NVIDIA (NVDA) 23 Aug 232024 Q2 Earnings call transcript

Good afternoon. My name is David, and I'll be your conference operator today. At this time, I'd like to welcome everyone to NVIDIA's Second Quarter Earnings Call. Today's conference is being recorded. All lines have been placed on mute to prevent any background noise. there a remarks, session. speakers’ the will question-and-answer be After [Operator Instructions]

Jankowski, Thank your you conference. Simona begin may you.

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

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

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

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

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

the let to call that, Colette. me over And turn with

outlook revenue billion exceptional $XX sequentially We of quarter. up our billion. XXX% year-on-year, Record Thanks, XX% an and had of above $XX.XX QX and up was Simona.

accelerating the of sequentially language demand year-on-year, up Center Let Record by start of providers and service and driven nearly compute from Internet revenue tripled me XXX% companies XXX% $XX.XX models. billion engine Data Data with generative up HGX primarily large and consumer platform, AI revenue first was Center. cloud year-on-year. for large

Major Cloud networking as growing Hopper driven as gold in based Networking the for architecture year-on-year, cloud on Tensor companies, and GPU providers InfiniBand number including Microsoft standard platform, end-to-end HGX Azure volume, AI. a revenue deploying, Core and almost are Google our Meta, AWS, Oracle systems by Cloud, well Ampere of our GPUs. doubled

platforms. There is demand accelerated tremendous computing AI for NVIDIA and

ramping to supply partners exceptional in needs. our Our have support capacity been

complex up center including data highly over Our been has XX,XXX and chain, the built supply past parts networking decade. HGX with

and the qualified have suppliers for We as also developed key additional [indiscernible] such process packaging. in capacity and manufacturing steps

and increase By in to strongest quarter our the computing. AI was geography, their the including through within center capital growth accelerated direct investments supply each expect Data to U.S. XX% demand Center data China networking was solutions. range historical next of of customers XX% compute as and We year. to revenue,

is potential achieving on take At a to China. regulation results. our the believe increased to let We recent exports for the the me reports on this time, regulations current address intended moment

Given the immediate an export would anticipate impact we products for on our strength restrictions material not worldwide, to Data adopted, have GPUs, of our financial do demand if Center that our results. additional

and prohibiting to and will result restrictions of term, in Center industry a compete U.S. for implemented, long GPUs in Data world's the markets. the over opportunity to if one However, lead of loss China, the sale the largest our permanent

The of they networking, stack of computing the is silicon, Our and the algorithms. generational quarter, to providers across full HGX decades HGX upgrade new for nearly data undertake interconnects, as systems platform accelerated era service in drove NVIDIA systems, demand their culminating exceptional for cloud innovation center AI. a transition strong software infrastructure two and of

generally very AWS, way now demand. are drove GPUs on the Microsoft GPU Core also Consumer generating providers, and center shortly. Tensor several Azure the are AI already purpose-built companies HXXX NVIDIA returns. available strong significant cloud others Internet the investments at Their by powered for with infrastructure data Instances in

recently driven increase a have since Meta, that than launching spent Reels, highlighted time Instagram. AI example, in For recommendations XX% on more

NVIDIA Enterprises are consumption serve cloud Whether deploy driving pre-trained access for to models also software instances to their can powered or generative of libraries, or AI strong applications in in direct, demand on-premise partners as well with infrastructure. and seamlessly the through AI, the cloud run on-prem acceleration racing our we NVIDIA as on enterprise customers APIs.

customized the announced Snowflake all create generative enterprises path with Data AI a Snowflake their accelerated partnership provide proprietary using Cloud. to applications to own data, We with within securely

NVIDIA the make Data enterprises custom models, services, LLMs right chatbot, able will Cloud. language from for Snowflake to platform search AI advanced With and be NeMo the for summarization, developing including large

Virtually, every industry AI. can benefit from generative

sales, of can AI Microsoft by to AI as support customer leverage For services, field. software dollar in set legal create multi-hundred systems and are of trained new their example, will in Billions tens market Copilot education the and available office AI workers just to such billion boost millions for over announced X opportunities those engineers. and assistants customers. of of be professionals Copilot billion our productivity

can Getty visual images such tools to media trained a and into some from a AI XD partners to the models artists integrate the foundry organization, services content NVIDIA from prompts largest content for NVIDIA responsibly generative using generative NVIDIA and communication generative WPP, create generative content Adobe developing and creation. Picasso, engine while marketing custom as is seeing Images for Omniverse text are WPP designers design. designers AI using and in AI earliest We enable entertainment. applications of and AI of world's marketing,

consulting and the deployment Lighthouse AI tracking ServiceNow accelerated Shutterstock We've XD the is services Accenture to and and scene of Accenture enables Picasso also and enterprise with AI NVIDIA NVIDIA generative build tools that provider users development unites capabilities. using to help create with the platform partnered content launch enterprise AI with of Lighthouse and to ServiceNow Visual the program, fast background with services. AI. automation engine computing

We a enterprises for are AI powered simplify also Face to just Hugging tune NVIDIA NVIDIA VMware Face offering for And integrated with train for to and platform of service models models and will Hugging yesterday, AI new new and by called the enterprises. running VMware. cloud. advanced and HGX AI computing fully from enterprises new creation collaborating major AI enterprise featuring custom a accelerated Private with with software software announced VMware NVIDIA, Foundation multi-cloud NVIDIA offer a

intelligent to VMware's the will summarization. management AI and software hundreds to and of generative as have assistants, of search run customize enterprise access thousands and such needed infrastructure, AI chatbot, applications cloud models customers

GPU the data center standard center new ecosystem the LXXS new also industry NVIDIA and We built announced for servers AI enterprise-ready DPU server data BlueField-X featuring processor. NVIDIA infrastructure

makers limited industrial (ph). applications, is is is AI processor and not training [indiscernible] a server most for the NVIDIA leading accelerate video including digitalization. and LXXS processing by designed world's designing, the data high the center to data inventing visualization, compute-intensive to LXXS through supply volume center out shipping universal Omniverse standing system and

world's companies, and ready ARM-based production fully AI Swiss including National Atmos this Nearly Foundation Lenovo. supercomputing and be Grace the IT will VMware, our The It and AI enterprise Superchip also and from XXX servers including be National Center. the Cloud OEM are AI servers. configurations with for shipping to Foundation. Private optimized in available soon Dell, available enterprise (ph), Labs NVIDIA entered quarter servers customers, HP NVIDIA CPU Computing computing leading enterprise GHXXX will of Hopper is which Grace Hopper combines GPU full multiple ready

SoftBank and XG/XG and a NVIDIA GHXXX applications. And for are collaborating based platform on on generative AI

DGX system, our HBMXe analytics. the the the NVIDIA our version a eight Superchips GHXXX XXX all of calendar to end for its switch over AI Hopper by one, year, Meta be generation of the [indiscernible]. are prior latest just model, among use a DGX connecting with gain GHXXX, the of second Grace as Google supercomputer compared memory data together announced Cloud, Grace class Superchip language and first to in huge GPUs is the large access. of of giant available QX We work AI enabling This generation and new first systems Microsoft to systems expected the The jump be available new of to Hopper [indiscernible] will recommendator memory XXXX.

to to than Thanks the and Strong Ethernet InfiniBand more double for InfiniBand GPU capabilities, HGX networking optimization delivers growth in-network AI. infrastructure traditional computing was of driven systems. primarily performance connect end-to-end its by

worth For is [indiscernible] billions network. of of dollar pays AI the for throughput hundreds infrastructures, the InfiniBand of from the value increased and

GPUs. leading only addition, choice of In InfiniBand can to of It hundreds thousands of is network AI practitioners. scale the for

qualification NVIDIA optimize AI AI their in multiple centers Ethernet-based announced traditional designed consumer is It AI we with platform an couples DPU, performance, Ethernet Ethernet cloud performance ramping switch for to For with BlueField-X and accelerated a DPU major OEMs or companies. X.Xx Ethernet. BlueField-X CSPs Spectrum-X, and the the Spectrum versus success. is optimize achieving workloads. networking overall seek efficiency major and Spectrum-X Internet that better to across data power

Now seasonality. moving end GeForce up revenue believe desktop. year-on-year. last GPUs customer demand End year's to sequentially fueled of XX Gaming $X.XX billion solid laptops We by growth was consistent gaming. and RTX for XX% demand after and slowdown. was Growth XX% to Series and was with has global returned

large a upgrade opportunity of us. ahead have We

installed or GPU the our an about Just performance. with have RTX XX% base XXXX to of XX% upgraded of RTX higher and

several shipments have This key for XXXX posted by year, outpacing quarters build season, of back-to-school world. schedules with laptops of Laptop GPU-powered the is as growth to and the and GPUs the reality bit, and popularity, QX from GPUs now led regions QX our around overall reflecting the are the their RTX desktop stronger a revenue back-to-school shift NVIDIA's in likely gaming strong GPUs. in laptops. holiday gained the

the support, low we The TI blockbusters XXXX and the IV DLSS X. $XXX. down and XX RTX Ada price XXXX Baldur’s Gate bringing launched DLSS including to such GeForce and points as of architecture games the continue Lovelace desktop, expand. to Diablo as new RTX added RTX to In games ecosystem GPUs, as GeForce

XXX over and games RTX apps. now accelerated There's

At can a characters. it model to We of NVIDIA AI COMPUTEX, use this AudioXFace. or are harnesses gaming. ACE Riva to Cloud technologies, foundry AI intelligence Developers service we custom NeMo, service. and a bringing and Engine games, number Omniverse to generative announced NVIDIA for AI Avatar non-player And including bring

can of moving powerful NVIDIA AI teraflops NVIDIA XX% include QX, will refresh GPU memory. architecture performance They ramp with in gigabytes desktop Now Revenue drove NVIDIA Ada with $XXX down up than inside. year-on-year. more Omniverse sequentially a AI to laptop systems up XXX RTX workstations in of new with QX. of and These growth million The or providing Professional workstations X XXXX to GPUs, configured coming initially X,XXX and XX% was enterprise be of RTX rolling out strong Visualization. in

Xx to announced and Xx performance RTX up NVIDIA up the three The the desktop faster Ada generation. generation. core compared training XXXX also the based AI to XXXX, previous on to offering new throughput We RT workstation GPUs XXXX, and

visualization to expanding data content In opportunity design large AI, traditional development language in science our technology. generative XD workloads workloads the such in are pro addition for as and creation, and new RTX model

One Omniverse Omniverse quickly is [indiscernible]. much XD the of interchange is NVIDIA this Apple, AI. platform. of the that keynote like for month and HTML is [indiscernible] in earlier for Pixar is key themes world, OpenUSD's the universal becoming XD form for the This OpenUSD. Autodesk, graphics standard was and a the Together, the where Alliance conversion Adobe, is the NVIDIA OpenUSD universal language Jensen's positioned. is native

OpenUSD accelerate to to including workload. RunUSD new generative We and mission adoption. ChatUSD to bring and cloud upcoming APIs, is Our and Omniverse AI announced OpenUSD's development

down Moving to XX% was and XX% $XXX million, year-on-year. sequentially automotive. Revenue up

We a growth MediaTek with integrate associated The demand, ramp GPU lower of the bring with overall experiences will partnership partnership Solid the new SoCs automotive on passengers vehicle to particularly a a energy to The a and wide platforms year-on-year level. or entry decline based new [indiscernible] announced driven makers. number self-driving sequential China. of reflects line drivers of from vehicle automotive in was chiplet. of luxury segments covers car. range and MediaTek product inside new develop NVIDIA's by

and the driven to XX.X% non-GAAP data margin expanded sales. higher gross P&L. margins rest GAAP to to Moving of XX.X%, center gross by the

repurchases and amount significant form cash expenses operating up GAAP returned is a benefits. drive expenses margin. software $X.X up We share Our Sequential also complexity, reflecting primarily Data which our were X%, shareholders gross Center of approximately the in and non-GAAP include billion and of compensation dividends. products helping increased to and operating X% were

as an $XX Our approved to our QX. stock Board of just additional repurchases authorization $X Directors of to end billion of add has remaining the in billion of

Let for Data of the the across quarter where third me is outlook for on tremendous turn customers. XXXX. Center fiscal to industries our and platform broad-based AI Demand

year. next into extends visibility demand Our

we as to partners cycle with the supply to continue capacity. lower will supply Our several over next our ramp times quarters and work add

help types for enterprise. the growing workloads the demand will cloud many address Additionally, to from of GPU new LXXS

expected $XX billion, For revenue to X%. minus total is plus QX, be or

Center GAAP respectively. growth sequential expenses billion driven expected XX billion, be gaming basis to are with expect and XX.X% approximately GAAP to minus $X ProViz We to points. operating respectively, by gross non-GAAP XX.X%, Data and and be and largely margins expected also be are plus contributing. and $X.XX non-GAAP or

and the of included an $XXX to or excluding on available any gains our in Further are rates income to items. GAAP GAAP expected X%, are be excluding commentary are information discrete non-GAAP non-GAAP income details other other from IR expenses minus and and investments. million, and and plus expected tax losses non-affiliated be XX.X%, website. CFO approximately financial

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

X in Chicago, September X Goldman the on X, San Virtual Citi conference the Francisco, as both Conference as Tech on Sachs September Evercore well We in Jefferies Semiconductor the XX And on Tech the will attend September Conference Conference August on York. on XX. New in BofA Summit September the AI

Our the November Tuesday, for XX. of our discuss is of fiscal earnings XXXX to third scheduled results quarter call

we Could for questions for call for will you poll Operator, us? Thank please open the you. questions. now

from Thank TD first [Operator We'll Instructions] question you. take our Ramsay Matt with Cowen.

is open. line now Your

ask afternoon. wanted regarding Yes. the model of a of Thank remarkable emerging Jensen, you to really Good Obviously, I much. application inference. you very results. large quickly question

lot the the the So smaller done have workloads on I it's market. a A -- that well pretty been of you share investors much majority the in understood smaller ASICs inference of training by model market guys lockdown think CPUs or have very of past.

how And Superchip many others accelerating inference segment well could really pretty inference. Thanks. GPT workload these large this positioned and super-duper product on that? that's But small us market talk between large Hopper of how products inference I Grace that. model with and other you're your for and about the and portfolio large models, model is there's think aligned model quickly And maybe to new you are seeing inference your for

lot. Yeah. Thanks a

unstructured it core, versions learned or large within So let's you large course. that compressed to These are you are it these of learned take the language several models is, studied. has it the distillation. It you a knowledge as it by ability structure then of process the models. happens It's pretty And amount has What derive a large understand has step language. and the within human language It it called create things, human create has it fairly large model, is models essentially phenomenal. -- But smaller does you its quick what back. corpuses encoded teacher-student of of language. -- that it a as can, and from at

and just distilled students forth. going so to likely this see these were And models, larger the see on from so have it's you learned they forward. case you're so very teachers models, from when derived or smaller that and and going and as or from professors you And

a you what's capability. and a called very has of generality start zero-shot generalization large model and from and And large it so amount

haven't And that so or lot specifically them. for models to the these it That's of questions a has it skills what magical. trained large makes and so applications language capability on, miraculously you perform

have is do good like These what the in capabilities called particular don't devices, skill, is On as computing excellent zero-shot have They models they these would generalize you capabilities smaller capabilities. to a you all you of kinds and down. them what well. other but so have as don't hand, on distill might

they own models. you all large from their have but capabilities, start so And unique very

Okay.

Vivek BofA to Next, Arya Securities. we'll go with

is now line open. Your

Thank you.

please if come you quick could XX%? you and supply said incremental every do it's much to how in how So think XX%, clarify expect of sense supply Just XX%, question. a year? much Colette, You you online next XX%, clarification quarter. the any it's a had growing up because just

spending, at Jensen, the we the look overall that then not hyperscaler when is And much. really buy growing for is, you that question

Center, billion, about confidence many servers are sustainable carve demand if next going? how seeing how of Just So growth next QX of billion, out that into is pie is just two accelerated? implied AI Data say us AI? your over continue giving what outlook years? give that take can are this give the the $XX sense is they some one of you generative years. we what to two look to that as $XX does that that to you Where the your So already So is more sustainable confidence one that for the I

ramping supply. the expect fiscal increasing regarding over So supply continue into our thanks well question for year. next as Yes, our next we that to do quarters as

to of terms other here. many suppliers, very percent, many the across HGX products It that spent support we coming are new have our we suppliers improving building is of we different supply. our something it's have pleased that so and them with long their time of In are that with market. have that the parts But an and many work not different a so we with both

for most energy effective time. platform We're The worth that centers because of at simultaneous same computing the the seeing process is $X that of of the doing And centers accelerated about AI. shifts into of lines of And way enterprise cost-effective, is data in $X and world computing. One and data and computing most two the performant in it's the accelerated along generative is transitioning the trillion the reason has installed, trillion otherwise. something cloud, most now. in

gives everyone and generative to of this platform enabled a AI along. two computing. by centers, data do by generative of way doing this transition came computing, compute the reasons worth and incredible to now sudden, about a you're $X then what purpose trillion of new each all AI, accelerated So doing computing, way enabled And call computing, to spend application of general $X.XX accelerated from capital year. it, trillion of It's shift classical seeing,

data world of around focusing it are on and today, the the most two that generative accelerated computing You're the centers taking computing trends seeing important spend and capital AI.

not thing. so seeing happening transition a same This near-term we're long-term And think two at this shifts I is a these time. the platform industry is and

Bernstein to go Rasgon Stacy with we Next, Research.

Your open. line is

really like like could for even question. of how systems if how to like at [indiscernible] wondering, Thanks you the or pricing the much that however Can I'm is, you HXXX? much versus actually me quarter, content us in guide trying units get GPU, Data going What the like guys. give or Colette, maybe taking Hi, versus just I was is is around forward. my you to that? any tell color Center want define DGX growth versus driving the

Sure, Stacy.

help. me Let

our very growth Data but HGX significant also that selling still as our Within systems Center well our both these seen. as the are Ampere we in Yes, we architecture, architecture. Data were quarter, had the systems a our our of include market. Center Those HGX our Hopper of of part architectures

does when that and from Now systems the as is both of that that, of in increases. about a substantially, quite is terms you unit, driving course, mean growing what think revenue the

the both Data of Center. these things So revenue of the are inside drivers

HGX going is last software The we we're for GPUs, revenue that will with and other sales additional add our will are that a was in this we're the that we even largest LXXS, forward. DGXs and many The but the But together consumer that rest to driver that continued that's importance there Our our definitely we market sell customers great are quarter about also of coming opportunities portion the system. new always systems with of sell. they enterprise our a Those of of DGXs, companies. again, coming GPUs growth Internet within talk types different seeing of have customers our we doing. the portion

add And Stacy, if something. I could just

mind. a robots to it’s trillion lift. HXXX a say Takes XX what parts, robot is X pounds it to you takes many in know image I supercomputer. a You nearly transistors supercomputer XX And know and test HXXX – in your it’s combination. the your mental build pounds, to well, build XX,XXX because But to

and marvels, the technology are them is really intensive. of things these manufacturing so And

And chip really, it’s to HGXs HXXX off out HXXXs of sent really the go we components, hyperscalers I world’s system think and comes so if as really they’re call will. fab, as if quite it a that large you a but

to with go Mark we Next, Lipacis Jefferies.

Your is line now open.

like with you platform. this if your it? there taking you've you hardware is of how versus part was a hardware percentage key seems part question congrats market success of in software chip I And was could ecosystem, us -- for if to along delivering the my you. understand your the lead care just dimension with then the and of on a the on software to it way us the what this. elements. years I many would had and the Thank Hi. differentiation question your differentiation? critical Jensen, your on two-part quantify wondering success. is software And evolution the Thanks like -- share a value And help NVIDIA is view in wondering the two, success the building I person on the platform invested the ecosystem of

question. Mark, really appreciate I Yeah, the

Let inference for you'd be It scale-out any and model It could data into scaling the a like a you any hyperscale out center. will, on for center. this is is, that could training every the of our machine a run end-to-end VMware. I metrics, scale-out me software deployment, on And of stack. enterprise the so center, data the if to from for framework it for like just AI uses of to called have about Enterprise. a processing, run we you company one time the This see some part learning data data example, do, use could the time do a that if be

our You on this any do of can GPUs.

or also computing GPU. multiple configuration field And just has sessions millions as or of in multiple We the per multi-GPU every and cloud per well multi-node. of instances GPUs as in and single cloud. millions GPUs have hundreds It the single runs a GPU compute about it of in

scale. per multiple center instances GPUs, entire from nodes GPU to multiple data multiple So to

software this X,XXX called among So has like enterprise has packages, AI time something each run something other. NVIDIA and software libraries like dependencies XX,XXX

time is combinations insane. I application example what to say of really number run and characterize number of would updated of optimized it combinations base will, for installed type But to quite work. that's two our elements mentioned, taken of And The I you is, company, accelerated -- it's that And computing several. continuously just And for stack. code one as would decades get take probably our as would and our I what our if are the here. us get X is to architecture.

inference. that people that's versatility architecture the the The so But when than to then do us the processing of workflow data, performance and from possible inference training quite post anyways, train to you for The focus of that But just computing a our tokenizing could of we'll languages it's flexibility, the much and with the or more applications. things it all said, these just of requires inference, -- lot it's actually data data training do where I intense amount the is you the it. the preprocessing use fine. makes before to systems, of -- of to processing for

we that deliver the reason And makes our of ownership. so for for many lowest accelerate different to it things. the combination the so is because cost us possible And architecture

second company the is base. installed our The of characteristic

because large largest that on so developers the that investments to developers so reason a users, end come software You is a that have business that get that of yourself, number installed make. base is for seek can platform? they the our could they the build to reach all software And the ask or they return why a it

training both reach. public-facing that so we use is for third and cloud. it in and way platform. all the search it so data develop are whatnot the then recommender train to use cloud And operate processing and because cloud, inference. cloud developers up today, or that to systems for consumption internal We're characteristic have delighted to use or in -- put to the our customers engines and CSPs public many the many and customers They

enterprise. the in so And we're in cloud, we're

had world's we bring the announcement. take the we're the -- and several enterprise. It's years And AI going to edge. VMware that. worthwhile we're all been a together, really enterprises at big way to for the we've system generative the to going look is to Yesterday, together a now, to working out bring of very world's together operating

of And anxious so And of to because in are another system is makers all NVIDIA's the reach, world's reason. put systems. reach platform their

we distribution so all and OEMs a of the our have forth And world's from of broad and so very on so reach. ODMs because and

a engineering. lastly, able able years. new were really this and stack different then software velocity, we and computing two we're we're to our It architecture our this and of because all complex scale of like environments. And hardware, models And every across sustain these to of do seems compute and introducing different and of accelerating networking usage all while the velocity

product a Now every introducing architecture, just six we're new a months. new about

properties so make on it build And ecosystem us. top their of these the business to possible company and their for

combination in makes those so And us special.

Malik Atif Next, we'll to with go Citi.

line Your is open.

you. Any a outlook. you on I on talked or guys tightness that for have my you taking talk and LXXS from product? how question job this about. of can LXXS incremental contribution the gross core question. help with? Thank And idea about profitability margin supply if the the Thank Great Hi. Colette, you results much can

Yeah, Atif.

models you. for deal large a really and LXXS that application. type processing HXXX designed language data. and just of take of a very great me The different is for is models designed large-scale Let for

transform is has It's to performance. a LXXS' LXXS' fine-tune not engine. got focus pretrained do be that's fine-tune incredibly a so well. able And focus. models, It to models, lot of that and it'll

server. meaning LXXS standard servers standard a is It's easy multiple into about world's data It get it and centers. to and in so rack, it's everything comes can GPUs it's You easy install for scale-out, in a to the install. hyperscale designed server, standard hyperscale

enterprise also with that's makers Lenovo that generative along with stack and many and going other we world's work and work the different for the the with why us did take building some and software enterprise is XX configurations is LXXS other did work and the IT systems. LXXS designed all to Snowflakes reason partners. VMware we to to Dell, and that ServiceNow enterprise around with about the and servers are it system XXX with BlueField-X AI so HPE, And enterprise. world's of the

so course, AI, case. for language And it's course, of of type will. if you but really is different LXXS a scale-out, generative of a It's, different It's, use models. designed large

great to start -- And to is so and really hyperscalers is going get are clamoring the off LXXS to deployed. the a and enterprise LXXS world's

Morgan Next, Joe Moore with we'll Stanley. go to

line is Your open.

Great. Thank to amount numbers so your the the demand customers. talking of me I these guess that's to that about you. thing of some is remains unfulfilled, remarkable

good getting. in a as demand, cases, quarters. a in tripled sort are, what of of than your As couple for you some multiples There's of are more people these numbers revenue

year. talk think you see of So do And equilibrium demand you extending can there you supply-demand have to about How that? talked get here? much you you next Do is? about into unfulfilled visibility into when line sight

Yeah.

year. by capital increase about AI. – offload accelerated -- about computing computing divert efficiency, the get year through generative think the planning world from way boosted. to improve that, way visibility accelerated to demand. The best and easiest next-generation of in CSPs off leading demand, to The available is data to transitioning easiest think efficiency their cost center the builders. with the will their and their CPUs, That's doing the the to computing. energy into so workload excellent And already the we're have their demand improve next is to companies Because is the much We the infrastructure center general-purpose but your data you're and to way going budget CPUs for throughput,

And this of beginning the the capital now do so point -- transition tipping seeing AI. diverting accelerated recognizing investment you're here, their is to companies and what recognizing this generative and computing

broad-based is this so that's is the ahead application over a platform, new This centers about way. opportunity And computing happening. way that a will, all the you new shifting a a transition data to the us. of And responding computing probably easiest world and driving that's think are but this to demand, the if singular in isn't

Goldman we Hari go Toshiya to Next, Sachs. with

Your now line is open.

revenue, Hi. Jensen. then you CSPs the clarification think Thank and quarter, about one last quick for CSPs question have consumer question. like consumer a XX% one Data XX%. of said business. Internet larger Center your I had enterprise another at you it I taking of percentage may remarks, on Internet XX%, your Colette were Colette, had your and for sounded been for Based

or you that helpful. that that, confirm super If kind would of can clarify be

And Jensen, a you. question then for

there bit there because concern the years. out demand could a there there's sustained Data in out and that your be the in increase profile a there your Curious and be there as key the enough have a generate if I'm of is forward. Center ask on breadth visibility Thank question in depth customer position guess for your their the confident breadth that projects, reasonable you applications into use of to cases Given I or a AI, I business you. customers a how of the curious your are pause will engagements you investments. return enabler to enough going support

Okay.

at So regarding customers thank our types we terms as the Center look our in our well networking that of you, question business. in it on have together. And combining we Data compute our as Toshiya, of

our XX% will enterprise large high a within our And our and of then our piece that CSPs, And next computing. largest be the performance consumer of be Internet revenue contributing bit little the CSPs last will QX. than more Our companies. category are

recognized Using Toshi, I'm from now that is reluctant time the is so first future of general the is and of It expensive, and too the are forcing general not the applications and for general computing computing. it's too no It's way question longer computer to energy at just computing science guess purpose too I'll go slow. scale to best answer principle performance purpose about the costly, forward. some the purpose perspective. brute

the center. for could by finally, called accelerated using It's AI. And the has it, into new But turbocharge accelerated what used kinds computing is And data way that's you of the world applications doing and kicked generative different all already computing offload of it CPUs. it. in a be

order ton throughput higher and and that's is save responding really energy magnitude, the in industry money cost You and the to. order of a of in of is what magnitude and

customers, to tons. to be And your and of the is, offer to Generative the new focus applications, new general power. provides well, purpose to investment number a computing. AI helps money it way is generating save way divert Lots capital way productivity, a services and data of best accelerated AI computing lots center computing ready accelerated the applications invest you save of of the and deployed. of from generative lots forward, new to developers, Going libraries. and on generating of It's in

centers. seeing is the to this, best and world's the world is this way cloud specialty whole going service think data GPU I around resources, recognize -- for the forward you're the GPU so centers of data that new This specialized crop deploy providers. capital clouds true for deploy a And

that ones because recognize to they're seeing generative service incredibly forward famous put the the well. same they you're AI. doing into world all thing, of their capital and all But and the invest best going providers the over way CoreWeave the it's GPU is accelerated computing to regional And specialist is One now. it

want it support that several with order VMware CPUs possible distributed high-performance of the But for the a to VMware GPUs, just do VMware. the We're been support security GPUs make BlueField to enterprises, system, but we've to software-defined not do as all to enterprises supporting NVIDIA's to well it, also enterprises networking. of and for years of capabilities as virtualization that. of the have approach computing center virtualization And working operating you seeing in data system, management the that's and for

we've we force, on as customers is, libraries this large around by several And VMware which all working hundred all SKU be And world. VMware's some know, is as of special they thousand that reach the been a new because quite SKU now sales generative the going is AI offered VMware be LXXS, generative be for a able center HP, a AI new possible Foundation. to enterprise this be and Lenovo's with and Private to any enterprises. it engage offerings state-of-the-art will combination going that based Dell, server have AI And called data on in could And new SKU to makes AI.

think I a going very seeing platform predict quarter-to-quarter. hard answer question is happen now I is But exactly that so shift. very, the the that we're And clear what's think trend to to to

to we'll with Timothy Arcuri UBS. Next, go

Your line is now open.

Can rate is the other networking compute networking your of -- to talk your attach something with words, of compute more is about lot. can of In prioritize half the maybe Thank the a GPUs? to you allocation your you half? you. solutions shipping? to than use this Thanks solutions And that less like shipping that your you're than half,

XX%, our the so that kind it, And we hate customers that what GPUs. let the allocation very the use the significant, building to enormous large don't is, XX%, billion is of prioritize networking for they of would We customers for free. to are the Well, backwards, infrastructure, use. throughput decide efficiency no-brainer. of higher say to working reason like And that InfiniBand I because a is savings. networking $X infrastructure translates to XX% for some Basically, InfiniBand

choice. largely you And infrastructure InfiniBand so, you have AI language a application, a if will, if dedicated terrific or systems, large is single large or it’s models to really

had Spectrum-X. have you but generative to is you recently of to so of solution will, a bring of different you center, Ethernet – the hosting users AI it, and announced the the of if Ethernet your going within and it, not really Ethernet, to environment some just enable it’s excellent that However, manage get capabilities can we if InfiniBand to all for Well, of called also, we excellent an capabilities, we’re that data we there of you lot capabilities. to the you’re way allow core

ramping just as along requires goes is to Spectrum-X it It is with that software switches. and spectacular. of is my a start. supports and really Spectrum-X And it’s and just both you performance bunch and know, really dear BlueField-X a tremendous project to all makes heart, off BlueField-X it the additional BlueField, Ethernet possible So our a now. Spectrum-X of whole it.

a think of I is be is the with in run. lot software and it it’s going computing run. is in-network a being of home and fabric home BlueField-X, realized a to concept the This putting computing

line comes question Reitzes final Ben with from Our of Melius. the

line is now Your open.

can about helping a reception the Hi. business? afternoon. is much. putting right question you talk me then does to Thank with that here. about seeing momentum it's now talk you that bit. Thank and materiality is you And for regard in already software you What And the how you're Colette, and also business? like run Good seem going? the the evening. of question, margins is it your Good the DGX Cloud. Can My very rate

working leading are of to ultimately, Thousands the of several DGX all of recognize world's between the achieve able world. enable of of our one, And segment, really me land number the work are the AI. Cloud's AI close we like of one course, to world's companies in with a generative DGX let companies startups. and clouds. the is many strategy, world's XX,XXX CSPs. them start -- fastest-growing them We they be is us there. and around that AI things: would strategy Cloud's some to XX,XXX We're generative with partnership start-ups. to

Cloud simultaneously partners. all as AI of them our could of in so we one and DGX built work blend footprint help partners we so And a with world's clouds our that inside the cloud leading easily

closely to our second to ourselves and is performance allows benefit improve designed historically the which like clouds, computing CSPs hyperscale The high-performance work is for that for and not together distributed multi-tenancy designed it generative of really AI.

able work and architecturally hand the our performance performance the in be to terrific. has really engineers have And to closely work powerful, improve been really so to computing to hand networking

none DGX require design uses And language our we to need significant. that then uses our systems. our are infrastructures very our infrastructure self-driving AI NVIDIA software team, of even and compilers engineering possible It's model well that the days generative without AI, AI course, to NVIDIA infrastructure large compilers And our these team, our our thirdly, develop. of research quite amount of publicized team, our And is ourselves. AI team, and software car optimizing Even optimizing chips. been requires

so our internal forth, own AI, AI. so so the on Omniverse so -- and of so consumption And robotics our on and teams, needs forth, all team,

DGX so internal and quite as Cloud. we that consumption And well, our is large land in

it. CSPs of love our the for success. been work it, cases, internal And has own multiple to great to the And and And drivers, ecosystem it's just off Cloud us AI around and way have more the and a enormous engineers our love engage it's with it so clamoring an are DGX multiple all of to world. use closely developers

of gaming In And software our Center revenue. that almost remember, software as our products, of answer all or products, within of our our question if and Data see let's your products part well, products. regarding of made part our opening I our GPU we systems a they're remarks any can automotive whether is future

a are the correct, software across continues upgrades stand-alone services, in business. standalone both to that You're we're grow well. as there selling we providing software where also And it

marketplaces. more our Now of availability we're business, with probably think DGX, software dollars and looking selling, AI of of such as at products included hundreds to many at millions this such the our even annually for PCIe are we're versions to of point, that with as CSP I our our seeing, our be we see NVIDIA enterprise HXXX. And going we're

I see So continue we're off grow do we'll and believe to this start, a great going forward. to

remarks. does back for And question-and-answer Huang additional over to or turn I'll Jensen any call conclude closing the that today's session.

Accelerated energy better data software era which never-before X centers and performance, of possible computing shift global achieve to are to in and A AI are now enabled accelerated an The is accelerated going generative industry and computing accelerated driving computing generative from generative The purpose AI, $X Together, to computer simultaneously is industry begun. centers through and making AI. a platform magnitude new, will accelerated efficiency Data computing broad-based shift. a of shift platform general platform has computing a trillion transitions, order transition platform applications. cost. computing. driving enabling new

tremendous. is demand Our

that capacity. -- world’s been can a and this The significantly makes to What over vector architecture has and our for our industries data substantially energy NVIDIA architecture. giant speech computer two decades center industry increase year. TCO for for training, platform created the this of year recommenders to and will from rest of are build vision, NVIDIA the performance are: accelerates and NVIDIA preparing Supply special model, has real-time databases. everything processing, next world’s the versatility translates expanding best lowest new one, to upon. We computing data and AI every inference, efficiency. production

in developer’s hundreds has for the data of more NVIDIA a of applications base developers platform. in end and users is create enterprise valuable their reach. more large base. installed worldwide. More generative to workstations, NVIDIA and PCs, has NVIDIA System significant OEMs computer AI. invest like has in because robotics. reach millions edge, computing achieved instruments offer computing suppliers NVIDIA ecosystems. OEMs, clouds, preferred installed and Developers centers, GPUs NVIDIA demand CUDA-compatible Each scale and velocity. that industrial Scale accelerated fundamentally Three, unique need business. models make and customers. is can and and Two, is grow confidently significant we invested NVIDIA reach. market XXX%

and Our road ecosystem to partners a can focus to expertise, scale grow. them deliver and trust help that reach the strong map have we

because We the of results additive capabilities. these of accelerating are

universe about products adding two every new six months We’re versus to and AI. of upgrading expanding the address generative years every

scale-out possible the switch, and servers. Grace the AI and universal on Ethernet next-generation for on infrastructures. already AI best Ethernet our software training enterprise NIC inference BlueField-X models, language LXXS working Customers want increased new large of we for of output who GPU for up computing helps performance we’re our Hopper. customers consists and accelerated generative our are HXXX of cloud Spectrum-X, ramping with which Super scale, While and

our generative demand the and leading HPE, ServiceNow, and the are generative partners, to Together privacy, enterprise. the we AI Dell, AI to world’s enterprise Accenture, VMware WPP security Adobe, companies, world’s Face, model sovereignty. IT but system NVIDIA and Lenovo, Getty, world’s We’re AI and with Hugging enterprise bringing the extending with that enterprises Snowflake,

entire to are AI on NVIDIA our greater building the forward to It dollar and industries, largest generative industries quarter. for customers, next drive shift productivity of enable trillions and computing. physical use Omniverse world’s they We how AI NVIDIA, in build achieve dollars gains. this time We’re of progress significant heavy exciting where assets digitalize our can opportunities look partners the ecosystem to to but cloud, an is generational the companies world’s realize the starts Generative updating you automate operate multi-trillion and most to the in and in productivity.

today's concludes conference This call.

You may now disconnect.