NVIDIA (NVDA) 17 Nov 222023 Q3 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 NVIDIA’s Third Quarter Earnings Call. All lines have been placed on mute to prevent any background noise. After the speakers’ remarks, there will be a question-and-answer session. you your Jankowski, Simona Instructions] may conference. begin [Operator

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

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

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

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

to our financial With Colette. over reconciliation non-GAAP commentary, measures of measures a turn is our You these me the that, which website. GAAP find on to CFO can in let posted call financial

external and data gaming export lingering and year-on-year. macroeconomic $X.XX set a Starting XX% workloads growth generative Year-on-year broadening visualization up supply with face leading center. work very of recommendation was performance billion down declined of and in by chain and systems We driven as was and channel and controls for reflects automotive new $X.XX data revenue, internet Thanks, delivered XX% year-on-year. Simona. revenue providers challenges was language the our challenging sequentially corrections primarily and AI. inventory QX disruptions. large sequentially consumer down through we conditions. of X% solid platforms Revenue This pro cloud while models, and XX% U.S. as center record such billion, companies

As computing and grows the AI internet definition companies our will end NVIDIA cases. use number public need different traditional hyperscale scale expanded be of cloud and to convey deploying the market to service

as center and to and align high-performance commentary growth automotive will with going to forward vertical customer key contributed our energy, workloads also industries, computing, We data accordingly. relating analytics. simulations such Other driving, autonomous

enterprises. restrictions using generation available and and energy. to demand engineering nodes a the enterprises HXXX Xx products HXXX. customers for and for AI constraints. Microsoft with infrastructure the in service in revenue, NVIDIA Services, supply These impacting inference AI incredible During rising AI thousands second Cloud-based stack Google scale Enterprise, bring working AI NVIDIA previous value server latest rescale capability performance ownership Earlier soft, adding HXXX the the accelerated software the GPUs, tens training hyperscale AXXX together scientific GPU easing into training U.S. and announced third and announced China. Azure AXXX Quantum-X AI leading to NVIDIA continue HXXX adding will the Difficulty] high-performance full including driven that products deploy includes for offer systems the Web industries. Packard also a compared and large performance fewer Microsoft in delivered started its and which industrial exports Amazon will AI the industry including in and both, to China. NVIDIA China, instances Oracle used including and and highest of train, HXXX-based simulation year, to and the platform. and QX. cost other X.Xx cloud shipping or our and software software versatility cloud-based sales AI new We starting it state-of-the-art are suite AI of certain inference flagship today, networking scale center offers China AI Hewlett the to more thousands broadly of XXX GPUs, alternative multiyear and the build thousands are AI largely we on of entities is and less HXXX government high-performance to lower include in equivalent at complete remains supercomputer as new HXXX Infrastructure. growth any the benchmarks. in collaboration based Xx makers [Technical workload new enterprise MLPerf help total also This based by delivers and offset the Supermicro. an product Oracle Microsoft available restrictions while That of computing advanced models. tens Networking NVIDIA Early based for month to Lenovo to incorporate and per address posted Oracle and destined bringing Cloud be to architecture and communities gigabit InfiniBand AXXX next server stack, quarter, this said, strong Dell, will by first cloud. enterprise quarter. Cloud, of offerings, computing from across adopting NVIDIA Azure cloud our quarter of demand expect to data Hopper on our systems current HXXX to impacted we

VMware, platforms DPUs BlueField rearchitected an has leading important building to per networking InfiniBand this now years whose milestone quarter new virtualization Ethernet gigabit XX supports momentum. and Quantum-X platform, DPUs. with are We over Spectrum two vSphere, run server last and second achieved on our the Our been

These Arista, the is Palo Alto growing design and available list of XXX systems, Check pipeline top on including Our Moreover, the AI of first is GPU Red new based GPUs. win The new Omniverse top including number OVX, Henry, the GTC, HXXX Computing one joint partners most on the Institute’s the softer our enterprise Juniper, NVIDIA system efficiency architecture. of or and the the Ada first XXX featuring powers Flatiron NVIDIA-powered system infrastructure XX% virtual platform energy-efficient servers. Rover The BlueField number latest announced LXX week featuring systems The Point, is PowerEdge will Omniverse Enterprise. of of supercomputers we the Networks NVIDIA to the is At the is Supercomputing XX% operate NVIDIA Inspur, customers by the XX OVX expanding, System, Hat. be available which designed the reference top first Lenovo of Martin Dell list. from systems ever systems. We build on Supermicro NVIDIA this highest the to and XXXX. released worlds of and total systems designed receive XD OVX among XX Land computing. will early using accelerated at demonstrating GreenXXX energy Jaguar ‘XX Lockheed and number Lovelace of list, be DPU are and has

and easily customized are model and BioNeMo our this with for These chatbox, language access We content NVIDIA are applications AI and large expanding co-development, AI models and property biomolecular developers services deploy entering month. adopt summarization, both early predictions. generation, to which enable software services, protein offerings language structure tech large further

macroeconomic a ago, relatively and help solid conditions down products lockdowns was lower partners to was channel expectations. Revenue levels exit $X.XX gaming. XX% in Sell-through for QX. weigh year billion We XX% normal in consumer in of with and as we and gaming sequentially Americas our to approach reflecting China down to Pac softer to inventory EMEA, on are as continued but sell-in Asia the current align demand. believe COVID track Channel Moving levels inventories demand from to on

an new had Lovelace Ada exceptional Our GPU architecture launch.

the amount Ada in and feedback XXX in GPU, gaming XXXX generate next sold first positive X, the for to entire features working quickly up play. out The became XX RTX many The hard technology GeForce RTX locations community. of are series game The frames rendering We that tremendous mid-October with neural the uses and a and keep from faster family, today. is demand. available RTX and to AI GPUs member available DLSS the Ada

helped for and raised bar RTX gaming. Our graphics the technology third-generation has computer supercharge

of $XXX game this Championship For Steam continue the where reimagined than just of of League and Dr. energy DLSS will hit fans is portal Strange tracing Maverick; Modern Legends Warfare record it top with prior the There in The -- Steam’s office on number X record sold on the sales, XXX with combined we tremendous million blockbusters, Francisco in packed and the of Multiverse old a XX Madness. more classic And fundamentals weekend exceeds peak of RTX in now month’s State the games in term. community Esports to and Golden with has million franchise number Activision’s most fuel a gaming surpassing the now World January. of users simultaneous in example, long applications for play. the total arena Top X The XX,XXX movie full million, ray in XXX. made Gun: XX of out box openings Duty topping minutes XX-year the Warriors set San the gains. over that opening Call believe strong of wish-listed

total games GeForce QX, cloud We XX continue over expand gaming bringing library, service. to the X,XXX. the In NOW over added to we the to

gaming new including Logitech, and We G, also Razr on GeForce the cloud Cloud Chromebooks devices, Edge. gaming launched NOW handheld, XG

on design, workloads. service channel engineering demand operate was ProViz. and our on to comprised runs sequentially we help to Cloud long-term down computer, fueled align software These reflecting current a on adopt for cloud increasing is enterprise NVIDIA the enabling improvement twins NVIDIA stores. by power safety graphics our challenges, twins creating network. German and largest wireless train intact, and computationally anywhere National some intensive as lower and near-term analytics optimize its XX% a delivery expectations. advanced inventory the of Omniverse rail any applications network design, QX. design of Omniverse offering, device. Cloud we HEAVY.AI AI, create believe and scale physics metaverse partners the Services, and first to in dynamics to Bahn, using remains teams global down to of workloads operator Leaders infrastructure center Services to Deutsche Despite from world’s digital of $XXX publish digital monitor Moving graphics simulation, to NVIDIA Charter’s Lowe’s opportunity for graphics AI And help using HDX of with are to Charter delivering retailer, a system simulation. twins OVX network, operate edge. in the computing build levels continue the sell-in and their and GTC, and models expected to company, AI from are Revenue network Omniverse Omniverse ago, is reliability. digital continue Communications Railway distributed for network, XX% million to Home year announced for Omniverse. it and artists, industries and metaverse a low-latency At Omniverse NVIDIA data developers

design in-vehicle year Volvo’s $XXX its multibillion-dollar way increase an centralized with to the the and by will along platform. driven and XX as announced Orin both, to the with sensors. Automotive wins and Moving DRIVE be and to first It production also performance in announced of our great technologies XX% systems, SUV leap flagship Revenue X,XXX SoC Orin-based Geely-owned offering a in Xavier, NVIDIA up AI DRIVE be infotainment of simultaneously customers’ Orin a roadmap. customer. running Volvo that continue Thor as ZEEKR XX% our by our Superchip, million has of Other recently At available sequentially Volvo capable energy delivers DRIVE containing momentum ago. next our automotive. platform. computer Auto, and core while in to the automakers to reducing leverages we Hozon DRIVE DRIVE the Thor introduced model Growth Hopper NIO, unveiled This automotive GTC, Ada software-defined XPeng. DRIVE consumption. XXXX was with automaker both is and and use NVIDIA ramps is include architectures. and of Cars and introductions to solutions performance automated scale. announced models Grace on all-new Thor the powered new successor first cost for from DRIVE EXXX drive increased a model teraFLOPS automotive architecture is Polestar

related XX%, increases partially Sequentially, and warranty primarily operating in in inventory margins to non-GAAP GAAP XX%, was due $X.XX margins Gross reflect higher remaining related cash million. Moving We offset expense gross gross under by it demand to to in to up was returned GAAP was we QX, P&L. of at ‘XX. rest margins expenses XX.X% growth repurchases $XX coming and relatively expenses keep higher a China, of million through operating Year-on-year, largely compensation in to benefit single-digit both of plan repurchase headcount and levels were approximately salary the center the data billion approximately XX.X%. quarters. operating and form flat $XXX data share and end were had GAAP the expenses non-GAAP center charges shareholders dividends. and non-GAAP the lower $X.X share the infrastructure over to expenses. our of percent, growth we and authorization the these billion At December up

XXXX. quarter Let me fourth turn the outlook of fiscal for to the

still continue Further X%, the on December to we expect discrete expenses commentary me GAAP to items. other Xth, IR Vegas. on and be automotive, gaming minus other Tech excluding community. revenue inventory we $XX Xth are tax be financial income expected Virtual XX%, details non-GAAP the losses Conference end expect investments. closing, sequential We’ll information income plus million, expected be In expect are the expenses Arete offset revenue correction. driven growth million to events CFO Non-GAAP website. our XXth, rates in included to the are to we ramp to or expected work of modest be or basis continued We In our billion, is points. All-in, expect the automotive, and below and center margins are shipments expenses demand by with expected of or minus And million billion. and an November X%, be Capital be and non-GAAP approximately data growth through minus to HXXX, softness GAAP non-GAAP Las and upcoming attending on on to reflect sequential China. in financial approximately our Orin are in million. X%. XX operating January billion. expenditures the are center. we early continued as expected GAAP the excluding Suisse design available plus approximately and of gross $X.XX $XXX be expected $X our gaming, resume production $XX.X GAAP in on nonaffiliated $X.XX wins. by to are respectively, and Phoenix $XXX gains the in expected Forum operating any data approximately and plus to channel be highlight Revenue let JPMorgan for Conference Credit to

Our fiscal for earnings fourth and the results XX. to February discuss our call scheduled are Wednesday, quarter XXXX of

We the for please will Operator, now you open questions? call poll questions. could for

Instructions] of of first Securities. [Operator Bank Arya question the line you. from America Thank with comes Vivek Your

first. Colette, I just wanted clarify to

business last a you at for billion a quarter, I gave think $X.X rate about us your quarter. gaming sell-through

I somewhat think is you said China weaker.

about now update of spending and is hyperscalers concerns large us gaming. to that slowdown. right lot question I for hoping for their rate you. a So, then, pointing was could you on And Jensen, the sell-through A what cutting

U.S. flat CapEx slightly data or in business do your is say, the year, you grow down let’s can cloud next think still and center if, So why?

the for Thanks Yes. question.

first sell-through sell-through start on our billion you business. business. quarters we me would for indicated, the gaming see two We Let had normalized if approximately together, put in $X our with

channel -- was Asia that China was QX desktop, Notebook was excuse and to China continue bit particularly areas. our business, lockdowns overall Now, sell-through still in challenge also during relatively softer, that indicated solid. solid. it although the a quite China We’ve quarter, and in solid sell-through me, relatively

stronger continuation Ada well QX end as the holidays adoption though, upcoming of by into demand, the as We though driven expect as enter we the

two is to indexed fundamental dynamics. business Vivek, our data center

has no first purpose with do longer general scaling. computing to The

scale is computing that saving slows. saving the of path so And as while recognized so, continue level we acceleration is efficiency achieve to can cost to necessary necessary efficiency purpose general energy money as workloads Accelerated computing the power. and scale, generally forward and increase

the content AI. in or quite using is or dynamic just using seeing phone AI. some who’s voice. is important The of to AIs like called interacting device surging recommender item important breakthroughs a cell now that deep is or a is second One computer systems, to to somebody in recommend and with best product sectors essential And very which demand a we’re

the of really You recommendation appropriate understand person request the them. need the make to and the context the nature, making to

be which led to with The with -- second do GPT-X, the of to led the languages We could years to human a example is the this It be a It language led recently, all of ago genes And other has called language of to biology. of now to could representations models large now Transformer, learning saw be language. models. which of associated first of of the language of I could have which started is the whole This kinds. the ability chemistry. bunch that. LM, one human language breakthrough learn which It of just several with invention BERT, genomes.

AI. has to do third with generative The

years, we’ve You dedicated know first that perception XX ourselves to the AI.

generative course, perception, the make Now, AI the But goal is to a contribution the is of of of beginning the now of to and generate -- this goal product of ultimate understand AI. create era to something is context. to

You more reduce whatever our images productivity or generating we the of with videos or all work performance productivity we they’re really is the kinds to it important whatever and ever. to with, over make have to and improve whether probably place, our enhanced see augment generating enhance text cost performance generating all than to ability to do

then, and for all my expectation AI these of than And could two company reasons. strong things demand than indexed and of cost power important we’re to that important more for ever. seeing the demand surging more are is and is our And course, which see are ever, that things, efficiency these which And so, both efficiency, you is productivity.

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

bundle enterprise talk focus we how started monetization? see And if with think to the can model, the flow side through NVIDIA with particularly now on should about curious about You how and on HXXX. the of of Enterprise should things? you kind around this software we I’m the AI timing Omniverse

Yes. Thanks, CJ.

Enterprise. excellent NVIDIA making AI in progress We’re

we In clouds. announcements associated fact, this several quarter that probably you made saw with

we start-ups language one announced But large more get into our that service all clamoring startups collaborations, It architecture for Microsoft like because they agility And stack our could cloud, full tipping that and more NVIDIA’s has the and everything so the of clouds. a have training to to single out enterprise our software which at into the years, so would of fantastic this the with building know And scale to developers if partnership reach so integrated installations all -- has and our the ecosystem several over and GPU scaling and internet announced company do and has ecosystem. because model do providers. partnerships could of world’s NVIDIA today, that there. have clouds. and we’re talking up to could You has large and that’s many quickly. of kinds. we’re their our with AI rich Every customers AI has startups to we we do scale, and we rich we And flexibility one point been than ever, of quarter, -- several initiatives, stack every more more of

of last software our the and to year over stacks are on of moving stacks to so, so, the cloud. we’ve our working or platform cloud all software And the been

NVIDIA ago, with so enterprise. of together included. standardize ourselves full that going so a That’s we that we -- very the part for doing similar take announced work to We, Microsoft could world’s the month that stack type large the software a same out today, partnership announced to all and the stack, And on a are Oracle. we’re of

cloud. integrated Omniverse been to model we would and software stack NVIDIA to that And do You of that their instance And the in so NVIDIA engage move large cloud, it’s leader also The the much cloud, the also now our customers platform it goal NeMo announced software BioNeMo of use computing If has has Rescale, it -- the license in all quickly like software. this recently on-prem, language GCP. is stack. full we and into saw customers hour; into like available to could could language utilize a announced much, per software can in they integrated in large and GPU Monite model put software and the license our would cloud is so through to they high-performance if more into we NVIDIA engage AI they subscription. cloud.

about now so, The And it. both available is practically work super excited partners NVIDIA’s you rich engage we software ecosystem have like in are with cases, would we everywhere that because to it ourselves, all their that consumption offer. for new they global, other could opportunities these connect both and also and this into to them services new bring both, but of APIs cloud and the other

And is stack so, great making progress. our software really

next the Caso of line with Your [ph] comes Credit Chris from question Suisse.

to in of the In that Wonder talked prepared you due you some inventory some some in documentation, And centers. a general. of your inventory took data was that if plus you spoke charges China this about more internal inventory then in hand on obligations. you being color and remarks purchase give the also quarter portion the about could

the what that couple need quarters? inventory, in be to for the general, can in next worked And you does to that of if be clarify So worked what And was your down those then if down? over implications charges. are needs still internal

Thanks for question, the Chris.

change in primarily, we looking for entry China. as just highlighted within that, So, data of forward for quarter. of reserves is to related we expected the center it the inventory booked due prepared million our demand all remarks, an in to business, $XXX our Most of

look this down. the AXXX, wrote which a good the center we we when data products, of portion also at So, was

our a architecture, even our architecture of and looking we that our has and have more Now networking upcoming just business. market, in architectures our coming that of terms that due hand to to our increased, Ada inventory inventory on at Hopper the lot is

building say. those as been come and for architectures to to to have such market We

looking We a are going expected always our demand each think just that done forward. job, expectation this in our levels I forward. at for quarter quarter end inventory based at at on solid we’ve the going of least But

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

Colette, can two-part have you I -- a question.

stockpiling because I with First, have in you data the a of lower ask the of guidance? of now the AXXX center is data is rate. the there any that version sort effect AXXX transfer modified

bit line. sense that could thought, one than revenue quarter, is can more? charge like the but get pretty because So, stocking sounded that might you while net that you it just the neutral. came Because last in customers I made to can it guess China in a little imagine less And revenue go still it. took they was that charge, be that a was a second part And related into of much you inventory

is inventory the related what to charge charges So, the to? just down related working Is AXXX that faster?

Sure.

working customers where that that very, the projects you center Most So, business of are build on being very They’re usually that let we with computing a is our data of and terms just see. we’re about business statement products accelerated held in for out that. and we to first specifically indicated. It’s AI. units see talk not specific me four needs their

can nothing So, that I’m see. going to no, we answer

can our long as needs or a provisions. At about inventory, inventory It’s to regarding question not always to term. current the softness looking future second at you build quarter the in quarter, us see. We’ve end It last quarters of see product in the for we beginning three for the been were two usually demand. takes statement China. Your

that So, are that’s always a of ordering. inventory case we the

it lockdowns, of economy going at continued forward, to think for our terms do led hard not what for center to what need time a seen in in we take looking China, now of So write-downs. us we’ve continued challenges data look we’ll was

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

commentary Collect, some center gave on on issues. I sounded like of had had China softness kind the you maybe a the It sequentials. question data

Are of sequential sequentially then you driven resumed said all gaming the gaming center. company of overall commentary? the like sequential QX if of But You given auto, little growth said grow Like about think we them How one actually growing flattish? kind guidance a three is just into and the growth. can by data down? for or those all just is they that how bit, segments do

Thanks, Yes. Stacy.

So, provided QX. from question your our to that for QX the sequentials guidance is regarding we

inventory again, are correcting as the have right get will grow not we’re which difficult bit. growing still are seeing working those of on guidance, million. three channel business say we’ve indicated bit. numbers levels, But the all will planning likely three to a that only correct, terms But different just little to visualization and those flattish you’re about is we is likely market It’s pro And to very growing the we going amount. in little for our of that increase. be to grow think we to platforms a As just of our $XXX platforms

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

talk embraced where vision? get a realizing maybe cycle? Grace is where this the Grace’s Could or and solution about vision or workloads, or about for of sense sense center will for a just with you’ve Thank an deliver Jensen, give do the GPU of Grace Grace data a I vision think where set an you workloads talk -- that of think. you xXX, for first importance us in I a within DPU that you. you, deployed solution the all integrated And -- most is off-the-shelf for articulated versus a vision, fastest CPU, will penetration what have

That’s GPU Grace’s GPU fast Grace its of between factor off special moving is effective memory, designed to GPU the Grace capability and capabilities also not and architecture Thanks coherent is Grace. expand XX. of memory memory are Mark. the to allows our data charts. by possible Hopper without a our that which GPU,

at to are applications data -- to related data large operates it It milliseconds example, processing learning. live that on It petabytes designed of data And for at at people be for using all -- speeds. hot. service. hundreds related Those processing fast, so, was make your needs which very can training, very Grace to effective for all so It’s also quite within a designed high systems, of recommendation is AI recommender is machine is you of time. millions a

kind and do really in terrific. will samples of We applications have I said so, those And that track before that. we’re on are to we think -- QX, Grace, production still I’ve

Your of Sur Harlan from question with line JPMorgan. next comes the

your out, partner driving BlueField data networking compute general the sales, purpose strong and in announcements exposure driving trends. business, million few Your is like years, guys DGX pretty and center like know is as term, hyperscale very strong growth believe, to solutions pointed very, VMware. networking has per and you cloud and own past also that quarter But near attached the spending I large $XXX over NIC compute about we to more team

So, and business for the what’s few over networking the growth quarters? next the outlook visibility

-- I The if first, thanks could Yes, take that. your for question.

heavily high-performance Our is networking, to indexed know, computing. as you

not the we -- don’t commodity We’re vast networking. of serve majority

solutions All high that a are and of of very end, for data. lot networking designed move our they’re data centers

center. on a has very the center a days, overall AI of you you the likely that applications, and substantial a network is that Now, if implication deploying have are number of these you your large data provision hyperscale data throughput it bandwidth

the high-performance So, billions provisioning in networking throughput, economics. of billions more of the their in which service to translates make savings dollars small frankly of they or increases incremental investment dollars

data pays -- for that’s AI in days, focused we provisioning and And is are AI networking and centers, fantastic really it on. that networking quite provisioning and where right disaggregated itself focus well, these But we high-performance applications away. and the AI AI application so, in high-performance with services --

by itself completely that NVIDIA is of that in world. for pays because the XXX it network that investment going and put were efficiency as network. NVIDIA’s largest is that You per reason instantaneously. infrastructure to And powered InfiniBand building infrastructures the if into Microsoft slow overall The networks, not you’re is for the significant might be gigabits one is And have infrastructure the AI second that dragged you high. noticed the are the by to obviously, of so

Mellanox. of announced when focus first all is quite important. so, we acquisition way And the networking places in It back to really the where we goes the

at in doing a doing a a were they to days, I we’re year, the quarter, the And quarter. do coming what quarter. dollars time, old hundred in practically up $XXX think few million used in a about about a now, they million

It to And networking. of an data kind indexed enterprise center it to you the overall spend about so, but highly is indexed that growth high-performance of and adoption. tells is AI

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

set the expand us AI about Jensen, like that is Microsoft want what When think I question further. of we in you deploying. InfiniBand your networking networking? fabric broader traditional that, a center Ethernet market, AI build help the for we a to networking opportunities the spectrum look compete that Meta announcement maybe doing does in positioning relative data sits of on understand And footprint the little on switching? against you’re kind can world like how where at for the today, on to they’re bit

The like math this. Thanks, Aaron. Yes. is

XX%, data the is you’re center and is spend $XX of entire these we the data improved to that huge. center. the large when by overall the across And going done If recommender numbers do processing systems, models are an on and language efficiency billion infrastructure

the we a workload very distribute runs across multiple it GPUs, so, multiple And and nodes long time. for

the so, overemphasized. the be importance network And can of

overall of entire because XX% to is the lot stack between efficiency. differences in computing very improvement the a us a Magnum We so, edge is ton network NVIDIA’s itself, of in in done software not the overall achieved itself. the what We itself, moving call running of call InfiniBand, difference of significant do data network we just is the easy in-network within software in And around. allows the it efficiency, -- at achieve, IO, in computing which to difference network software which the the with

hundreds billions so, on of millions if or on the of you’re the infrastructure, difference even dollars the And spending profound. is really quite dollars infrastructure, of

line Your with the from next Ambrish comes BMO. question Srivastava of

before, a data I grew had again, Colette, a said business the side, Mellanox of couple clarifications. that is actually was up was was it or you quarter as reported the And assumption out. the the networking center called quarter-over-quarter. Q-over-Q in it on quarter it compute fair because down

a to firewall. on So, the was it then what to Is fair really I USG the AXXX Isn’t trying understand or that of high-performance set what just it. i.e., government trying be of a is a you. $XXX sure trying going firewall, to band. customers? AXXX against to spirit compute, I make do, government I’m to that was million, is supposed a to the And is the going had the different Initially, -- assumption? clarification Thank

Thank you for the question.

at our seeing our should quarter. of quarter growth, last also terms flattish. So, about QX, compute flattish about in our is the with you but networking, Yes, looking look for we’re compute growth at is

reprogrammed is ensures the control. AXXX, export U.S. It It application for be hardware clear that reprogrammed always limited. hardware, customer meets test AXXX’s exceed hardware the the in cannot AXXX Ambrish, capabilities. is or that it government’s determines And of it it. hardware to

company meets business feeds it The make clear of and affected that the and and spirit. sure the AXXX million about was meets AXXXs government’s always test ensures But because in remarkable uncertain We for it our AXXX U.S. our the surely affected. did could that supply customers clear concern to so, time. our . control. situation not in the this about raised customers hardware export to whether of we were chain introduction and tests swarm letter in not execute, were we through And our $XXX

Your Stein next question Truist from Securities. the William of comes with line

that product in pace I’m product supply we well. traditional to a as this available gotten over improved quite performance product the there’s as bit ramp of of where whether hoping cycle growth HXXX can discuss year. as lot the and sort next progress look significant demand of We’ve you pent-up of for a a there’s this like the should questions

more of sort a rollout delayed more see perhaps look that trajectory does we start perspective, So, let’s or typical in, this say, maybe growth of growth where relatively the expect half from substantially we ‘XX? should of second

HXXX ways. ramp several than different ramp is the in AXXX

deep cost because that for this is significant data cost center operational Ampere transformer performance engine that’s of The first -- power models. now over is I the TCO, limited, is every so same. pent-up are designed are incredible Customers earlier, the are generative that and trying demand recommender as that models. to quickly systems models to models ramp because spoke I as large and AI clamoring of because language the Hopper and that of benefits, savings energy benefits for of the about new because because and the these The -- we do possible, the the and latest AI Hopper

up quarter. hands semi-custom to help It’s produces We company a stand cloud supercomputers on in deck supercomputers and miracle to NVIDIA supercomputer volume. to one a the the It’s in service is years. to ships every of high that are provider cloud the only Remember, ship the unheard every supercomputers. world all providers three in ship service

glove up And working are one so, them and in with hand them, one we’re stand racing of of Hoppers. every to every

have them Hopper expect QX. to in stood We up services cloud

are And ship some -- and production ship expecting to ship Ampere. in volumes That’s transition then volume QX, so, faster QX. a expecting we in large we’re expecting than we’re to to

And so, dynamics I described. it’s because that of the

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

does for reflection customer Thank is differently, the decision within is revenue is way something And could or center internal business hyperscaler I in a us detail that guess, the if Jensen, hyperscalers how that the the had to give that, guess, what mix your new Colette, what to guess to demand a you’re decision? there of you a properties business hyperscale about differently. and I of of about your And base. I thinking sort bit more the talked reporting commenting industry heard out I you about you’re I drove and script going mean changing that the on center you business. -- data demand talk break that maybe or data where of mean versus wondered you. of derivative what maybe vertical things little in just I about

for thanks question. Matt, the Yes.

bit Let we believe in clarify go little forward be what at center terms and discussing looking we business. of data should our a when we me

X, could becoming Our talk is companies. and that talk tend complex. on our customers we there’s a business to data large is we purchasing. reality they’re are based But of add what center And we larger very about about the larger companies hyperscale, lot different to that and when discussion X

and what building at cloud, really important are focus because the cloud purchases also Additionally, area where to looking our our is customers is is, our our is, purchasing. on where research education the higher looking our enterprise cloud at for this is where an

some of in a better way trying describe we’re to to the seeing also you better of these installments look hyperscales. seeing we’re the what large cloud give the we’re a So for and that color of in understanding

Yes.

that’s Colette double is just what dynamics click two Let which on absolutely There me are right. major said, happening.

around AI grown And doing has the world, adoption public internet they’re companies the in which service scale companies. and these of they’re number service the not services, lot, cloud First, service internet internet but it the companies that NVIDIA are by computing offer a companies.

has to factor second computing. with The do cloud

now cloud-first -- a the this are all to -- is platforms cloud, can is of a us, It are hyperscaler sell customer; this world quarter We the with reason the enterprise that a a be therefore, in that we a sell are exactly every cloud at since CSP, the point multi-cloud to are has also tipping hyperscaler both both strategy. announcements two they new quarter, available in partner. GTC a of year now last a made why computing. the and about Almost the this things all

the would of stack, business likely NVIDIA’s hyperscalers, business, the cloud partnership And enterprise of their On ecosystem we and their the sell service the have public of customers all side cloud their with their and richness for many clear NVIDIA’s public because very using enjoys likely relationship now that have would public internet with with and pretty overall the vast really us. be side of -- of it’s cloud us side because business of the the values so majority customers the they consumption. of full

the side public early it’s enterprise -- CSPs, the is the And cloud clouds only the very, of to to world’s at world, very so, their the public lifting to be world’s the clear enterprise because going very large. cloud that business is of innings

sell very with of for them internal public continuing our importantly, increasingly, the to hyperscalers consumption cloud so, relationship will course, And but to with include, for our sell side. with CSPs, relationship

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

GPUs the crypto in numbers potential your Great. some part blockchain but foresee business Thank looking you network, from at now, going could talk impact. the impact your liquidation that important being if any mining you to Obviously, at do you And are that’s down see you. gone road? for the an of do of forward? Wonder backward any point

road. We is don’t market. always There important blockchain see our the down a business being part an to resell expect of

XXXX upgraded secondhand the sold sold there years If you that reason somebody bought eBay, today, the that couple is for or today. are be major That is sites, a example, if somebody they XXXX ago resell to of upgraded XXXX to all could the because cards for enjoyed graphics XXXX the look of a at XXXX And bought a to at sale right any time. for price. is and

like when the And with That we’re minutes, the now it so, is my range, inventory because know, there. targeting the a the sense And used availability I’m the of for prepared never volume demand, of our homerun. -- our early very of were out all sure And ends a been lower that yet where would going and as sold it. likely and market. is inventory half than Ada signs we secondhand of And the the are, seeing XXXX market. right is supply drift the But zero. And lower usual, graphics price XXXXs shipped within of has cards also always Ada volume -- is affect clearly and was world. it they the top you’re large upper we in that you around launch were larger

the and charts. something gaming about strength and reception the And that vibrancy reception market. health the has XXXX of XXXX so, and says the off the the today of the And been of

So, launch. enthusiastic the we’re Ada super about

more products many to Ada have We come.

last Your Hari of from comes Goldman Sachs. line today question the with Toshiya

had messaging for I squeezing much I so your two ones for quick me you were mixed in. Thank On Colette. supply, some in think there Great. remarks.

I supply speaking were one think networking business, supply then And you talked a the being think at about talked you easing. when point. you to about I headwind

realize, think OpEx. to caught the you’re then -- million. a can you supply on hoping just if bigger stock-based but point. topic, was was in I it speak It’s kind becoming to of your it about And So, quarter, pretty secondly, demand up is mundane $XXX compensation, of I I at piece this

be modeling you. forward. that we So, curious Thank going should how

can at issues past, look one we supply every and of this has support and to our see it taken there. we next we that supply the our us improved really probably was quarter the have in for that When each is our probably to ago, so quarter, constraints Sure. a our that quarter, year this getting and had better. we pipeline Networking have customers

supply. getting decisions a regarding customers, to still supply parts up supply that’s looking chain in made discussion issues. data setting certain of We’ve our a as our capacity that that come their them for center, data challenges constraints And they’re therefore, been difficult. a When even through. purchasing also Now center has very

areas So, clarifies would we’ll see, that our arrives. our talking very be to what of it’s when stock-based In supply. about stock-based difficult compensation it compensation, two hopefully predict regarding what we were what

when also of employees to that provided our incoming in have it’s a date terms priced. single our and a We is to but employees, year once

to But employees’ So, important difficult continue compensation will of part is it’s determine. an stock-based our to be. compensation and

look compensation So, perspective. overall we it at from an

of the for when we’ll employee right we now a the do So up reduced have size same and until additions the that we with hiring now. focal, see few about level

now remarks. back will closing to Jensen turn the for Huang you. call over Thank I

everyone. Thanks,

in We next for quickly OpEx the macro alternative few China environment, the flat quarters. our to customers correcting adapting center keeping and products data inventory to offering are levels,

computing off the are around networking AI-powered record in data and center highest in demand engine world’s is and XD The the AI. formed for vehicles just great is with computing for on deploy the road our AI and accelerated build look meet computing general systems, sold enjoying computing. language general NVIDIA’s in crawl new stocks to help Transformer recommender Our ray first engines for XXXX generative a progress platform. Orin business, is than multibillion-dollar and path updating is NVIDIA power quarter. pioneering large and purchase NVIDIA to purpose our alone tracing libraries a revolutionary results. the Accelerated NVIDIA be foundation and the the run surging as the software slowed demands forward computing start automotive These Omniverse, with vital no the platform next synonymous platforms lines a you and Scaling autonomous Ada has designed reinventing to ever. next products waited cost graphics resumed phenomenal. our robotics, is with more viable, its models on NVIDIA that is standpoint. time is We quickly. and by work of Limited throughput both is platforms computing AI. out launch RTX Gamers Lovelace putting and longer to or services. to and world, AI companies AI Hopper, growth. more long just forward. from commuting to through

This you concludes today’s Thank conference call. attending. for

disconnect. may You