生成式人工智能市場現(xiàn)狀(英文版)_第1頁
生成式人工智能市場現(xiàn)狀(英文版)_第2頁
生成式人工智能市場現(xiàn)狀(英文版)_第3頁
生成式人工智能市場現(xiàn)狀(英文版)_第4頁
生成式人工智能市場現(xiàn)狀(英文版)_第5頁
已閱讀5頁,還剩69頁未讀, 繼續(xù)免費閱讀

下載本文檔

版權說明:本文檔由用戶提供并上傳,收益歸屬內容提供方,若內容存在侵權,請進行舉報或認領

文檔簡介

StateoftheGenerativeAIMarket

September2024

1SG

Introduction

StateoftheMarket

GenAIImplementationRoadmap

DifferentiationinaWorldofGenAI

2024StateoftheGenerativeAIMarket|2

TableofContents

17BuyingfortheEnterprise

17

BuyingAIImplementationServices

19

KeyConsiderationswhenBuyingAIToolsorPlatforms

21DataMatters

21

DataPlatformstoSupportGenAI

23

DataGovernanceandDataEngineering

25NavigatingToughTerrain

25RisksofGenAIfortheEnterprise

26AICompliance:AGlobalPerspective

27NewFrontiersofAIGovernance

29EthicalApproachestoGenAI

30DifferentiationinaWorldofGenAI

3Introduction

4StateoftheMarket

5MarketAdoptionbytheNumbers

6KeyUseCases:TodayandTomorrow

8What’sNext

9ThePilot-to-ProductionDilemma

10TransformingSoftwareDevelopment

11

GenAIImplementationRoadmap

12ImpactsonBothBusinessandIT

13

GenAIfortheBusiness

32KeyContributors

14

GenAIforIT–AILandingZones

16

GenAIforIT–AIOps

?2024InformationServicesGroup,Inc.AllRightsReserved.

1SG

Introduction

StateoftheMarket

GenAIImplementationRoadmap

DifferentiationinaWorldofGenAI

Introduction

IthasbeenalmosttwoyearssincegenerativeAI(GenAI)

emergedwithforceontothetechnologylandscape,andtwothingsarenowclear.First,GenAIwillsignificantlyimpact

andtransformeveryindustry—itisjustamatteroftime.

Second,wearestillintheearlystagesofadoptionandtherearesignificantchallengestoharnessingthepowerofGenAIduringthistransition.Whilethereisclearlyanabundanceofhype,itisalsoclearthatthistrulywilldrivedramaticchange,presentingopportunitiesandrisksforeveryenterprise.In

manyways,itissimilartothelate1990swhentheinternetwasyoungandtherewasmuchweedingoutandgrowingneeded.This2024StateoftheGenAIMarketreportwill

guideyouandyourenterpriseonthekeythemesand

developmentsthatareshapingthelandscape,highlightingboththesuccessesandtheobstaclesforsuccessfully

navigatingthiscrucialnewtechnology.

OneofthemostnotabletrendsintheGenAImarketis

theincreasingsophisticationandaccessibilityofGenAI

models.Whileadvancementsindeeplearningalgorithmsandcomputationalresourceshaveenabledthesemodelstobetterunderstandandgeneratecomplexcontent

acrossvariousdomains,therealityisthatleveragingthesecapabilitieseffectivelyisfarfromstraightforward.

Fromnaturallanguageprocessingandimagesynthesistoautomatedcodegeneration,enterprisesfacechallengesinintegratingGenAIsolutionsintotheirexistingworkflows,ensuringdataqualityandmanagingthecostsassociatedwiththesetechnologies.

Tohelpaddressthecostissue,morethanhalfofthe

softwareprovidersevaluatedinourGenAIBuyersGuide

haveimplementedsomeformofcostcontrolandmosthaveinformedusoffurtherplansinthisarea.

TherapidadoptionofGenAIalsobringsothernew

challengestotheforefront.Itisprovingverydifficultfor

enterprisestohireordeveloptheAIskillstheyneedamongtheirteams.Concernssuchasdataprivacy,algorithmic

fairnessandthepotentialformisusehavebecomemore

prominent,promptingtheneedforrobustgovernance

frameworksandexplicitethicalguidelines.Manyenterprisesfindthemselvesill-equippedtonavigatethesecomplexities,riskingunintendedconsequencesandnegativeimpacts.Byacknowledgingandaddressingthesechallenges,enterprisescanmakemoreinformeddecisionsanddrivemeaningful

progressinanevolvinglandscape.

?2024InformationServicesGroup,Inc.AllRightsReserved.

2024StateoftheGenerativeAIMarket|3

ItisimperativeforbusinessestoapproachGenAIprojectswitharealisticunderstandingofthepotentialpitfalls,investinginthe

necessaryexpertiseandinfrastructuretorealizeitsbenefitsresponsibly.

Bringyourquestionstothetableasyousitdownwiththisreport.What’sactuallyhappeningouttherewithGenAI—

beyondthehype?Whattypesofusecasesarehavingan

impact?Shouldwebuildorbuy?Whatkeyriskisinourblindspot?Andifeveryonekeepspacewithadoptionanyway,isitevenpossibletoachievedifferentiation?

ThisreportisdesignedtobeakeyresourceasyounavigateyourAIplanningandevaluationoverthenextsixmonths.

Readitstraightthroughorjumptothesectionyoucare

mostabout.Haveanunansweredquestion?Wanttovalidate

yournextsteps?

Dropusalineandlet’stalk.

Introduction

StateoftheMarket

GenAIImplementationRoadmap

DifferentiationinaWorldofGenAI

2024StateoftheGenerativeAIMarket|4

StateoftheMarket

?2024InformationServicesGroup,Inc.AllRightsReserved.

1SG

Introduction

StateoftheMarket

GenAIImplementationRoadmap

DifferentiationinaWorldofGenAI

2024StateoftheGenerativeAIMarket|5

MarketAdoptionbytheNumbers

SpendingonGenAIinitiatives

willincreaseby

50%

in2025comparedwith2024.

4.3%

of2024ITspendingisonGenAI,andenterprisesexpectthattogrowto

6.5%

bytheendof2025.

DataSource:ISG2024MarketLens??GenAIUseCaseStudy

EnterprisesexpecttocaptureasignificantshareoftheanticipatedROIfromtheircurrentGenAIinitiativesin2025.Howdoyoucompare?

ExpectedROICapture

2024Firsthalfof2025Secondhalfof202520262027orlater

Efficiency·

Innovation·

CustomerService·

CostSavings·

BusinessGrowth·

ExpectROI

tofallshortof

expectations

23%

27%

22%

22%

27%

26%

22%

21%

29%

25%

31%

10%

10%

18%

17%

17%

17%

14%

27%

31%

17%

16%

3%

3%

7%

7%

1%

1%

4%

5%

?2024InformationServicesGroup,Inc.AllRightsReserved.

1SG

Introduction

StateoftheMarket

KeyUseCases:Todayand

Tomorrow

OurISGGenAIUseCaseStudy,conductedin

August2024,focusedonwhichusecasesglobal2000companieshaveinvestedintodate,andwhichhavereceivedthemostfunding.While

GenAIusecasesspanabroadrangeofbusinessdomains,threeofthetopfivemostwell-fundedareasarefocusedoncontactcenter,efficiencygainsandcontentgeneration.

Theusecasesgettingthemostfundingtodayareaimedatefficiencyandprofitability,notrevenue.AndAItoolscontinuetobeprimarilydeployedinprocessesthathavea“humanintheloop”(HITL)tointermediateandvalidatetheoutputbeforeitisdeliveredtocustomers.

DataSource:ISG2024MarketLens??GenAIUseCaseStudy

?2024InformationServicesGroup,Inc.AllRightsReserved.

GenAIImplementationRoadmap

DifferentiationinaWorldofGenAI

2024StateoftheGenerativeAIMarket|6

MostcommonusecaseHighestinvestmentusecase

GenAIUseCasesin2024

53%

Customerservicechatbots

43%

AutomatedITtesting

41%

41%

Audio/visualcontentgeneration

Customerservicesupport

35%

HRsupport

33%

Customercommunications

32%

Businessprocesswork?owmanagement

31%

Documentationcreation

30%

30%

ITsecurity

Marketingtext/graphicgeneration

29%

29%

29%

Billingandinvoicing

Marketingresearch/customerinsights

Marketingcontentmanagement

28%

Frauddetection

27%

Softwarecodegeneration/translation

26%

Contactcentermanagement/monitoring

1SG

Introduction

StateoftheMarket

GenAIImplementationRoadmap

Thepracticaloutcomesofcurrentinvestmentsareimprovedquality(e.g.,betterresponsestocustomerinquiries,betterprocessmanagement)and,tosome

extent,increasedefficiency(reducedresponsetimes,fasterresolutions).Butaslongastheusecasesservetoaugmenthumans,thehuman’sabilitytooperatetheprocesswillultimatelybethelimitingfactorforhowdramaticallytheAI

toolsandprocessescanscale.

ThisfocusonqualityimprovementsandHITLprocessesrevealsthat

enterprisesaretakingafairlysafe,liability-limitingapproachtoGenAI,for

now.Intheseusecases,AIresultsarereviewedbeforetheyaretrustedwith

thegoalofimprovingthequalityofoutputovertime.GiventheconcernsaboutAIhallucinationsandunclearIPrestrictionsrelatedtomanylargeAItraining

sets,thisconservativeapproachisreasonablefornear-termusecases.

12%

Innovation

32%

Efficiency

GenAI

InitiativePrimary

Motivation

17%

CustomerService

21%

BusinessGrowth

18%

CostSavings

DataSource:ISG2024MarketLens??AIStudy

?2024InformationServicesGroup,Inc.AllRightsReserved.

DifferentiationinaWorldofGenAI

2024StateoftheGenerativeAIMarket|7

AsGenAIcapabilitiesandusecasesprogress,twoneedscometothefrontforenterprisesandtheirGenAIinitiatives:

1.TheabilitytooperateGenAIwithouttheconstraintsofHITL

processes,soenterprisescanbeginachievingreturnsthatscaleinanon-linearway.

2.Aprogressionfrominvestmentsthatfocusonqualityand

efficiencyimprovementstowardinvestmentsthatfocusongrowthandnewproducts.

Atthispoint,weexpectenterpriseswillcontinuetodiversifyGenAIusageacrossHITLusecasesandembraceAI-drivenservicesthatincorporateefficiencyimprovements.

Overthelongerterm,enterpriseswillneedtoobservefurthertechnicalprogressandgainforthemselvesabetterunderstandingof

associatedrisksbeforetheybegintryingusecaseswithgreaterscaleandROI.

1SG

Introduction

StateoftheMarket

What’sNext?

Giventhatpastresearchhashighlighted

theimportanceofrevenuegrowthasatop

enterpriseobjectivefortheadoptionofAI,

higher-valueusecasesinthefuturewillbe

thosethatdonotinvolveHITLprocesssothatenterprisescanachievemoredramaticscaling.WepredicttwowavesofAIadoption:

1

Wave1:Themajorusecasesfrom2024willstilldrivesignificantvaluein2025.

TheseareHTILprocessesrelatedtothecontactcenterandcustomerexperience,whereGenAIprovidesinputinreal-timetodrivefasterresolutionsandmore

accuratesolutions.GenAIisstilllimitedbythecapacityofthecontactcenterorcustomerserviceagent.

2

Wave2:Emergingusecasesin2025willprimarilyfocusonaugmentingexpertise.Supportingcompliance,forecasting,

marketresearch,supplychainplanningandsoftwaredevelopmentareall

domainsinwhichhumanexpertise—ratherthanhumantime—canbethelimitingfactors.

?2024InformationServicesGroup,Inc.AllRightsReserved.

GenAIImplementationRoadmap

DifferentiationinaWorldofGenAI

2024StateoftheGenerativeAIMarket|8

HighestAnticipatedGenAIUseCasesfor2025

Emergingusecasefor2025

Topusecasein2024

28%

Customerservicechatbots

21%

Businessprocesswork?owmanagement

19%

Customerservicesupport

18%

Marketingresearch/customerinsights

18%

Customercommunications

18%

Softwarecodegeneration/translation

17%

Planning,budgetingandforecasting

16%

Supplychainoptimization

16%

Regulatorydocumentation/compliance

15%

Contactcentermanagement/monitoring

DataSource:ISG2024MarketLens??GenAIUseCaseStudy

1SG

Introduction

StateoftheMarket

GenAIImplementationRoadmap

DifferentiationinaWorldofGenAI

2024StateoftheGenerativeAIMarket|9

ThePilot-to-ProductionDilemma

ScalingGenAIfrompilotprojectstofull-scaleproduction

deploymentsremainsfraughtforenterprises.Presently,

neitherenterprisesnorsoftwareorserviceprovidershaveestablishedarepeatableandreliablepathfrominitialpilotusecasestolarge-scale,production-readyworkloads.Theresultinglackofmarketevidenceofsuccessfuloutcomes

meansthatenterprisesremainhesitanttoinvestatthe

levelslikelyneededtogeneratethoseoutcomesinthefirstplace.Consequently,wedonotyetseetheamountofvaluecreationneededtodriveamarketchainreaction.

Enterprisesexpressastrongdesiretouseandbenefit

fromAI,buttheywantsignificant,large-scalewinsratherthannumerousminorvictories.Despitepromisingresultsfrompilotprojects,thesizeoftheseeffortsoftenremainsmarginalrelativetoenterprisegoals.Insteadofsuccessfulnoveltyprojects,enterprisesarekeentoidentifyaclearpaththatleadstosubstantialmaterialgains.

TherealityisthattraditionalITorganizationsarenotalwaysalignedwiththenext-generationopportunitiesthatAIcanpresent.TheaverageITorganizationtendstowardbeing

atechnology“fastfollower,”andtypicallyrealizesnew

?2024InformationServicesGroup,Inc.AllRightsReserved.

capabilitiesroughlyonpacewiththemarketasawhole.Doesthatdescribehowyourenterpriseoperates?

Thissuggeststhatwhilewaitingandwatchingfor

evidenceofdramaticoutcomes,enterpriseswilldowelltostrategicallypartnerwithsoftwareandserviceproviders

thatcandemonstratesolidwinsagainstspecific,well-

definedbusinessgoals.Partneringcanreducetheriskofwastinginternaleffortonunsuccessfulprojects,provideaccesstoexpertisethatimprovesthelikelihoodofpositiveoutcomesandallowenterprisestoharnesscutting-edgetools,frameworksandtechnologiesthatmaynotbe

availableinhouse.

CurrentStatusofGenAIInitiatives

8%

Evaluationoftools/providers

7%

Notlive,

testingphase

43%

Livepilot/trialwithoneormoredepartments

27%

Movingtowardsfullproduction

15%

Fullyinproduction

DataSource:ISG2024MarketLens??GenAIUseCaseStudy

1SG

Introduction

StateoftheMarket

GenAIImplementationRoadmap

DifferentiationinaWorldofGenAI

2024StateoftheGenerativeAIMarket|10

TransformingSoftwareDevelopment

Thefastestrisingusecaseoutsideofcustomerserviceverywellmaybethekillerappenterpriseshavebeenlookingfor:GenAIinsoftwaredevelopment.

ITdepartmentsarebeginningtoharnessthepowerofGenAItostreamlineoperations,enhanceinnovationandoptimizeworkflowswithinITinfrastructure.Asaresult,software

developmentisemergingasaleadingapplicationforGenAI,drivingsubstantialbenefitsandrevolutionizingthewayIT

departmentsoperate.70%ofourrespondentsreportusingChatGPTforsoftwaredevelopmentactivities,with33%usingGitHubCoPilot.

IntegratingGenAIintothesoftwaredevelopmentprocess

cansignificantlyenhanceproductivity,efficiencyandquality.

AI-drivenanalyticsstreamlinestakeholderinterviewsand

requirementsgathering,whileautomatedtoolsimprovesystemarchitecturesandthedesignofuserinterfaces.AIassistants

supportcodegenerationandbugfixing,reducingmanualeffortsandimprovingoverallcodequality.AIcangenerateandexecutesoftwaretestcasesandimproveregressiontesting.Additionally,AItoolsoptimizethedeploymentprocessesandproactively

managesystemmaintenance,reducingerrorsanddowntime.

Overall,GenAIdeliverssubstantialproductivitygains,estimatedat

30–42%,byautomatingpredictiveinsights,facilitatingrobustand

andsecurity.

7%

4%4%

2%

PlanningandDesignCodingTestingDeploymentMaintenance

Requirements

error-resistantcodingpractices,andenhancingsoftwarequality

%ImprovementwithIntroductionofGenAI

3%

3%

.

18%

.

31%

.

7%

25%

27%

11%

3

4

Comprehensivetesting

Continuous

improvement

Automatedand

AI-driventesting

coversawiderrangeofscenariosandusecases,ensuringmorethoroughvalidationofsoftware

beforedeployment.

AIsystemslearn

frompastdata

andcontinuously

optimizeprocesses,leadingtoongoing

improvementsin

qualityandefficiency.

KeybenefitsyoushouldberealizingwithGenAIinthesoftwaredevelopmentprocess

1

Earlydetectionofissues

AI-drivencodereviewsandtestingidentify

potentialissuesearlyinthedevelopmentcycle,preventingdefects

fromprogressingtolaterstages.

2

Consistentapplicationofbestpractices

AItoolsenforcecodingstandardsandbest

practices,ensuring

consistencyand

reducingthelikelihoodoferrors.

DataSource:ISG2024MarketLens??GenAIUseCaseStudy;ISG2024GenAISoftwareDevelopmentStudy

?2024InformationServicesGroup,Inc.AllRightsReserved.

SG

IntroductionStateoftheMarket

2024StateoftheGenerativeAIMarket|11

GenAIImplementationRoadmapDifferentiationinaWorldofGenAI

GenAIImplementationRoadmap

?2024InformationServicesGroup,Inc.AllRightsReserved.

HowisGenAImoneybeingspent?

36%

Applications/software

(incl.softwareasaservice)

21%

Infrastructure

(e.g.,storageandcompute)

25%

Personnel

(incl.contractorsandstaffaugmentation)

18%

Outsourced

(managedservices)

?2024InformationServicesGroup,Inc.AllRightsReserved.

1SG

Introduction

StateoftheMarket

GenAIImplementationRoadmap

DifferentiationinaWorldofGenAI

2024StateoftheGenerativeAIMarket|12

ImpactsonBothBusinessandIT

WhereisGenAImakinganimpact?

IT

Facilities

16%

65%

HR/recruitment

Customerservice/contactcenter

28%

59%

Production/manufacturing/engineering

31%

Product/servicedevelopment

GenAIImpactedFunctions

54%

Finance/

31%

accounting

Sales/businessdevelopment

Procurement/

53%

36%

purchasing

44%

48%

Sharedservices/operations

45%

Marketing/PR/media

Compliance/legal/riskmanagement

DataSource:ISG2024MarketLens??GenAIUseCaseStudy

1SG

Introduction

StateoftheMarket

GenAIImplementationRoadmap

DifferentiationinaWorldofGenAI

GenAIfortheBusiness

BusinesssoftwarevendorsacrossvarioussectorsarerapidlyenhancingtheircorefunctionalitybyincorporatingGenAI

features.Whilesomecapabilitiesarealreadyavailable,

productroadmapsindicateasignificantaccelerationin

theintegrationofGenAIoverthenext18months.CertainGenAIfunctionalitieswillbeofferedatnoadditionalcharge,butotherswillrequireextrainvestment,enablingproductproviderstocreateandmaintainAI-enabledapplicationsasconsumableoptionsforcustomers.

Wemeasurethatenterprisesarewilling

tospend6%–9%moreonaper-userbasis

forAIfunctionalityacrossawiderangeofbusinessapplications.

Theseadvancementsaddressthepressingissues

enterprisesfaceinformingtheirAIstrategies.ByleveragingAIandGenAI,softwaredevelopersaimtoempoweruserstoworkfaster,smarter,moreeconomicallyandwithgreatersafety.AI-enabledapplicationsspecificallytargetenterprise

challenges,suchasalackofAIexpertiseandtheneedforacohesive,company-widevisionforAIimplementation.

TheadvantagesofAIcanbetailoredtodomain-specific

needs.Forinstance,incustomerexperience(CX),AI

enhancescustomerinsightsatscale,transformsself-

servicechatbotsandaugmentstheagents’abilitytoservecustomersinrealtime.Infinanceandaccounting,AI

improvesprocesses,suchaspredictiveanalytics-driven

forecasting,errordetectionandtheever-importantmonth-endclosecycle.Meanwhile,humancapitalmanagement

(HCM)benefitsfromAIthroughenhancedemployeeself-service,onboarding,andskillsassessments.Inrevenue

management,AIelevatesforecasting,real-timemonitoring,planningandoptimization.

DuringGenAIinvestmentplanning,enterprisesmust

notoverlookthedevelopmentroadmapsoftheirthird-

partyvendors.Ignoringtheplansofthesevendorscan

meanredundantinvestmentandinefficientallocationof

resources.Afteraccountingforvendorplans,enterprisescanthendeveloptheirowndynamicAIstrategiesthat

clearlydelineateareasdesignatedforinternaldevelopmentversusthosethatwillusethird-partyapplicationsor

serviceproviders.

DataSource:ISG2024MarketLens??GenAIUseCaseStudy;ISG2024MarketLens??AIStudy

?2024InformationServicesGroup,Inc.AllRightsReserved.

2024StateoftheGenerativeAIMarket|13

Onaverage,enterpriseshaveimplemented

151

GenAI-enabledapplications.And,theyestimateitwill

increaseto

356

GenAIapplicationsbytheendof2025.

1SG

Introduction

StateoftheMarket

GenAIImplementationRoadmap

DifferentiationinaWorldofGenAI

GenAIforIT–AILandingZones

ManyenterpriseslookingtoadoptAIatscalestruggletounderstandtheinfrastructurerequiredtopowertheirownAIusecasesthatwilldrivethemajorityofAIvalue.The

globalsupplyofcomputeisnowtryingtokeepupwiththedemandofAI.WhileGPUsandthepublicclouddominatethemarkettorunAIworkloads,neitherwerenatively

builttorunthemwiththeappropriateperformanceandcostconsiderations.

EnterprisesshouldexploreseveralkeyconsiderationsforwhereandhowtohosttheirAIworkloads.

WheretoHostAI:On-premvs.PublicCloud

Thepubliccloud(i.e.,AWS,Azure,GoogleCloud,etc.)hasbeenthedefactolandingzoneforAIworkloadsover

thelasttwoyears,andwhilemostenterpriseslookingtoexpandtheirAIcapabilitieswillcontinuetoincreasetheircloudfootprint,manyarelookingatalternativestothepubliccloud.

28%

ofenterprises

expecttotrain

theirLLMsin

privatecloudsor

on-prem

21%

ofenterprises

expectto

purchaseGPUs

foron-premiseAI

development

Publiccloudprovidershavespentbillionsincapital

investmentsinanticipationofmassiveAIadoption,and

theeaseofaccesshasmadeprovisioningthisAI-ready

infrastructureinthepubliccloudtheeasychoiceforproofofconcepts(POCs).Morerecently,privatecloudprovidersandnicheAIcloudprovidersareenteringthemarketto

provideenterprisesmoreoptionsforadditionalsecurityorperformance.

DataSource:ISG2024MarketLens??AIStudy

?2024InformationServicesGroup,Inc.AllRightsReserved.

2024StateoftheGenerativeAIMarket|14

AI-ReadyInfrastructureComponents

WhethertheAI-readyinfrastructureisdeployedina

publiccloud,on-premiseorelsewhere,enterprises

faceadditionalconsiderationsfortheirAIdeployments,whicharedescribedtotheright.

Choosewhatisbestforyourenterprise

Ultimately,enterprisesmustmaketheirinfrastructuredecisionsbasedontheirownrequirementsfor

performance,availability,budget,fungibility,

integrationandsecurity.Theserequirements,and

theunderlyinginfrastructure,mayvarywidely

basedontheselectedusecaseandthestageofAI

development,suchastrainingversusinference.In

2024,enterpriseshavemanychoicesandevenmoreconsiderationsfordeployingtheirAIworkloadsontheappropriateinfrastructure.

EnterprisesshouldallowtheirusecasestoguidetheirAIinfrastructureinvestments,understandingthat

infrastructuremodernizationisadeliberatejourneyratherthanaproblemthatdemandsimmediate

resolution.AIinfrastructureisadvancingatarate

rarelyseeninhardwaremarkets,soenterprisesmustevaluatetheirinvestmentsbasedonthepoint-in-timecapabilitiesandpriceoftheinfrastructureagainstthevalueitwilldeliver.

1SG

Introduction

StateoftheMarket

GenAIImplementationRoadmap

DifferentiationinaWorldofGenAI

2024StateoftheGenerativeAIMarket|15

Compute

AIaccelerators,orspecializedhardwareforimprovedAI/MLprocessing,arecriticaltoeffectivelytrainandrunAIworkloads.GPUscontinuetodominateAIacceleratordiscussionsduetotheinherentsuperiorityofparallelprocessingoverCPU-basedserialcomputingforhandlingAI.However,newentrantsintothesiliconmarketlikeCerebras,Groqandbignameslike

Google,MicrosoftandApplearebuildingperformantAIcomputingforspecificusecasesliketraining,inferenceorend-usercomputingplatforms.AIacceleratorsareincrediblyexpensiveandmaybecost-prohibitiveformanyenterprisesatscale.

Storage

AIexcelsatmanaginglargenumbersofparametersandlarger,morevariedstructuredandunstructureddatasets.

Enterprisesmustleveragehighlyscalable,performant,availableandresilientstoragesolutionstoload,storeandaccesstherequireddataandmodels.Thesesolutionsmustbeveryusabletofindandorganizethedatastoredwithinthem.

Enterprisesareincreasinglylookingtowardsoftware-definedobjectstoragesolutionsfortheirAIstorageinfrastructureovertraditionalNASorSANsolutions.

Network

ThedistributednatureofAItrainingacrossthousandsofnodesresultsinasignificantportionoftheAItrainingtimespentinthenetwork.TrainingAIworkloadsefficientlyrequiresveryhigh-bandwidthnetworks(100Gbps+)thatcanperformattherequiredthroughputandlatency.Enterprisesmustensurethattheirdatacentershavetheappropriatenetworking

setuptomanagethetrafficrequiredfortraining.

Facilities

Theseinfrastructurecomponentsmustresidewithinafacilitythatiscapableofhandlingtheadditionalenergyandcooling

requirementsofAIworkloads.TrainingandrunningAIworkloadsdemandstremendousamountsofenergy,andthechipsrequiresubstantialcooling.Manyenterprisesarelookingtoliquidsolutionssuchasimmersioncoolingordirect-to-chip

coolingtoaugmenttraditionalmethods.Companieswith“green”initiativesmustbecogniz

溫馨提示

  • 1. 本站所有資源如無特殊說明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請下載最新的WinRAR軟件解壓。
  • 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請聯(lián)系上傳者。文件的所有權益歸上傳用戶所有。
  • 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁內容里面會有圖紙預覽,若沒有圖紙預覽就沒有圖紙。
  • 4. 未經(jīng)權益所有人同意不得將文件中的內容挪作商業(yè)或盈利用途。
  • 5. 人人文庫網(wǎng)僅提供信息存儲空間,僅對用戶上傳內容的表現(xiàn)方式做保護處理,對用戶上傳分享的文檔內容本身不做任何修改或編輯,并不能對任何下載內容負責。
  • 6. 下載文件中如有侵權或不適當內容,請與我們聯(lián)系,我們立即糾正。
  • 7. 本站不保證下載資源的準確性、安全性和完整性, 同時也不承擔用戶因使用這些下載資源對自己和他人造成任何形式的傷害或損失。

評論

0/150

提交評論