NVIDIA (NVDA) 22 Feb 232023 Q4 Earnings call transcript

Good afternoon. My name is Emma, and I will be your conference operator today. At this time, I would like to welcome everyone to the NVIDIA's Fourth Quarter Earnings Call. [Operator Instructions]. Thank you. begin may you Jankowski, your conference. Simona

and discuss Good call our is webcast for of you. me Vice the XXXX. from Huang, President afternoon, Kress, Financial and fiscal live to webcast everyone, conference on to financial being transcribed conference prior NVIDIA's consent. be today remind until It for President Jensen replay are The is without I'd to Relations the and The NVIDIA's for welcome NVIDIA written or property. Executive the available Executive call will website. fiscal call With be Officer; first Colette today's and the results XXXX. Thank Chief of Officer. fourth like reproduced can't our Chief of NVIDIA's quarter Investor that quarter call content you

may statements During on our number significant this risks based and to current expectations. results materially. make we and a may are subject of differ call, uncertainties, forward-looking These actual

For most a our and we on disclosure release, could earnings reports are X-K future discussion available update based any Exchange and All and with assume made to in today's obligation as statements. we the XX, to no us. Commission. the Form our that please XX-Q required file refer business, that February such of Forms the XX-K factors recent may as financial results Except affect on our Securities today, to statements of currently and XXXX, information by law,

will call, we this discuss non-GAAP measures. financial During

commentary, which website. on GAAP these financial to CFO measures a financial in our of find our posted reconciliation non-GAAP can You is measures

to that, me over Colette. the With let turn call

plans. was the was flat $X.XX was year-on-year. down X% X% sequentially, you, up billion providers paused as Hyperscale softer up we COVID on believe service lower timing recalibrate total at by and Revenue posted with grew Full build to though is sales GPUs CPU is growth, bit end from strong. year less general of Though Starting demand QX the revenue macroeconomic center revenue $XX the market with for XX% our issue revenue Fiscal Thank we reflects of year. expectations and was line year AI overall $X.XX center. and a data for up than sequentially was and billion but end expected $XX decline demand which see Simona. some customer issues. domestic billion, largely Networking purpose China, data year other driven short expectations, year-on-year. billion, that prior generally uncertainty, infrastructure. reflecting in infrastructure tightening in this XX%. revenue their was cloud our sequential a as revenue our sequential was The strong the

our service of X-quarter cloud With outpaced providers, of revenue. continuing approach. expanding are Oracle we Center customers of CSPs. to GPU customers Data about moved cloud Data adoption growth list a Revenue On serving basis, as fast-growing specialized a cloud-first as customers an to more year from that significantly whole CSP enterprise Center CSP a including drove trailing and XX% grow, last

which transformer-based inference training out large much testament Adoption of AXXX, time up of of development than the declined to strong. of GPU H-XXX ramp, H-XXX, This our its models. which and just inferencing as the just cluster is sequentially. revenue is higher that already new performance center language H-XXX arrived than The scale is In flagship transformer much Xx the faster was for serve second H-XXX a large and exceptional models. of engine of language XXx on as in in quarter the of of the AXXX

range of neural with video do documents an summarizing every AI's firsthand possible and customer ads or like wide creating generative [indiscernible], questions. whether people answering allowing new generating can worldwide, models images productivity experience to more or of faster. has for at is almost ChatGPT showing copy, will network point. a types games Open improve interest industry what's adoption help marketing AI AI like Generative inflection text AI. tasks, in captured AI These applications

language billion can are forward opportunities. methods and networks to world. customers models algorithms, these large We with look neural to the bring generative expertise AI in most the Generative advanced with today's data XXX processing over NVIDIA's supercomputers, across spans enterprise. capabilities training parameters helping AI that

the major will consumer in addition companies, are hyperscale In we the driving with engaged center strong significant every enterprises and Internet cloud provider, through start-ups. that accelerate opportunity is to with many data and growth working The year.

made and we During sector, partnership industry software. in in enterprise of financial accelerate AI a fraud Together, AI, and the to machine in and verticals. NVIDIA, the learning leveraging with Bank Deutsche speech are We including of announcements notable of largest including announced customer applications, on-premise services. use the financial one detection services we both the quarter, cloud, our a NVIDIA's stack, process developing virtual agents, service full computing range bank AI automation,

platforms inference results discovery. financial that we see InfiniBand by networking, generation also asset for for captured In and We NVIDIA AI. services in such industry our price key announced latest as benchmark a HPC optimized AI fueled leading for growing applications Ethernet demand

for scale, exponential the workloads. in-network high-performance computing cloud and multi-node unmatched grow model sizes driving power-efficient InfiniBand at choice to Delivering the generative accelerated clear continue networking AI. latency performance, scale for out Generative foundation need capabilities, AI rates, to is

networking our integrating and XX growth higher XX is platform the scale transition is and gigabit customers. start, enterprise second Quantum bringing by led second advanced switches. smaller platform per the full Spectrum off momentum cloud, Ethernet per most with great a expertise In demand to its as X world's supercomputing Ethernet, For accelerated stack speeds, NVIDIA across high-performance gigabit our In deployments, and customers gaining X adopters InfiniBand driven our next-generation fabrics. it as is to quarter,

and frameworks more of X.X We our version and contact center, for model BioNeMo NVIDIA transcription currently Upcoming for customers. We NeMo software intelligent virtual pretrained cybersecurity. with are workflows model and released AI large and remain new AI with and early which our language services, enterprise support expanding offerings assistance, access audio in NVIDIA focused XX on services. include than

Now to cloud more a Jensen and about our software business. bit talk to

Thanks, inflection accumulation at Colette. the around to for processing Yet, AI deploy AI's algorithms, within and enterprise every put most. level an reach model and a remain point. of insurmountable AI The share supercomputer Generative enterprises I to our infrastructure, urgency you Today, an AI the help breakthroughs obstacle with to capability model and of brought techniques to want develop business training data has the of versatility customer. next technology of AI strategies. has triggered sense world

storage, NVIDIA advanced our AI to We to the -- of most AI remaining by AI cloud as and the the offer with a services, hosted world's while service offers world's easy cloud security largest cloud access service platform, are advanced partnering major by directly the services close and enterprises service offered through offered and networking, NVIDIA to clouds. go-to-market partners, network within most world's clouds. providers NVIDIA

through supercomputers blueprint or Customers can NVIDIA supercomputer, are an engage is AI built have build. around world. browser. time-consuming NVIDIA AI NVIDIA Google and already the DGX others Oracle AI NVIDIA and open software AI DGX the Cloud the Cloud, supercomputer, Azure, to being way. cloud announcing on are DGX GCP is AI DGX services AI and hard we and way easiest Cloud available pretrained factories your Microsoft to fastest of the the at acceleration own your AI Today, layers. model Infrastructure and just the library supercomputer,

for With engage and new pretrained full training enterprise their customers model, platform software our offering and enterprise access AI layer, the AI who businesses. customers NVIDIA's AI for AI to generative proprietary customers generative we of other their to or scale NeMo models models, AI at layer, private computing can to At the will build and BioNeMo, can And NVIDIA public want customizable model any large language AI business AI deploying be across models cloud. services workloads.

upcoming to so be will share services in. details our about cloud tune NVIDIA We GTC more sure at AI

Now Colette gaming. it let turn to me back on

GPUs the XX% regions. by our most which a is $X.XX XX% ago. Thanks, impact XX%. seasonally strong behind is revenue year Lovelace was of demand was correction, channel and billion us. in the of up year XX Fiscal revenue down RTX down Sequential growth solid from And Jensen. of driven Ada on GeForce $X.XX inventory the largely the reception decline Gaming sequentially in of Series was reflects year-on-year architecture. The strong quarter based fourth billion

we in that While China encouraged was the signs market. somewhat related by recovery by of to disruptions early impacted COVID, are

architecture, XXXX the in and graphic. to outlets online Gamers selling the reflecting new flagship popularity many top with out stock. The claim shot enthusiastically Steam Ti retail has to the are RTXXXXX, for spot for GPUs, AI quickly gamers' desktop XXXX desire of responding quickly up high-performance on RTX XXXX,

based phase performance in reached shelves, on month, the Ada laptops leap and efficiency. this of generational Earlier power first the NVIDIA's architecture retail largest-ever gaming delivering

the we For enthusiast-class GPU segment, are as performance fast-growing slim to XX a first basic as inches, limited laptops and time, previously to apps. tasks bringing

will thanks and fifth-generation great XX of over in, most bringing GPUs our power back-to-schools In creator power the gaming the we RTX class are a our XXX Series to models, first, GPUs, setting for to another performing up laptops, XX laptops, All Max-Q efficiency season. technology.

AI-powered architecture for over graphics, our AI-powered XXX features massively There tracing now technology performance. AI ray NVIDIA's RTX The and X, real-time supporting applications and are third-generation boosts games which graphics. DLSS

games, advanced to recently XXXX, performance X Cyberpunk resolution. X Xx added most DLSS at a in With boost XK the rate frame enabling

X. million service enabled in graphics multiple XXX second It now tier. stream XXXX XX than titles NOW members XXX month, and DLSS up members horsepower a to with has and Last RTX cloud from Our Ultimate tracing it cloud ultimate gaming high-performance countries. ray to more can in over per GeForce performance. at full new in expand the users, continued frames membership dimensions,

PC a and which we with of lineup important it like Microsoft's Call GeForce Activision bring And to XX-year close add made acquisition, Microsoft's NOW just Microsoft. blockbusters titles Duty agreed like of Flight Halo the partnership Simulator. will Minecraft, of Overwatch. Xbox to yesterday, an includes games, We upon and to And announcement

of sequentially was of year XX% Moving to the from manufacturing reflects Fiscal impact the down billion the correction, ago. $XXX Pro first of Sequential up XX%. strengths was desktop was Year-on-year year. and in revenue the decline by end half Revenue verticals. to XX% channel the year Visualization. a $X.XX automotive down driven expect which and we of growth workstations industrial million in inventory with

released downloads far, real-time XXX NVIDIA's to of design AI-powered Omniverse almost to continues with XK, including through in so latest number and XD and for for Omniverse tracing, Interest path cloud third-party enhancements, applications. Omniverse AWS. Search build XXX,XXX for large untagged support search a databases, The Omniverse has connectors containers features

primarily driven a by AI Revenue up up a from $XXX record automotive. ago. was growth to from million, XX% and XXX% move was solutions. year automotive Let's Sequential [indiscernible]

at ramps program traditional year vehicle New and drive million of customers this up XX%. helped Fiscal was growth. revenue OEM both electric $XXX

Hyperion autonomous to the At Foxconn announced partnership with growing use Foxconn partnership vehicles. . and and global electric for NVIDIA CES, scale demand its Drive based we X compute control provide Foxconn for a be develop a This will Orin platforms. vehicle volume, will for manufacturing strategic to producing sensor meet NVIDIA architecture units the electronic for Drive, on platform. Tier automated manufacturer Drive will NVIDIA

reached important certification functional for the experienced operating our one milestone higher industry. bodies safety, most performance industry-leading safety transportation. assessment SÜD, NVIDIA from platform required system this quarter. meets also standards We in Drive and autonomous the and of an rigorous With received automotive the The TÜV

expenses P&L. infrastructure the XX%, of up were rest gross data primarily margin non-GAAP was margin year margin non-GAAP XX.X%, GAAP higher gross the GAAP operating the gross margin compensation expenses XX.X%, to Fiscal gross were XX.X%. and QX and Year-on-year, XX%, was and GAAP XX.X%. was operating up center Moving to due non-GAAP expenses. was and

at GAAP non-GAAP Sequentially, them were quarters. plan down non-GAAP level GAAP coming up the XX%. were expenses We operating keep flat over and to this flat, Full operating were expenses expenses XX%, year up expenses X%. operating relatively and operating were

our under through cash shareholders QX, returned and dividends. approximately the repurchase share billion December had form in of billion the end repurchases We to of share %X At we XXXX. authorization remaining $X.XX

the Let me first the quarter fiscal of outlook look for to 'XX.

center Non-GAAP $X.XX We and be to and X%. growth led and plus minus approximately be billion, by driven points. expect expected minus operating to to expected our XX.X% respectively, to are and of CFO Revenue plus commentary $X.XX billion. and or and be and discrete any are available billion. to income of expected by full plus GAAP on be gross approximately be are divestments. X the losses XX.X%, are excluding non-GAAP major rates are margins XX GAAP to fiscal or market minus million for quarter expected to billion Capital $XXX expenses in $XXX is data expenses year to the website. are items. details are expenses $X.X million, basis non-GAAP expenditures GAAP approximately an non-GAAP million expected income operating of other financial be $XX for GAAP in to X%, first each XX%, expected the Further in nonaffiliated strong XXXX. be and sequential to range and billion or excluding IR the gains be gaming. approximately of platforms growth $X.X our other information expected included $X.X tax

community. me for the In highlight upcoming events financial let closing,

Boston. in in will Francisco Morgan We on be the March and X attending San Cowen Technology Conference Healthcare X Conference the March on Stanley

XX. Jensen's GTC We virtually March will also on keynote kicking off host with

of the May for year earnings call discuss to first of is 'XX our XX. scheduled Our quarter fiscal results Wednesday,

the will we please questions? would Operator, up call for poll you questions. open for Now

Instructions]. [Operator

the of Your first Aaron Rakers Wells question comes from line Fargo. with

this focal cloud on is strategy. going to software key Clearly, your the effect a of and point call, monetization be

we socket. I suite, software AI think think, at $X,XXX as up, CPU the I look enterprise priced at think, around per I straight is it,

got can model. past, to over think million couple couple bit. in if little think, to the do company's a to, quarters you've talked unpack I a hundred that how metrics so? next pricing like start Just that contribution little of the higher curious just model relative a monetization Colette, the consumption you for or of curious, you've cloud bit think I the business we about I'm

Jensen So to I'll be talk will more to start and believe this great topic and it because also discussion over turn our I a at GTC.

more of in continue our good software. software, we're growth onboarding working in increasing our even plans partners our progress see partners, with both and in we results, making terms to Our QX quite

probably of software in being to the Jensen. terms in it level our QX our software are day of terms there's in about You correct. and a And there, each We've more as revenues unpack turn going getting was of stronger talked we're hundreds millions. I'm even But to levels. record to

It learning, all, essentially today. Yes, of validations, back, AI a operating systems AI first starts step system to to the to data of NVIDIA is training, inference. from taking processing

every And it's accelerated this body so runs accelerated. in we supports every of model of, It is that And software It it completely know on-prem. runs every and framework, cloud. everywhere.

is your learning AI, is By efficient, cost using operations more NVIDIA it effective. more entire and machine

You save using by accelerated software. money

in way the adoption believe enterprises. of It hosting extensive the from enterprises software. body AI all within announcement but the It by utilize models, Our learning accelerates have the today through for infrastructure leading machine pretrained in the this the NVIDIA's everything generative providers that of is accelerate enterprise. and very deployed NVIDIA It we of infrastructure people's not is cloud, pipeline, world's the putting through a the of service ability faint not is to it be adoption heart. hosted AI enterprise's from in of can system broadly, we software, cloud operating which accelerates

this And believe new business about will model. that really our part of we're so of the extended accelerate excited software. it adoption We

comes line question of Vivek with next America. the Arya Bank Your of from

Just wanted meant in year-on-year data clarify, on Colette, to center grow basis if could also you a QX?

a that high, it hyperscalers? accelerator there gets The the in does for Jensen, are of kind very question if to the more other other is the generative my And market? extreme, ASICs then market AI, or from then it large, NVIDIA X just cloud out very on of the small ones. attract if it competition doesn't handful options to main for market intensity related computing size And relate limit

question. the for Thanks

QX. that guidance of in a past year-over-year for growth data First, talking center. our data terms expect expect data And provided we do probably for sequential We center QX. growth sequential center growth. We a year. our accelerated great year-over-year strong a expecting our growth center, actually we about also our data are in

because they large models quite are language large. Large called are

new By that days, XXXx over million Law, large systems, processing with faster. researchers entire language times million we've delivered million a up new decade. new frameworks a in coming accelerated new x with would span, best and interconnects, by a AI the that scientists, a algorithms model faster, last Moore's data working on decade. processors, However, advanced made times have in processing models, across and new AI remember its and we've

which, infrastructure, a What couple now would a engine, its of InfiniBand models. large new And in large Hopper, even about the with processing beginning, leap it we're large another second transformer course, per rates, XX data have gigabits taken introducing to of still NVLink in And switches months need the happens the XXX able you new in days. its take of we're language it's and infrastructure.

think training by this NVIDIA's going DGX the quite technology the and to this make And -- NVIDIA accessible. access democratize and capabilities, of DGX so the supercomputers this putting we're with I infrastructure, cloud, cloud into capability accelerated really with

So thought. that's X

The is quite its of really models fields, we different are developed language their and domains, is Different whether knowledge of prior other foundation countries need of of large. the to that imaging language domain its that of quite be all models. development them With cultures the biology be own models second accelerate used now Different or exciting. could large body have different foundation fields, have each it's to physics, or large number different. with or course, models, which a is of these one its

the remember models be And data first world also train data. will world the body And company. to number that have It company. AI inside their proprietary. is valuable harnessed the of of The leave new the thing It's to time. are their the own belong in the very will other in they proprietary data for to companies The never most that company.

to in our like is the the make who so proprietary available strategy we would DGX create proprietary capability in world goal so to cloud -- to that so and we can enterprise, company models. -- And every infrastructure put the this every and data our

second The We've long thing a for competition competition. time. had about

be the one, dimensions. computing you types. private you AI can out their use inference, you is develop and supports for every approach, X everywhere. of It's cloud all use training, as all different several to meaning for could know, Number the models It framework. it to It architecture, can allows edge. you It on could it it Our for models it supports different cloud. our architecture universal, to quite use to it on-prem. autonomous cloud every developers It's deploy system. It's cloud, way This an everywhere. is

interesting. The second is or very pre it of sexy maybe application it no to part part don't people preprocessing talk an large because service. as idea AI and post-processing the about processing not post and application. it as itself a is into a Most that in an There's and not of turn it's

However, out consumes that the it half workload. of and preprocessing turns oftentimes overall or post-processing X/X

by of data the end-to-end pipeline. the accelerate versus Whereas so The from the about accelerate accelerating data accelerate but because end, everything the way that to if accelerating could to workload, for all accelerating end is just why when And as passed the And workload, routinely you the to accelerate maybe limit it, pipeline to which way hear processing, twice things processing, pipeline, part entire end. hear up. XX, you're of not preprocessing accelerate applications, to of CUDA half end learning accelerate XXx, XXx half to even the from reason only the if is reason all to XXx speed you deep speed using just is instantly able we NVIDIA you entire you fast. faster, we're ingestion, up, entire workload the the if the XX, you XXx, post from preprocessing, the

differentiated so accelerated accessible operating in the most our we're And universality edge, quite of the that the way. And good every because really all high, our think to importantly, cloud, computing the of to -- comparison in is get architecture fact it this from close. computing cost. the and platform, is can is architecture throughput, our cloud lowest the a end-to-end because I to so that fact very utilization we're makes It's not service even the such providers, not workload accelerate the -- you and use

those yes, and answers. -- X the are So

Your Muse next question line comes of from Evercore. with C.J. the

as iPhone. the I inflection you about kind an guess, ChatGPT Jensen, like point of talked

evolved you growth with And then for as have And post part large-scale and with of the those few with have enterprises? about kind outlook so and in product curious, X your last the A, hyperscale Hopper evolved how you cycles ChatGPT memory, high-bandwidth conversations think secondly, for Grace transformative months? your engine how

And piece stuck the led that and work, job the wonderful a and of of to inside job, capability. of with with its accumulation versatility a model great team service surprised They is did all did ChatGPT the OpenAI a it. it. everybody a with a its great with breakthroughs

for model industry that people course, not of speak not a can can program. can language, to prompted this understood. Python, our surprised remember, And it never very even speak English, single Cobalt, human it a for output output a in be by, Blender, But well the is this skills were just translate, trained and model just perform few output human and in XD tasks language to but was people close -- Python language, What AI and is but or do. of the language capability within that surprising that

writes for another a program a it's program. that program So

is, programming translate We it to request whether it for now and I everyone, that explain computer perfectly take performed. when This computer -- particular say decide that not. maybe human or prompt of a be you realizes process this whatever process to instructions your you we've new a and could almost it want platform, language human to anyone new -- it sequence directly, now of for realize your era that world a language, new to computer this computing, you computing is, democratized programming or is who could the language task whatever waits human-explained computer new in good

because that of is to computer software in type just world. people provider, every democratized all excited frankly, they're, because over what has CSP, is an is and reason, actively I service. or service that programming this the Every either or their utterly that write many any develops company, so really model integrated application application so this revolutionary alerted Because on their can single working enterprises ChatGPT of everybody of who shocked explained, every single that like it's to Internet be for alert something a its integrated And into program. is single AI program software into into or

is, this you can so as And imagine, utterly worldwide.

The activity around gone build and inference the XX activity the to Hopper Ampere language we has last roof AI infrastructure the and the large that models, using through Hopper days. inferencing around just in

there's the XX last of enter has XX, no as views dramatically fairly, a changed so of days. this year that the whatever as our been we And result are year question

question next Cowen & Your Company. with the Ramsay line from comes Matt of

more going you could what hyperscalers. I and top the Jensen, couple product more business, the questions that next services years, host the and talking to all wondering of [indiscernible] But been compute kind X on or if the Hopper services business business that with a on segment selling transition Grace that about from for back, systems I data you're to Cloud. center on drivers a wonder selling evolves when systems product, and think of center think maybe mix a your And of wanted I margins over ask level. systems we've the the in your to guess, mean the time? step InfiniBand your cards software? different And we're might X of and over with these things how you pairing to do the can business DGX data and of watching I level meaning your of with of

Yes, the appreciate question. I

switches First sell GPU the as with in a we Hoppers a of NVLink. the a service NVLink large fairly connected because providers computing or business what context Center is are of cloud with is conceptual to GPU all, know, a panel, Data our connected panel Amperes X actually X that's of that you business only

so essentially high-speed chips X board interconnect. with And X represents this into It's connected GPU. X very together a GPU chip-to-chip

been we've GPU. for if quite is so that some you And And working will, time. X multi-die on, computers

an cloud we infrastructure So when and going thing have We're because GPUs. actually we to excited about so about the many service the for by hosting continue And think think about GPU, to HGX NVIDIA is providers do we to that. our GPU, are that's really that companies oftentimes start-ups and the on-prem with a offer We're directly would at into relationships those X be with really directly to And working the work least with. enterprises both world, every that cloud or love XX,XXX the we AI in to of today cloud deploy into multi-cloud. around industry. all able and

that open is the And AI effectively for for And model salespeople customers going world's by This very, in the team AI AI full software, the us from arms. best have entity, And having are to to have operating entire infrastructure way all a NVIDIA be to now clouds. DGX models. the an the infrastructure stack extremely the system, we're infrastructure and instantaneous to They they NVIDIA across are acceleration up they have very we're open X with CSPs. NVIDIA's so advanced. good of people a And expertise is Within who their the cloud, access span. customers, attracting most to exciting for that the them. they welcomed the

CSPs. model this like for for wind. And customers. it's great a the great great really run It lets a It's is us. us model a And for model so

how AI build no as it. time we hard much matter supercomputers you at look to take It's how DGX continue supercomputers, look it will does matter advance no takes AI to on-prem. continue at and time you As it. It

that and and possible. a so of prefetch to have now And lot running we customers as as the really ability fast up get

comes Your line Arcuri question next Timothy the UBS. with from of

there now they've contracted to question TAM. right who's to the end, on there. doing model. obviously, least but I of and of and does all seems on has lot more XX,XXX in that Most a are about had focus working maybe companies what like at Jensen, on training these on training a deployed thousands is it big maybe, of the models a widely this They're of your video text, tens GPUs high inference cloud And models music. somebody the

seems So TAM I'm billions easily last of easily the is tens does the But thousands in TAM you numbers incremental hundred the year. of this wondering several kind like what in to of gave it GPUs the dollars. of and

and think $XXX you said TAM TAM. $XXX software I billion hardware billion

how you of what So the think be? new kind TAM do about would

exact is part easy was of X million it on so develop weren't generative at app There's of difference the and big mainframe all This so computer you TAM probably good program mobile I end the mobile App point the numbers those There because of it towards device year, of the accessible. distribute with cloud. a to easier going because applications. the, million same the with This the than will, X converging workstation applications. way last The very every people million because to inflection revolution. cloud, to And deploy if the and this that is is easier There happened Remember, cloud, PC revolution. no happening and are with X single think thing stores, of now that really incredible we're because Internet counting AI part and and the that amazing million breakthroughs computers There mobile applications question versatility emerged. more Store, still. middle easy and iPhone happened of anchor the change, industry. develop arrive happened happened This in because the weren't AI. for the platform weren't sooner and happened Every to that PC X later. to and applications. was moment use, capabilities to applications

In quite an I XX, did to create million computing no is days. of are XX computing kinds thing. ChatGPT, platform, things. XXX people mean, people era reached X this in using extraordinary And all it

emerging. And so in just torrent that new companies is a I a you're are new think era. way, every seeing no new of that what There's now this is, question computing applications and

so this than more and And is today even the expressed, sooner really -- explained think we TAM realizable that before. it and I

Bernstein. of with Stacy next comes from Your line question Rasgon the

both overall question shipments? at AXXX. X of that clarification, like same a that have I Or statement? the in you point The an said Was for after revenue's and then Colette. H-XXX a time was higher clarification than quarters

that you of were order are decade. about ask single then guys of an their today, over kind euros actual by Is and the specifically software The XX-XX. that low that were software that Mercedes And you how supposedly and talking had thinking frame? that the I of wanted mid-decade question. and the Mercedes billion of by about similar time billion the my they revenues And end that? magnitude software Drive auto, mid to event guys euros revenues MB could should of for or from be be for Mercedes revenues splitting the deal Is know I the modeling opportunity. digit we

Stacy, Thanks, the for Great. question.

And process this with in me your many quarters about a for question us get and H-XXX was start. ago. a in to of back H-XXX production It them began Let first We AXXX. of start them to was that initial level QX. QX. you began had time shipments Many great

great QX, of important the to many So QX up and running ramp H-XXX What to saw. our a they were get within that we in both was CSPs see that means H-XXX our of for instances. was of all cloud wanting get time an is focus and us

AXXX saw for quarter forward, H-XXX in strong sell amount. in at QX, what in QX And so a a we less saw to we larger continue architectures both just of it of tend to going but We actually was

them their that Mercedes talk Your what diligently XX about the We've the with us. to opportunity They in software see about X can they about a work. They do also looking at the extended about can as And of Benz. Connect. did out Drive you right. probably years very have had on phases, the connection additional of you're as they that to working market. talked do with position ready They with with software getting what questions opportunity. very joint opportunity front come been to they in we about pleased well and that I'm

our a long-term what thoughts aligns partner that sharing over it that are So of time. with with revenue and

-- every future One luxury sensor and that only that install single has, to about upgradable large that be of supercomputer, with forward. every base installed the is will of from going them one I This so of luxury for them customers Mercedes of to of a one AI end what something set, entry rich fleet if is Mercedes the their things the single cars, add, an every forever the with the could all car to in renewed every across the the contribute from an brand highest say doing. to board, the Stacy, could wisdom that, way

Mercedeses can fleet just represent could today has. And benefits of think, entire extraordinary. millions on you benefits. represent to continue and they're with is also looks that's for revenue-generating tens you programmable, that of completely if building And Mercedes I for what it of large road will the were imagine for, that Ola luxury The If the that would that installed opportunity. revenue-generating it's to that it be like going the what OTA, would base renew customers' -- vision cars

the Lipacis Jefferies. line Your next from of comes question Mark with

AI I think generative a we facial for you then Omniverse processing, like think drives Is engines, language products? expect your next view? us seems engines, comes now your can forward, with it natural process I workload ecosystem and what level for if workload out brand-new should new like share demand the back this or to going drives that every And cycles. a demand for and and recognition you, recommendation Jensen, year your your

a generative the than be products comments is is AI for in your over your ask I the found view reason drive lot, your script thought of just the last I demand you -- your that where what XX because and you And going better to days. is mentioned interesting seems about now it services to kind that

to and you're what you little can drive extent seeing next past today? think us workloads applications drive with it of So what the of -- to that share you're that would demand, levels care you bit or a could there's working on new that new

Yes, Mark, I really appreciate the question.

First of and that all, new you I that have new at applications would we've I share never that you like about with don't GTC. to shared know workloads

And talk about. think by surprised and to delighted going so applications that's to and you're be the we're very to to GTC, hook going quite my I that come

for a not there's it why way. accelerated a is completely a with do multi-domain reason NVIDIA that We CPU is that that fluids. you're It that about which accelerated platform. computing CPU number the functions kinds accelerate, flexible. mathematics, multi-domain. hearing a the There constantly XX% applications. we offloads can We're But only offload because not control an systems. is is, graphics. Now CPU computing of like the and highly could X% makes that case is XX%, The do paralyzed really are one. all completely neurons. number reason We're it one, new We heavy works general platform different do can highly, them. can do could we're do could be that particle things And we computing units, to could computer applications purpose We that . We

that two, seek reach. universal one. like a only from to compatible Number special and Literally, it reach means out many like the developed that to reason on-prem. the the They accelerated workstations, to, they why are only tools the And one attracting and to single so just because They after This single They of Every They we can experts using develop gamers the ecosystem the is them. millions architecturally see reach platform. the is programming PCs then have people like like acceleration, the reach. fact our very you base ease they large. across it all is help cloud computer accelerated, something the like And they the applications. users out who -- the is developer compatible, can that would tool they of around our only like they the world, to installed number so keep reach. CUDA software new fact they platform cars computing There all which every many importantly, platform, is can the accelerated. that so architecturally the software. of

years. to At to at finally, applications advance XXx keep has the of have that CPU years at continued really, That that my sustainable. bring every and still career, this when they people then And do important whereas up. really afford be in rate in slowed every the would very the has to with rate point. of performance the then years computing is first slowed. and a XX X world, the time to needs of like a And same urging going they power Remember, XXx X about Everybody advance can't back power tremendously.

can power power. consume you it, amount workload. to we continue of any can't for accelerating the You use By decrease

use And so driving all applications. we and accelerated really of these is to multitude people keep new reasons discovering of exciting computing,

the of Your next question with from Atif comes line Malik Citi.

center. data a Colette, I have question on

to You the saw plan year. the in but year-over-year build January and acceleration some guiding weakness through April on in quarter, you're

you So based or your generative AI coming And if or also, the can the you us if on acceleration. vertical. seeing the AI you're enterprise Is confidence can just sales model? H-XXX that on for ramp in services rank order through talk the new what about

growth, accelerate do forward. going the It as going growth well. QX for When in grow we're yes, likely in Thanks and will there think about year-over-year to expect our question. Sure. we sequentially QX

have see market. we we of Yes, as coming do that? what drivers So to cycles the product multiple

H-XXX in now. We market have

are GPU the as in our new our half that are networking. We with continuing likely And computing second we launches coming of have with our sometimes the well then fueled with year. grades

it's AI, those CSPs, enterprises, start-ups. those our generative be be Additionally, one among interest whether those customers, definitely be sparked of whether

focusing forget an here given the this be on not We there's that just then end on AI, part lastly, let's our error revenue a And that accelerated growth focusing Moore's of year. of expect to continuing. of Law,

going economy very turn enterprises is to additional the the as things to for to first any So I'm it Jensen and of existence fueled he'd to add. this cloud as the probably to enterprises has [indiscernible]. improves, by like it if the be important they can see

did you great. great. No, That was

with Morgan last today the of Stanley. comes line Your Joseph question from Moore

to next few give the Jensen, that what of over much the improvement the to what about you scale language looks in into that models years million these ability some that But performance. Can that road? talked is better some going expense of of much XXXx frame. in decade. models are to to the large of these your up? do and I over times know just customers about it's that the sort Hopper And you time like larger train a talking be you the complexity kind that down how Xx can extent us transformer with insight your X hardware last reflects

all the cranks when in, form the X backwards. future, First, AI a comes is to data building a of model. of be improved going Where it, exactly there X for on I'll thing intelligence. I refined only. Data intelligence, physical great in. or that the data out will and time factory. does called to was be that's production over And It updated goods. value, -- cranks an that that number and will produces manufacturing. the start people comes grow thing be of AI, the manufacture almost infrastructure new That world. raw in to material In of or every something and goods. going center reason are comes is manufacture a it just happens infrastructures There is believe the just the company It soft on

so expect the And factories AI to all I see over world.

Some be hosted in of it will cloud.

large, be be will are and will on-prem. there Some then some that be of some that There and smaller. are mega are will that some be it large, there'll

that And happen, fully so expect number to one. I

Number two.

why company us trying AI accelerate up chip that's the of And a ways NVIDIA XX going new interconnects, one just of to to a is I x. reasons we're new through developers is new systems, the to course new and of and AI hope do the models, with Over computing lot with because algorithms just operating we're million too distributed next years, working systems, for coming I believe chips, new new new the by There's algorithms that. problem not another complex. solve

entire because software. And You network stack through X innovate great of center the across that that can across for way a the entire all that mind the company, stack. across data we And to stack. the way in we entire have across playground into it's scientists quite the the all single the can computer chip, think reason think for the from really

decade. So really going coming this, is breakthroughs you're growth my of simultaneously, next the everywhere. these that the you're to company, incredible AI going in see because AI in of in But AI the gigantic models and platforms the adoption see to expectation factories

now back concludes Huang session. turn to Jensen I remarks. Q&A will This over the closing our call for

capability has Thank generative you. AI and and model large new versatility language computing The breakthroughs from emerged. has elevated remarkable the accumulation of of level. platform to transformers, AI a A

and The their businesses. we've every AI, solutions challenges and ready. New AI over been we generative working And industry for decade. level already rate. This products towards of are is activity just new Enterprises in about around accelerated moment activating a apply being to companies, significantly. which applications new are astounding the long-standing are was to has an invented high, to reimagine at

transformer fabric full Our InfiniBand production, and Hopper are Hopper in CSPs the and Quantum racing cloud to with services. is new open their engine AI supercomputer

to look work to we we meet the growth accelerating forward through our strong year. for As the demand GPUs,

the GTC. miss Don't upcoming

applications there. new software, you more much partners lot and to and have systems This our See CUDA best on customers, new yet. AI you will ecosystem tell and be about a new We chips, NVIDIA and GTC Omniverse.

concludes conference. This today's

may You disconnect. now