentResearch.

31

LargeCompaniesHaveLedtheAdoptionSurge

Source:CensusBureau,GoldmanSachsGlobalInvestmentResearch.

32

AdoptionisHigherinIndustriesSettoBenefittheMost

Source:GoldmanSachsGlobalInvestmentResearch.

33

JobPostingsHavePulledBackMoreinAI-ExposedIndustries…

Source:IndeedHiringLab,GoldmanSachsGlobalInvestmentResearch.

34

…ButtheImpactonLaborMarketOutcomesAppearsSmallandInsignificantSoFar

Source:GoldmanSachsGlobalInvestmentResearch.

35

EmploymentIsContractinginSomeSectorsWhereAnecdotesSuggestAIIsSubstitutingforLabor

Source:BureauofLaborStatistics,HaverAnalytics,GoldmanSachsGlobalInvestmentResearch.

36

TechísEmploymentShareHasDeclinedBelowTrend

Source:BureauofLaborStatistics,HaverAnalytics,GoldmanSachsGlobalInvestmentResearch.

37

YouthUnemploymentinTechExposedSectorsHasRisenRecently

Source:BureauofLaborStatistics,iPUMS,GoldmanSachsGlobalInvestmentResearch.

38

AlsoNoSignofAI-RelatedLayoffsorUnemploymentYet

Source:CensusBureau,GoldmanSachsGlobalInvestmentResearch.

39

LaborReplacementTendstoAccelerateDuringEconomicDownturns

Source:JaimovichSiu2018,GoldmanSachsGlobalInvestmentResearch.

40

FiveDimensionsofOurLaborDisplacementRiskScores

1.ConsequenceofError

2.TaskRepetitiveness

3.TaskCohesion

4.ValueofAI-ExposedTasksvs.OverallWage

5.Backvs.FrontOffice

Source:GoldmanSachsGlobalInvestmentResearch.

41

OurLaborDisplacementRiskMetricsVaryAcrossOccupationsinanIntuitiveManner…

Source:O*NET,GoldmanSachsGlobalInvestmentResearch.

42

…AndPredictJobsThatAreMoreorLessatRiskofAutomationinaMannerthatAlignsWithAnecdotes

Source:O*NET,GoldmanSachsGlobalInvestmentResearch.

43

OurScoresSuggesta6.4%DisplacementRateifthe20%MostAt-RiskJobsAreFullyAutomated

Source:GoldmanSachsGlobalInvestmentResearch.

44

FasterAdoptioninDMs,SlowerAdoptioninEMs

GSAIAdoptionScenarios

PercentCumulativeAdoptionPercent

100

90

80

70

60

50

40

30

20

10

0

US

OtherDMs

AdvancedEMsOtherEMs

100

90

80

70

60

50

40

30

20

10

0

2020202520302035204020452050

PercentMarginalAdoptionPercent

9

8

7

6

5

4

3

2

1

0

US

OtherDMs

AdvancedEMsOtherEMs

9

8

7

6

5

4

3

2

1

0

2020202520302035204020452050

Note:AdoptionratesshownasashareofeconomywidefirmsexposedtoAIautomation.

Source:GoldmanSachsGlobalInvestmentResearch.

45

OurBaselineImpliesa0.4ppPeakIncreaseintheUnemploymentRate

PercentagepointsEffectofAI-DrivenLaborDisplacementonthePercentagepoints

2.0

1.8

1.6

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0.0

UnemploymentRatebyScenario

Baseline

——FasterAdoption

LargerDisplacement

FrontloadedDisplacementFasterxLarger

FrontloadedxLarger

2020202520302035204020452050

2.0

1.8

1.6

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0.0

Source:GoldmanSachsGlobalInvestmentResearch.

46

MarketImplicationsofAI:EquitiesBiggestBeneficiary;EffectonRatesDependsonWhetherAnticipated

Source:GoldmanSachsGlobalInvestmentResearch.

47

DisclosureAppendix

October16,2025

48

DisclosureAppendix

RegAC

I,JosephBriggs,herebycertifythatalloftheviewsexpressedinthisreportaccuratelyre?ectmypersonalviews,whichhavenotbeenin?uencedbyconsiderationsofthe?rmísbusinessorclientrelationships.

Unlessotherwisestated,theindividualslistedonthecoverpageofthisreportareanalystsinGoldmanSachsíGlobalInvestmentResearchdivision.

Disclosures

Regulatorydisclosures

DisclosuresrequiredbyUnitedStateslawsandregulations

Seecompany-speci?cregulatorydisclosuresaboveforanyofthefollowingdisclosuresrequiredastocompaniesreferredtointhisreport:managerorco-managerinapendingtransaction;1%orotherownership;

compensationforcertainservices;typesofclientrelationships;managed/co-managedpublicofferingsinpriorperiods;directorships;forequitysecurities,marketmakingand/orspecialistrole.GoldmanSachstradesormaytradeasaprincipalindebtsecurities(orinrelatedderivatives)ofissuersdiscussedinthisreport.

Thefollowingareadditionalrequireddisclosures:Ownershipandmaterialcon?ictsofinterest:GoldmanSachspolicyprohibitsitsanalysts,professionalsreportingtoanalystsandmembersoftheir

householdsfromowningsecuritiesofanycompanyintheanalystísareaofcoverage.Analystcompensation:Analystsarepaidinpartbasedonthepro?tabilityofGoldmanSachs,whichincludesinvestment

bankingrevenues.Analystasof?cerordirector:GoldmanSachspolicygenerallyprohibitsitsanalysts,personsreportingtoanalystsormembersoftheirhouseholdsfromservingasanof?cer,directoror

advisorofanycompanyintheanalystísareaofcoverage.Non-U.S.Analysts:Non-U.S.analystsmaynotbeassociatedpersonsofGoldmanSachs&Co.LLCandthereforemaynotbesubjecttoFINRARule2241orFINRARule2242restrictionsoncommunicationswithasubjectcompany,publicappearancesandtradinginsecuritiescoveredbytheanalysts.

AdditionaldisclosuresrequiredunderthelawsandregulationsofjurisdictionsotherthantheUnitedStates

Thefollowingdisclosuresarethoserequiredbythejurisdictionindicated,excepttotheextentalreadymadeabovepursuanttoUnitedStateslawsandregulations.Australia:GoldmanSachsAustraliaPtyLtdanditsaf?liatesarenotauthoriseddeposit-takinginstitutions(asthattermisde?nedintheBankingAct1959(Cth))inAustraliaanddonotprovidebankingservices,norcarryonabankingbusiness,inAustralia.Thisresearch,andanyaccesstoit,isintendedonlyforìwholesaleclients?withinthemeaningoftheAustralianCorporationsAct,unlessotherwiseagreedbyGoldmanSachs.Inproducingresearchreports,membersofGlobalInvestmentResearchofGoldmanSachsAustraliamayattendsitevisitsandothermeetingshostedbythecompaniesandotherentitieswhicharethesubjectofitsresearchreports.Insomeinstancesthe

costsofsuchsitevisitsormeetingsmaybemetinpartorinwholebytheissuersconcernedifGoldmanSachsAustraliaconsidersitisappropriateandreasonableinthespeci?ccircumstancesrelatingtothesitevisitormeeting.Totheextentthatthecontentsofthisdocumentcontainsany?nancialproductadvice,itisgeneraladviceonlyandhasbeenpreparedbyGoldmanSachswithouttakingintoaccountaclientísobjectives,?nancialsituationorneeds.Aclientshould,beforeactingonanysuchadvice,considertheappropriatenessoftheadvicehavingregardtotheclientísownobjectives,?nancialsituationandneeds.AcopyofcertainGoldmanSachsAustraliaandNewZealanddisclosureofinterestsandacopyofGoldmanSachsíAustralianSell-SideResearchIndependencePolicyStatementareavailableat:

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