Tesla (TSLA) Q1 2024 Earnings Call Transcript财报电话会议记录.docx

《Tesla (TSLA) Q1 2024 Earnings Call Transcript财报电话会议记录.docx》由会员分享，可在线阅读，更多相关《Tesla (TSLA) Q1 2024 Earnings Call Transcript财报电话会议记录.docx（27页珍藏版）》请在课桌文档上搜索。

1、Tesla,Inc.(TS1.A)Q12024EarningsCallTranscriptApr.23,202410:30PMETTesla,Inc.(TS1.A)Stock,TS1.A:CAStockCompanyParticipantsMartinViecha-VicePresident-InvestorRelationsElonMusk-ChiefExecutiveOfficerVaibhavTaneja-ChiefFinancialOfficer1.arsMoravy-VicePresident-VehicleEngineeringAshokElluswamy-Director-Aut

2、opilotSoftwareConferenceCallParticipantsTonySacconaghi-BernsteinAdamJonas-MorganStanleyAlexPotter-PiperSandlerMarkDelaney-GoldmanSachsColinRusch-OppenheimerBenKallo-BairdShreyasPatil-WolfeResearchMartinViechaTeslasFirstQuarter2024Q&AWebcast.MynameisMartinViecha,VPofInvestorRelations,andmjoinedtodayb

3、yElonMusk,VaibhavTaneja,andanumberofotherexecutives.OurQ1resultswereannouncedatabout3.00p.m.CentralTimeintheUpdateDeckwepublishedatthesamelinkasthiswebcast.Duringthiscall,wewilldiscussourbusinessoutlookandmakeforward-lookingstatements.Thesecommentsarebasedonourpredictionsandexpectationsasoftoday.Act

4、ualeventsandresultscoulddiffermateriallyduetoanumberofrisksanduncertainties,includingthosementionedinourmostrecentfilingswiththeSEC.Duringthequestion-and-answerportionoftodayscall,pleaselimityourselftoonequestionandonefollow-up.Pleaseusetheraisehandbuttontojointhequestionqueue.ButbeforewejumpintoQ&A

5、,Elonhassomeopeningremarks.Elon?ElonMuskThanks,Martin.SotorecapinQ1wenavigatedseveralunforeseenchallengesaswellastherampoftheupdatedModel3inFremont.Therewas,asweallhaveseen,theEVadoptionrategloballyisunderpressureandalotofotherautomanufacturersarepullingbackonEVsandpursuingplug-inhybridsinstead.Webe

6、lievethisisnottherightstrategyandelectricvehicleswillultimatelydominatethemarket.Despitethesechallenges,theTeslateamdidagreatjobexecutinginatoughenvironmentandenergystoragedeployments,theMegapackinparticular,reachedanalltimehighinQ1,leadingtorecordprofitabilityfortheenergybusiness,andthatlookslikely

7、tocontinuetoincreaseinthequartersandyearsahead.Itwillincrease.Weactuallyknowthatitwill,sosignificantlyfasterthanthecarbusinessasweexpected.WealsocontinuetoexpandourAltrainingcapacityinQ1,morethandoublingourtrainingcomputesequentially.Intermsofthenewproductroadmap,therehasbeenalotoftalkaboutourupcomi

8、ngvehiclelineinthenext-inthepastseveralweeks.Weveupdatedourfuturevehiclelineuptoacceleratethelaunchofnewmodelsahead,previouslymentionedstartupproductioninthesecondhalfof2025,soweexpectittobemoreliketheearly2025,ifnotlatethisyear.Thesenewvehicles,includingmoreaffordablemodels,willuseaspectsofthenextg

9、enerationplatformaswellasaspectsofourcurrentplatforms,andwillbeabletoproduceonthesamemanufacturinglinesasourcurrentvehiclelineup.Soitsnotcontingentonanynewfactoryormassivenewproductionline.Itllbemadeonourcurrentproductionlinesmuchmoreefficiently.Andwethinkthisshouldallowustogettoover3millionvehicles

10、ofcapacitywhenrealizedtothefullextent.RegardingFSDVersion12,whichisthepureAl-basedself-driving,ifyouhaventexperiencedthis,Istronglyurgeyoutotryitout.Itsprofoundandtherateofimprovementisrapidso-andwe,venowturnedthatonforallcarswiththecamerasandinferencecomputerandeverythingfromHardware3oninNorthAmeri

11、ca.Andsoitsbeenpushedoutto,Ithink,around1.8millionvehiclesandwereseeingabouthalfofpeopleuseitsofarandthatpercentageisincreasingwitheachpassingweek.Sowenowhaveover300billionmilesthathavebeendrivenwithFSDV12.Sincethelaunchoffullself-driving,supervisedfullself-driving,itsbecomeveryclearthatthevision-ba

12、sedapproachwithendtoendneuralnetworksistherightsolutionforscalableautonomy.Itsreallyhowhumansdrive.Ourentireroadnetworkisdesignedforbiologicalneuralnetsandeyes.Sonaturallycamerasanddigitalneuralnetsarethesolutiontoourcurrentroadsystem.Tomakeitmoreaccessible,wevereducedthesubscriptionpriceto$99amonth

13、,soitseasytotryout.Andasweveannounced,wellbeshowcasingourpurpose-builtrobotaxi,orCybercab,inAugust.Yes.RegardingAlcompute,overthepastfewmonths,wevebeenactivelyworkingonexpandingTeslascoreAlinfrastructure.Forawhilethere,weweretrainingconstrainedinourprogress.Weare,atthispoint,nolongertrainingconstrai

14、nedandsoweremakingrapidprogress.Weveinstalledandcommissioned,meaningtheyreactuallyworking35,000H100computersorGPUs,GPUiswrongword,theyneedanewword.IalwaysfeellikeawincewhenIsayGPUbecauseitsnot-GPUstands-Gstandsforgraphics,anditdoesntdographics.Butanywayroughly35,000H100sareactive,andweexpectthattobe

15、probably85,000orthereaboutsbytheendofthisyearandtraining,justfortraining.Wearemakingsurethatwerebeingasefficientaspossibleinourtraining.ItsnotjustaboutthenumberofH100s,buthowefficientlytheyreused.Soinconclusion,weresuperexcitedaboutourautonomyroadmap.IthinkitshouldbeobvioustoanyonewhosdrivingVersion

16、12andittellsthatthatitisonlyamatteroftimebeforeweexceedthereliabilityofhumansandnotmuchtimewiththat.Andwerereallyheadedforanelectricvehicle,anautonomousfuture.AndllgobacktosomethingIsaidseveralyearsagothatinthefuture,gasolinecarsthatarenotautonomouswillbelikeridingahorseandusingaflipphone.Andthatwil

17、lbecomeveryobviousinhindsight.WecontinuetomakethenecessaryinvestmentsthatwilldrivegrowthandprofitsforTeslainthefuture,andIwantedtothanktheTeslateamforincredibleexecutionduringthisperiodandlookforwardtoeverythingthatwehaveplannedahead.Thanks.MartinViechaThankyouverymuch,andVaibhavhassomecommentsaswel

18、l.VaibhavTanejaThanks.ItsimportanttoacknowledgewhatElonsaid,fromourautobusinessperspective.Wedidseeaseasonaldeclineinrevenuesquarter-over-quarterandthosewereprimarilybecauseofseasonality,uncertainmacroeconomicenvironmentandtheotherreasons,whichElonhadmentionedearlier.Automarginsdeclinedfrom18.9%to18

19、.5%.ExcludingtheimpactofCybertruck,theimpactofpricingactionswaslargelyoffsetbyreductionsinperunitcostsandtherecognitionofrevenuefromAutoparkfeatureforcertainvehiclesintheU.S.thatpreviouslydidnothavethatfunctionality.Additionally,whilewedidexperiencehighercostduetotherampofModel3inFremontanddisruptio

20、nsinBerlin,thesecostswerelargelyoffsetbycostreductioninitiatives.Infact,ifweexcludeCybertruckandFremontModel3rampcosts,therevenuefromAutopark,automarginsimprovedslightly.CurrentlynormalizedModelYcostpervehicleinAustinandBerlinarealreadyveryclosetothatofFremont.Ourabilitytoreducecostswithoutsacrifici

21、ngonqualitywasduetotheamazingeffortsoftheteam,inexecutingTeslasrelentlesspursuitofefficiencyacrossthebusiness.WevealsowitnessedthatasotherOEMsarepullingbackontheirinvestmentsinEV,thereisincreasingappetiteforcredits,andthatmeansasteadystreamofrevenueforus.Obviously,seeingotherspullbackfromEVisnotthef

22、uturewewant.Wewouldpreferitthewholeindustrywentallin.Onthedemandfront,weveundertakenavarietyofinitiatives,includingloweringthepriceofboththepurchaseandsubscriptionoptionsforFSDlaunchingextremelyattractiveleasingspecialsfortheModel3intheU.S.for$299amonthandofferingattractivefinancingoptionsincertainm

23、arkets.Webelievethatourawarenessactivities,pairedwithattractivefinancing,willgoalongwayinexpandingourreachanddrivingdemandforourproducts.OurEnergybusinesscontinuestomakemeaningfulprogresswithmarginsreachingarecordof24.6%.Weexpecttheenergystoragedeploymentsfor2024togrowatleast75%higherfrom2023.Andacc

24、ordingly,thisbusinesswillbegincontributingsignificantlytoouroverallprofitability.NotethatthereisabitofIumpinessinourstoragedeploymentsduetoavarietyoffactorsthatareoutsideofourcontrol,sodeploymentsmayfluctuatequarter-over-quarter.Ontheoperatingexpensefront,wesawasequentialincreasefromourAlinitiatives

25、,continuedinvestmentinfutureprojects,marketingandotheractivities.Wehadnegativefreecashflowof$2.5billioninthefirstquarter.Theprimarydriverofthiswasanincreaseininventoryfromamismatchbetweenbuildsanddeliveriesasdiscussedbefore,andourelevatedspendonCapExacrossvariousinitiatives,includingAlcompute.Weexpe

26、cttheinventorybuildtoreverseinthesecondquarterandfreecashflowtoreturntopositiveagain.Aswepreparethecompanyforthenextphaseofgrowth,wehadtomakethehardbutnecessarydecisiontoreduceourheadcountbyover10%.Thesavingsgeneratedareexpectedtobewellinexcessof$1billiononanannualrunratebasis.Wearealsogettinghyperf

27、ocusedonCapExefficiencyandutilizingourinstalledcapacityinamoreefficientmanner.Thesavingsfromtheseinitiatives,includingourcostreductionswillhelpimproveouroverallprofitabilityandultimatelyenableustoincreasethescaleofourinvestmentsinAl.Inconclusion,thefutureisextremelybrightandthejourneytogettherewhile

28、challengingwillbeextremelyrewarding.Onceagain,IwouldliketothankthewholeTeslateamfordeliveringgreatresults.AndwecanopenituptoQ&A.Question-and-AnswerSessionA-MartinViechaOkay.1.etsstartwithinvestorQ&A.Thefirstquestionis,whatisthestatusof4680.Whatisthecurrentoutput?1.ars?1.arsMoravySure.4680productioni

29、ncreasedabout18%to20%fromQ4reachinggreaterthan1KaweekforCybertruck,whichisabout7gigawatthoursperyearaswepostedonX.WeexpecttostayaheadoftheCybertruckrampwiththecellproductionthroughoutQ2aswerampthethirdoffourlinesinPhase1,whilemaintainingmultipleweeksofcellinventorytomakesurewereaheadoftheramp.Becaus

30、ewereramping,COGScontinuestodroprapidlyweek-overweekdrivenbyyieldimprovementsthroughoutthelinesandproductionvolumeincreases.Soourgoal,andweexpecttodothisistobeatsuppliercostofnickel-basedcellsbytheendoftheyear.MartinViechaThankyou.ThesecondquestionisonOptimus.SowhatisthecurrentstatusofOptimus?Arethe

31、ycurrentlyperforminganyfactorytasks?Whendoyouexpecttostartmassproduction?ElonMuskWeareabletodosimplefactorytasksoratleast,Ishouldsay,factorytasksinthelab.Intermsof-wedothinkwewillhaveOptimusinlimitedproductioninthenaturalfactoryitself,doingusefultasksbeforetheendofthisyear.AndthenIthinkwemaybeableto

32、sellitexternallybytheendofnextyear.Thesearejustguesses.Asvesaidbefore,IthinkOptimuswillbemorevaluablethaneverythingelsecombined.Becauseifyouvegotasentienthumanoidrobotsthatisabletonavigaterealityanddotasksatrequest,thereisnomeaningfullimittothesizeoftheeconomy.Sothatswhatisgoingtohappen.AndIthinkTes

33、laisbestpositionedofanyhumanoidrobotmakertobeabletoreachvolumeproductionwithefficientinferenceontherobotitself.ImeanthisperhapsisapointthatisworthemphasizingTeslasAlinferenceefficiencyisvastlybetterthananyothercompany.ThereisnocompanyevenclosetotheinferenceefficiencyofTesla.Wevehadtodothatbecausewew

34、ereconstrainedbytheinferencehardwareinthecar,wedidnthaveachoice.Butthatwillpaydividendsinmanyways.MartinViechaThankyou.Thethirdquestionis,whatisthecurrentassessmentofthepathwaytowardsregulatoryapprovalforunsupervisedFSDintheU.S.Andhowshouldwethinkabouttheappropriatesafetythresholdcomparedtohumandriv

35、ers?ElonMuskSure.1.arsMoravyIcanstart.Thereareahandfulofstatesthatalreadyhaveadoptedautonomousvehiclelaws.Thesestatesarepavingthewayforoperations,whilethedataforsuchoperationsguidesabroaderadoptionofdriver-lessvehicles.IthinkAshokcantalkalittlebitaboutoursafetymethodology,butweexpectthatthesestatesa

36、ndtheworkongoingaswellasthedatathatwe,reprovidingwillpaveawayforabroad-basedregulatoryapprovalintheU.S.atleastandtheninothercountriesaswell?AshokElluswamyYes.Itsactuallybeenprettyhelpfulthatotherautonomouscarcompanieshavebeencuttingapaththroughtheregulatoryjungle,whichisabsurd.Thatsactuallyquitehelp

37、ful.AndtheyhaveobviouslybeenoperatinginSanFranciscoforawhile.IthinktheygotapprovalforCityof1.A.Sotheseapprovalsarehappeningrapidly.Ithinkifyouvegotatscale,astatisticallysignificantamountofdatathatshowsconclusivelythattheautonomouscarhas,letssay,halftheaccidentrateofahuman-drivencar,Ithink,thatsdiffi

38、culttoignorebecauseatthatpoint,stoppingautonomymeanskillingpeople.SoIactuallydonotthinkthattherewillbesignificantregulatorybarriersprovidedtherewasconclusivedatathattheautonomouscarissaferthanahuman-drivencar.Andinmyview,thiswillbemuchlikeelevators.Elevatorsusedtobeoperatedbyaguywithrelayswitch.Buts

39、ometimesthatguywouldgettiredordrunkorjustmakeamistake,andshattersomebodyinhalfbetweenfloors.Sowejustgetanelevatorandpressbutton,wedontthinkaboutit.Infact,itskindofweirdifsomebodyisstandingtherewitharelayswitch.Andthatwillbehowcarswork.Youjustsummonthecarusingyourphone,yougetin,ittakesyoutoadestinati

40、on,yougetout.VaibhavTanejaYoudonteventhinkaboutit?ElonMuskYoudonteventhinkaboutit.Justlikeanelevator,ittakesyoutoyourfloor.Thatsit.Dontthinkabouthowtheelevatorisworkingoranythinglikethat.AndsomethingIshouldclarifyisthatTeslawillbeoperatingthefleet.SoyoucanthinkoflikehowTesla,thinkofitslikesomecombin

41、ationofAirbnbandUber,meaningthattherewillbesomenumberofcarsthatTeslaownsitselfandoperatesinthefleet.Therewillbesomenumberofcarsandthentherellbeabunchofcarswheretheyreownedbytheenduser.Thatendusercanaddorsubtracttheircartothefleetwhenevertheywant,andtheycandecideiftheywanttoonlyletthecarbeusedbyfrien

42、dsandfamilyoronlyby5-starusersorbyanyoneatanytimetheycouldhavethecarcomebacktothemandbeexclusivelytheirs,likeanAirbnb.Youcouldrentoutyourguestroomornot,anytimeyouwant.Soasourfleetgrows,wehave7millioncarsgoingto-9millioncarsgoingto,eventuallytensofmillionsofcarsworldwide.Withaconstantfeedbackloop,eve

43、rytimesomethinggoeswrong,thatgetsaddedtothetrainingdataandyougetthistrainingflywheelhappeninginthesamewaythatGoogleSearchhasthesortOfflywheel,itsverydifficulttocompetewithGooglebecausepeopleareconstantlydoingsearchesandclickingandGoogleisgettingthatfeedbackloop.IfsthesamewithTesla.Butatascalethatism

44、aybedifficulttocomprehend,butultimately,itwillbetensofmillions.IthinktheresalsosomepotentialhereforanAWSelementdowntheroadwhereifwe,vegotverypowerfulinferencebecausewevegotaHardware3inthecars,butnowallcarsarebeingmadewithHardware4.Hardware5isprettymuchdesignedandshouldbeincars,hopefullytowardstheend

45、ofnextyear.Andtheresapotentialtorun-whenthecarisnotmovingtoactuallyrundistributedinference.SokindoflikeAWS,butdistributedinference.1.ikeittakesalotofcomputerstotrainanAlmodel,butmanyordersofmagnitudelesscomputetorunit.Soifyoucanimaginefuture,perhapswheretheresafleetof100millionTeslas,andonaverage,th

46、eyvegotlikemaybeakilowattofinferencecompute.Thats100gigawattsofinferencecomputedistributedallaroundtheworld.Itsprettyhardtoputtogether100gigawattsofAlcompute.Andeveninanautonomousfuturewherethecaris,perhaps,usedinsteadofbeingused10hoursaweek,itisused50hoursaweek.Thatstillleavesover100hoursaweekwhere

47、thecarinferencecomputercouldbedoingsomethingelse.Anditseemslikeitwillbeawastenottouseit.MartinViechaAshok,doyouwanttochimeinontheairprocessandsafety?AshokElluswamyYes,wehavemultipletiersofvalidatingthesafetyinanygivenweek,wetrainhundredsofneuralnetworksthatcanproducedifferenttrajectoriesforhowtodriv

48、ethecar,wereplaythemthroughthemillionsofclipsthatwehavealreadycollectedfromourusersandourownQA.Thosearelikecriticalevents,likesomeonejumpingoutinfrontorlikeothercriticaleventsthatwehavegathereddatabaseovermany,manyyears,andwereplaythroughallofthemtomakesurethatwearenetimprovingsafety.Andontopofit,we

49、havesimulationsystemsthatalsotrytorecreatethisandtestthisinclosedloopfashion.Andsomeofthisisvalidated,wegiveittoourownQAdrivers.Wehavehundredsofthemindifferentcities,inSanFrancisco,1.osAngeles,Austin,NewYork,alotofdifferentlocations.Theyarealsodrivingthisandcollectingreal-worldmiles,andwehaveanestimateofwhatarethecriticalevents,aretheyanetimprovementcomparedtothepreviousweeksbuilds.Ando