NVDA - Earnings call transcript Q1 2021

Due to licensing restrictions, you must log in to view earnings call transcripts.

Operator

Good afternoon. My name is Josh, and I will be your conference operator today. At this time, I would like to welcome everyone to NVIDIA's Financial Results Conference call. [Operator Instructions].

Simona Jankowski, you may begin your conference.

Simona Jankowski

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

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

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

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

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

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

Jensen Huang

Simona. Thanks,

in the history. many challenges, stride barely our acknowledge incredible care first with thank of Despite health here like also day our during busiest Colette crisis, who are service colleagues at broken those to inspires Before responders, the I'd I NVIDIA. have want on lines periods our to quarterly bravery the this describes selflessness. front of providers, one of and their they results, every workers, who efforts

efforts related X are to in Our focused areas. virus the

families more $XX have money and those our thus million We've hands, in far need. X to by months our care in communities. than in donated NVIDIA our put we're taking people more pooled raises of employees' and First, and to

science on NVIDIA's every Second, uses being technology done we're NVIDIA capabilities virus. using the to when fight of COVID-XX acceleration deal unique second counts. great A for

the and AI Some analyzing virus microscopy body virus, examples, drug temperature identifying cryo-electron at including candidates, the elevated resolution molecular the with with of imaging cameras. many sequencing

disease last And a COVID-XX and for need technology system, virus, won't monitor. we detect because scientific third, early, essential test the be community the of defense the spread to end-to-end be an for killer outbreak. continuously vaccine, ready develop to a development the and contain is NVIDIA computational of that system accelerate can next the

computational Imaging cameras to are deploy Clara continue detect over NVIDIA and days to We develop to from can watch the genomics patients. will and minutes. Clara to connect will and Guardian Clara racing Parabricks AI models leading with to Clara microphones research infections, institutes partner health staff accelerate state-of-the-art hospitals AI to overloaded care to platforms, analysis help

of NVIDIA's We data completed part centers. the business is platform, now will center major Mellanox and XX. brand as technology now market and networking one NVIDIA's April reported be Mellanox of and of is acquisition unit Israel our on

computing industry us The new Mellanox NVIDIA expertise and accelerate end-to-end a data footprint much data IT combination NVIDIA in are the computing, data in perfect architectures today, to innovation. center tremendous has forces AI. larger and center a scale the and and center position major shaping full-stack for scale

of high-performance learning to computing was CPU micro a AI, are and to Previously, and data unit networking modern computing. accelerated compute the machine applications From cloud service node critical centers.

together, and servers, applications simultaneously. computing of servers, and entire these unit the architect new by to invent Connecting NVIDIA users era data computing the basic is center. force the are hyperscale that millions are computing and is is Mellanox of forward, an orchestrated end together one composed end. an now serving servers CPU computing. extraordinary elements storage the to This is are together. thrilled that and The And Mellanox company, we're high-performance data I'm can scale GPU Going future networking. elements now of Mellanox center

turn over me call to Colette. the let Now

Colette Kress

globe, the upside Thanks, QX of billion, Jensen. strong extraordinary $X.XX Against events up year-on-year, gaming backdrop data of unfolding the a around our center ahead our had we and platforms. sequentially X% XX% outlook, down the was revenue and very quarter. reflecting slightly in

$X.XX gaming. billion year-on-year Revenue XX% down was of up with Starting XX% and sequentially.

in these of COVID-XX. marked quarter, with the expectations pleased which We by the unprecedented are the challenge results, exceeded

demand give iCafes Let color. was extended with in an Early period. for me you impacted some as unfolded, closing China epidemic the QX, in

of spread in As rise platform, have gaming GeForce and gameplay much by more playing our on hours people surged. gameplay both and home, from the driven learning we more per virus seen played and XX% working globally, user. the started world Globally,

growth quite With shifted has many demand in retail quarters. its outlets to to revenue accelerated our closed, efficiently year-on-year globally. products Gaming for X fastest channels laptops e-tail

SUPER to April, in the laptops which have to global for to NVIDIA and available OEMs, XXXX desktop summer. professionals most in partners starting QX of the working meet announced first home. with ship the record our last a engaged in and use our early GeForce are channel with are high-end the laptops In partners GeForce-powered learning needs students at RTX We our been playing OEM and SUPER working, and QX. growing XXX availability These since XXXX new GPUs,

and bringing as GPU point advantage ray are of computing $XXX, market RTX, capabilities. accelerate. needs that based are mobile XXXX larger the OEMs power These to take at the and audience well-timed to a remote its and of RTX addition, In unique price just tracing laptops AI launches a features enables including on

RTX NVIDIA surfaces the looks like with World The monthly glorious strong and gaming effects XXX for gamers realistic growth collaborated said Nintendo it Technica Ars business, Switch fog. sequentially bring jaw-dropping year-over-year. a astounding tracing console global charts. to Mojang Microsoft popular with and ray driving reflects, RTX behold. to shadows also Minecraft as over and scatters We most naturalistic that XXX with over our Reviews off sales Light in YouTube. rise Minecraft, lifted the world's was of both game and it and and stunner, on billion PC million views it and reflections. total through to called are refracts with

our Sampling. with Core We lead speed not technology Our a tracing trained making use enhance on in algorithm apart, function it the RTX next our games to on our but universal major It images, introduced DLSS can and from missing ray the version only can bits AI called frame, Tensor and real easy it nongaming-specific up X-year stands every class implement. and with RTX be Deep Learning using its Super the performance. of the of represents step X.X In time, doubling to fill AI from GPUs. silicon original,

The GPUs value to our grow. RTX momentum of continue and

have RTX-enabled tide significant upgrade games, Cyberpunk. and over We the a major year of blockbusters Minecraft like opportunity rise next the including and with

GFN Lighting Let Destiny around other It world, GFN, of stay-at-home is X gamers to game streaming GFN; streaming the million X in the with games GFN in expands of XXX more ray both touch onboarded. me gives than on game this publisher-friendly, as market added line gamers CONTROL, only underpowered tracing. the hours users devices. gameplays most is fall. that the another boosted billions also greatest and to to one developer-friendly sign-ups titles launching on service number the Games, with our of with Since in games beta the measures. and and which get X,XXX access popular the most such our exited with These League has service, reach February, Fortnite, game which of played supports Epic quarter. It by include and

Moving Revenue Turing by X% laptop was growth workstations and sequentially. up adoption. $XXX to in revenue QX accelerated down Pro million, XX% and Visualization. year-on-year driven Year-on-year

seeing including in of take also for a RTX RTX while time have fraction more ecosystem tracing. software, continued designers others. take RTX major upgrade Chaos of design a versus expanding creative momentum Autodesk GPUs. driving Maya, systems Group's opportunity VRED, supports reach. Pixar's of XD and support software and CATIA, RenderMan, visualization are that We its release all automotive now latest advantage visualization Dassault's We the in ray for leading V-Ray of Autodesk sustained packages, their CPU-based Over RTX has enables NVIDIA applications advantage now produce like-life This many XX rendering and to Quadro-powered announced designs the for RTX, design to systems. our

initiatives. received FDA Siemens, increased ultrasound, entertainment, deployed We update to remote perform Health as clearance in diagnostics-quality others. among Oxford see care, demand at in strong Caption Higher and health media fueled in education, And Caption health to cardiac was including and easier making Health. demand verticals, Disney and COVID-XX workforce and media research ultrasounds. related higher demand part AI-guided by it care its like entertainment for companies an

and Automotive industry to from year-on-year from robotic expect quarter affect and down The X% our down the that was well, impact second we is automotive significant as sequentially. QX. the about and $XXX million, likely Turning XX% to automotive revenue seeing pandemic, in machines. revenue autonomous a declining X%

work will our autonomous continues. moves that believe that capabilities. every Despite have We important challenges, the near-term someday machine

During an Xavier by DRIVE driving innovative features, automated NVIDIA platform. Level the the introduced sports powered X the PX, AI AGX quarter, Xpeng compute all-electric with sedan

NVIDIA to and Our new software over-the-air DRIVE deliveries of its capabilities. delivering open, features proprietary month. run next platform Production programmable, also enables with driving PX for updates Xpeng software-defined while begin the

than platform power better on Xx will build driverless DRIVE a watts every Xx platform driver way of architecture performance will their down DRIVE platform systems windshield few just for in to our NVIDIA from extend efficiency. more the appreciate vehicle letting all Ampere based top-to-bottom platform sipping of Xavier Ampere Orin, the single all fleet. power and a delivering called With Customers one Our scalability, power. software-defined the architecture, Solutions the in next-generation robotaxis assistant them

area and has robotics, new their robotics a automate announced that AI in technologies. to selected utilizing of robots NVIDIA advanced on factories, visualization the platforms built the we BMW computing Group as Lastly, logistic

billion, last GPU. center. Ampere the to week, the XX% sequentially, data $X.XX crossing up for and revenue first Announced AXXX mark billion time. Quarterly XX% $X a is up year-on-year record first the Turning was architecture the

in is Although across inference and meaningful Overall, as strong. and workloads, well quarter. industry including data solid customers the vertical high-performance just training, also announced, across hyperscale broad-based demand production, throughout contributed computing. full and to revenue It center QX as was AXXX demand is was

have QX. to visibility into continue We solid

to as our boosting GPUs, the date in computing for largest high-performance a the by its over as over scientific and offers workloads, well exceptionally performance and versatile, X universal AXXX data training XXx It performance is of generations up most analytics, predecessor. graphics. as AI important leap cloud to accelerator inference including The serving

The It inference. and for Technologies, Dell and Web it Centre A by job. leading be AXXX National will in multi-instance technologies Germany Microsoft adopted every Google several centers, power leap Services, training interconnected larger AXXX GPU and system into cloud, third-generation the platform, operate new This by partitioned possible Computing multiple its to be also providers GPU Research Energy ideal allows supercomputing to our Cloud, X the versatility, bring both AXXX National Alibaba that as make builders, instances. Scientific including computing by Cloud and service as makes for smaller introduces getting GPU others. AXXX capability performance Argonne to giant HPE Beyond the training Azure, ever one among Supercomputing deployed Center elastic Baidu rightsized NVLink the computing for many tasks. AXXX Amazon can cloud each Laboratory. and and Conversely, as is world's including Jülich the

inference most our AXXX, analytics. demanding AXXX advanced elastic The workloads GPUs for the DGX most the DGX from XX third-generation infrastructure center shipped the X from AI independent system to to configurable and AI world. DGX software-defined in deliver and to data launched is data and the training We

native software, two Jetson by announced for built EGX for optimized larger the the computing platform cloud, off-the-shelf AI Xavier products commercial NX secure micro-edge edge. and servers. full is EGX AI: We EGX for edge AI for AXXX the servers; Supported the at and

of amounts of With data trillions factories edge makes stores, hospitals, NVIDIA processing securely, the fleets carry we XG to streaming working from securely ideal update and it remotely. networks, like deploy sensors. possible retail of can of EGX, massive and EGX farms for the the out is massive which also real-time and and Mavenir. Ericsson of computational our with challenge EGX servers manage partners on are

conversational universality other we CUDA complex the XX announced science. data three workloads: and the fast-growing software of systems, accelerate Additionally, recommendation most performance AXXX's and harnessing and and to AI important

quickly every With critical First, recommender build virtually NVIDIA Internet, on are users models. deep developers recommendation billions leveraging our that the is powering to trillions systems, service. internet framework of and Merlin of application a enables items deep engine pretrained state-of-the-art recommendators the

that vision framework understand run applications applications speech. application shift each the Jarvis and that conversational terminology it remote using both create, a is real-time to its unique NVIDIA to Second, and customers to learning. Amid makes are for telemedicine Demand deploy home, easy from GPU-accelerated these company and for AI developers surging. and end-to-end working

in initial the data loading connect cost Xx has of And and extracting, and the with with delivers acceleration GPU data training the data we data analytics for performance entire end-to-end to Spark more the performance an a scientists needed improvement big power scale finally using field test that bringing AI. the acceleration in Spark. potential the data Spark, inference, of third, to science data of achieved and an the GPU-accelerated analytics Adobe to savings are pipeline, a worldwide. XX% analytics, engine for announced big data to we that GPU XXX,XXX from processing than Native and Apache transforming uses

platform GTC NVIDIA's Digital, More content to success an and platforms. of the annual technology AI online from than industry to includes this online event, accelerated XX,XXX for who free which computing experts underscored momentum, of GPU research which shifted Our and hundreds tremendous leverage registered practitioners the hours online gain of developers our conference, continues by spring format. AI

Our biomolecular combining computing cases, COVID-XX. or changes Scientists test seek to brings vaccine GPU detect now the our simulation solutions. compounds in these X.X fight the deploying treatment. ecosystem for around Times racing pneumonia truly it against scientist Researchers to virus developers million and are platform for a are strong. effective the in computing metal sequence, platform's in is like world the utility a AI

researchers Summit, XX the Oak scale National is identify first more screened map using powered Austin This and who world's the NVIDIA Laboratory which drug targets to researchers at Health, The University virus first Institute X,XXX supercomputer, from than GPUs. Ridge XX,XXX by National XD was GPU-accelerated followed of the compounds the who Texas came at GPUs. to breakthrough by promising of fastest NVIDIA at of using used create atomic application

million their The VXXX in demand of as help compound supercomputing the up virus. the protein GPUs they at analysis cluster some California, speed Diego GPUs virus software high XXXx Diego, Oak San the ported pattern, more XX analyze spread in researchers to what to the San of University Ridge are combinations people in from day. susceptible They'll microbiomic are can understand a analysis

Okay.

from GAAP and non-GAAP of center up Moving to the driven margins XX.X%, GPU up by expenses year-on-year, operating GeForce year $XXX gross primarily XXX% year up cash a up from and GAAP rest the mix X% expenses and EPS XXX% from sequentially and GAAP QX was were year-on-year, flow P&L. $X.XX non-GAAP higher was QX million. was respectively. QX product operations $X.XX, million, was XX.X% billion, operating and earlier, non-GAAP a and data $XXX ago. $X.XX, sales. was XX% EPS QX were

let outlook, to me Mellanox turn make the comments I few acquisition. Before on our a

Mellanox year-on-year profile. quarter, The range. reported exceptionally the Jensen strategic XX% has to and margins GAAP March growth, gross fit strong strong non-GAAP has in and of accelerating $XXX company million in revenue to discussed, its mid- high XX% that the Beyond with cultural financial

computing. our acquisition combined shared We team gross as be to jointly our the per investments free flow. we the expect immediately We and non-GAAP accelerated innovate to aim on and margins, accretive the of earnings road in Mellanox for cash to share full future retain map accelerate non-GAAP vision

me of With which quarter a turn Mellanox. the from that, second let of contribution quarter full outlook XXXX, includes fiscal to the

impact We in potential our have ongoing the from assumed outlook COVID-XX.

operations minus contribution predict, be precise teens the a likely reopening closures see given magnitude we low gaming, from retail be of around to economy. that to plus of we shift stores, likely in be will the revenue. and second sales billion, to the Overall, is Viz X%. largely revenue range iCafes impact to ongoing channels. percentage quarter or expect be expect platform by to Overall, to while basis total XX% we our The impact or automotive our sequentially. QX of offset and difficult is the e-tail on decline sequential down expect Mellanox the partial $X.XX of In Pro to uncertainties We a of the revenue

to going QX part But and providing breakout are of QX. become center forward, our We help data an this it platform. market comparability integrated between will with

XX%, and of nonrecurring. GAAP which or GAAP be margins and plus XX primarily The gross most to increase non-GAAP XX.X% points. basis in costs, an decline in respectively, are acquisition-related expected minus gross margins are sequential reflects

be expenses to are increase $X.XX GAAP to $X.X expected operating in operating the non-GAAP costs. respectively. acquisition-related and non-GAAP be expenses respectively. are operating approximately stock-based sequential billion billion, billion, GAAP and and GAAP an and in expenses The compensation $X.XX change year for reflects expected approximately and full billion, $X.X

stock-based full For the acquisition-related compensation year, also influence. and costs

discrete increase non-GAAP GAAP and both be approximately X%, expected excluding $XXX an are are Capital financial IR $XX both commentary other details our million. and tax million in to items. OI&E on Further and plus to respectively. to the approximately are expenditures million, included $XX available expected million minus be non-GAAP $XXX expected are or X%, and of website. GAAP to CFO be rates information

quarter, me highlights. an have for our posted closing, the presentation results also and community. investor highlight financial key upcoming this New let In events summarizing we

by with webcast Next Evercore. presentation will Q&A and XX, product our May a recent Jensen we announcement Thursday, on moderated

Needham's will Investor June We June Secular Cloud XX; on June X Fourth on Conference Technology on at Automotive Conference be Conference Conference also Technology BoFa's X; TMT Nasdaq and on Winners Cowen's X; Conference Stanley's XX. May on Morgan June

please now for Can please. questions. you for open we questions, Operator, will poll

Operator

Rakers [Operator Fargo. Instructions]. with Wells Aaron

Aaron Rakers

center the of do data of characterize to context couple comments the about the of year. solid your back you And Congratulations it commentary visibility that over I'm think quarters, your the curious how relative the was the visibility a quarter. last would last how half side, Colette, trends maybe what today in around in maybe on quarter? calendar of into we visibility that's

Colette Kress

the for question. Will, Thanks,

correct. are You

have year. to a of out we indicated see in couple of fiscal came We overall the that period digestion the were starting quarters we visibility improved ago after prior

into QX, overall we still for and move we visibility QX visibility As results solid data centers. have our into

this our of most think platform same are we indication true additional period. about seen say we regarding launched of its into QX the relatively that's AXXX, what And time, going at particularly and the our I'd about products. a that excitement is excitement So their had they and and

year, you also in in we hyperscale broad-based know, results. terms well Now the well, seen half to of terms regarding the vertical And edge as as continuing see expanding inferencing of record of we the as we're second them both at have terms and in our in AI. the of QX levels. In growth both as of grow industries,

ramp. products, of Delta demand AXXX DGXs, including strong Our initial starting an our but Board, was the terms in of the just also

guide at we However, quarter time. only do X a

So to the feel about it's still us. for us But again, AXXX. very true front give demand that's early little the we certainty a macro of in a good bit for too of situation terms in

Operator

Research. from Bernstein comes Rasgon question next Stacy Your with

Stacy Rasgon

I to on follow-up first wanted your gaming commentary.

should were about what And that trying when the of pattern be and those you it You you headwind, to around each into of things COVID into I to segment? Because thinking that suggest a of what of typical typically offsetting or it Colette, kind did sort know completely gaming other. offsets. with guess those comes I flattish trying Were a business tailwind seasonality, say still and QX are switches a e-tail couple has of to factors? mentioned was up. QX? kind offset we potentially seasonal maybe kinds

Colette Kress

to see wants on to start, me I'll add let So Jensen it. also and if

sequential I between you're think our about and QX. QX talking

Some of the...

Stacy Rasgon

right. That's Yes.

Colette Kress

Right.

over as in seen the COVID-XX as related to in had of our COVID-XX, our carry may well we pieces channels in had Some iCafes. fact, into QX impact terms of an QX. that retail

However, as we discussed, e-tail. efficiently, moved overall to

we in for move of in see QX, laptops happen. down seasonally terms consoles QX the strength do and that likely been we will and QX. normally have as We QX and overall But between to desktop

Jensen gaming additional overall summary, to QX it grow in expect And commentary. for if our QX do has between I'll and business. he we turn So see over sequentially to

Jensen Huang

that No, was was That fantastic. great.

Stacy Rasgon

just of don't like in in levels was Yes. chance to strong kind prior kind it's it if somewhat? I on double grow any You guess magnitude, is business mean But mix years, that up -- like of past? I thinking really of up seen -- that that, we are do of typical follow seen be us Like like different any back though, be very of we've that the up like there the for the give you -- I in helpful? you would it can then. digits. growing. Obviously, think sense on we've the it's could what terms mean

Colette Kress

Yes.

we just terms to low right we'll I moving goes. -- growth, about our that's that -- see now, mid-single probably to think the up be have in when what and digits the think of quarter sequential probably we'll guidance the in how

Stacy Rasgon

That's Yes. helpful. very

Jensen Huang

look Stacy, the this. think you picture, at But say, I would really there's Colette favor. is our big is would dynamics I exactly are few if what thing the the I that working in well a that guidance mentioned. add

course, phenomenal. First, RTX is and home is ray tracing was just the run. of Minecraft that

have games We announced already pipe or has that in been shipping. XX the

think that going is to be this every is computer the and RTX the game signed conclusion and graphics it's I Just ray tracing, is foregone the way developers generation. in a This to about future. on next

so think And run. is home RTX I a

got the create second, is notebooks that The we great. gaming. notebooks doing XXX in We just

mobile games well games Animal Switch today in quite the world All X Fortnite, There doing perfect. call the a games what or in ability for timing either stay fantastic. are our With Top know to the is platform mobile workstations have creators. for And world. top to X the designed notebooks guys and workstation, Minecraft everybody and needing you home, course, that of then, XX have designers And timing. gaming perfect a Crossing. studio the We it was just NVIDIA and are at Nintendo NVIDIA mobile we just are The was platforms.

think just our to all got we turns have out. favor. then I And in it see we how working dynamics so the And

Operator

Moore comes question from Stanley. Morgan Your next Joe with

Joseph Moore

rollout quickly wanted was between a smooth a as I Volta, there in well Is of little Ampere ask transition? as bit it the to that the I transitional remember the as roll about DGX launched on the -- side. pause. on a as And when how there does well hyperscale of HPC segments side you is various

Just customer can you us up you tell how ramping the with see segments? different that

Jensen Huang

Yes. Thanks lot, a Joe.

So wasn't data all, Accelerated step sense is the now It first first of taking back. centers. a when computing we case common in launched Volta.

If you went on the called generation computing. way, back and the inference It serious in really it was really later training was version to until Volta, didn't that with come a on and TX. the Volta was training training. first high-performance learning We deep focused focused

over are in that course X data workloads diversifying of been now we've the the last years, centers. But accelerating

very recommender probably center or of a imaging the conversational video It years still machine is GPUs. or workloads take common that the world important most our have data tremendously, machine part learning Deep learning whether the accelerated the hyperscalers, accelerated GPUs The deep that expanded their systems expect using as number in learning point, The a last data used at deep with machine X world. you this If since of unquestionably, we've all is significant being model AI centers in learning or it's accelerated notion most People now But the now learning is look over be. esoteric. saw sense. of pervasive. didn't at and today, pervasive. it of to now something

of the number we accelerate quite so applications now diverse. is And

accelerated -- it greatly runs that they that's TX started forward the compatible really that's already of a have amount Everything And that commonsensical that's GPUs -- by of to came adopt runs on compatible When to or AXXX, generation. the would compatible. AXXX. GPUs. Ampere. on so NVIDIA architecturally runs was everything are VXXX on know, our very They them on to And backwards you runs generation that workload large to ramp as from We're it. to we introduce when data contributed Ampere center, we

I to so really the going is think smooth. And really, transition be

of On AXXX excuse hand, up. had for almost in the the sequentially a VXXX or a now quarter. on hyperscalers adopted HPC -- and of TX services, cloud which, industries, shipment. AI with computing, VXXX was are by other roughly AXXX way, top our TX in the their half It and business. quite me, in then vertical because And which TX grew broadly we VXXX great -- overall and is that,

out edge, which a All very small -- Much important, to it's we of smaller of a high-performance the had centers. great way also supercomputing also to but But the supercomputing -- part, quarter. the AXXX shipped course that's is computing. part

sense think foregone grow, adoption of the computing it is And and AI accelerated so sense all the common of for now I the cloud general -- a and hyperscalers acceleration number learning it of has summary conclusion. to The of workloads using of that is the continued machine the and has grown.

think a product. fantastic -- fantastic a a with receptive really so I very ramping with into we're really And market

Operator

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

Vivek Arya

Congratulations on growth the strong execution. and

I'm market? where a in then terms growing of in inference baseline as the to are curious of of for contribute? how clarification. the And product new and for just as journey Ampere workload quick XX% payer And expected a us sense much Colette, and we are is edge give inference you, the kind in you Just question, margin? kind where the for of AI Jensen, penetration? a a are gross

Colette Kress

on So margin margin as we QX. gross let into question regarding our first the the look and start gross me

ramp very another now similar Mellanox. to But and quarter overall data as also we as well. to would new as continuing overall go architecture non-GAAP QX has is overall we -- a at some XX%. baseline record are center likely changes guiding a This we are finished our margins margin Mellanox that. gross our incorporates We just Ampere an record within as gross see be right level, The margins as QX great transition margins overall this forward. we see to even

a it's the pleased very But guidance will margins we're right gross now overall see to for still However, where early QX. with at these little go. XX%

Jensen Huang

its of beginning journey. the at just is computing Accelerated

you characterize as look If several -- at segments. it would I

AI microservices, a is When Whenever movie, we that the AI. it or is services recommends has movie it song. web, watching First that a shop today you're recommends all you the hyperscaler a product. on enjoy which recommends a it

All recommends news they friend first XX the a of those website, -- recommends a recommend. or recommends or that websites or

why these Internet. your personalized data learning the the on to predicting are they can The data. today. recommenders are preference, reason so they that machine predict collect, a the that All collecting core It's more and based more much could of all Internet is powering the your they're preference

It used CPU tens more be superior change accuracy based are in forward, The prediction economics. deep of to all of few percentage much But on learning of could approaches. and billions result a approaches. preference going largely in dollars results based it's on

And micro deal. this And services learning very, hyperscale very or deep the AI towards so micro still services in shift is big is ramping.

know, And have XX% industry is seen. Cloud segment you that $XXX ever billion $X growing market transformation trillion IT today, Second into largest cloud. we as single of a about it cloud is is opportunity. the computing

perfectly positioned is is cloud that powers these forces. data perfectly, force the really our driving benefit two The powerful two that AI -- to and -- is the two really forces computing. from center We're business

is in to intelligence, that turn is and -- So industrial And IoT, connected single -- millions. a every case XG, whether or or They'll long the industry. and the industrial, centers. construction, warehousing today every the trillions sensors, the going or manufacturing And all second industrial data farming, single But the logistics those artificial micro And industry journey be centers will industry tech it's it's a to it's over be become to of today. has third there'll is XG In into be the a will go. of the distributed little ways be Then edge. data cloud combination not edge. they'll computing. tech future, industry. And

that announced that And know, we service just important factory lead automated this for Service, announced are working several of and working domains. transform We announced X last very quarter, Walmart U.S. in have edge We partners mail they're factory that a journey to the And partners who we but us us logistics service. We barely BMW, felt X has robotics, AI. with retail. announced started. into smart is with world's we on partners then people would their that largest hundred the the sorting future. And the Postal

examples just and applications the so to early Ampere intelligence ramp into. X these where artificial phase going And that of its is great of And is stages. next at are really

gigantic And data Hyperscale artificial so positioned well in the and edge cloud And to going segments AI. really are that all and computing AI, IT intelligence. I it's fundamental think make center the fair that forces to real X a of scale markets. it's today, we're impact say

Operator

Evercore. next question comes from C.J. with Your Muse

Christopher Muse

you guess AI? ask worldwide look you for question if think about I could broad-based. think you in long-lasting health I demand you of could two. And like more us calendar rate picture bigger for Mellanox do can what a the 'XX? specific help care, on for not and the growth really But then Colette, with COVID Jensen, how to impact

Colette Kress

in the to Can out help you you of You me? me? for me. first cut the middle your can of it C.J., repeat part sentence

Christopher Muse

No, sorry what handholding you about 'XX? think I'm growth for provide calendar that. we on a if for Mellanox curious little should could in about

Colette Kress

go we of quarter together. products the it's At generally we a know, team for time, building future start you as us. the X And excited to we're bring on Mellanox board this out, can just and little early beginning so

going had to tuned just their to margin, quarter them. great looks have overall half equally for performance as couple the the last of the over overall what March a quarters, stay them the great with had second of to a year we're well. last year. see they For They And

Okay?

Jensen Huang

industries for question. to think Yes, is it's And be going it's markets. quite pandemic the and tragic, reshaping structural. thanks This I really and C.J.,

really to that There's going be be think good are that areas a And a structurally it'll I going to good question remain. I your because is it's it, think I think very believe double down. now think I say where changed. once sensible. to about And to it's time few I

just the infrastructure enterprise them and world's need. moving to can't the much a provide resilience to to IT. cloud hybrid the The And Every single having digital first to They be have is more afford that they to computing accelerate. on going going is cloud, transformation is rely resilient. on-prem that company

one. And so that's

I now billion moves that to into when accelerates $XXX journey, be and transformation, is accelerate. And world year, surprised it's into this see $X which infrastructure wouldn't trillion the IT a XX% that growing

cloud to that. And is because so going computing accelerate AI of

unseeable. most of a the up each now country be year gearing every think systems And is to Ridge, to would scientific something about on The second of We system. does are a otherwise defense detect creating infectious Oak compounds based case they in importance of is I the And billion defense And in is defense take computational disease. that on lab methods? nation a radar. and an The take X our things going to for future, a are going based system It's computational company. defense accelerated that nations filter and government and computing that take what NVIDIA to national today day. think create in yet is are the systems it

that's And what do. you need to

defense find to computational to allows then, monitoring. to computational an to through And has containment mitigation the have ASAP. the system go find that early living you of the insight, need range way a warning system You to course, to detect from within entire accelerated

National labs And so gearing be are up. be are going going gearing to labs scientific up. to

hard are And going our to robots in labs. ability The labs, work of our getting third part us to the a to do We're in our have AI a our have those NVIDIA is to able time we'll remotely. helping has and be helping robots us that lot robotics. without have done. work

able little the robotics environments, all being possible be as outbreak. kinds of up outbreak. clean -- to for to disinfect these another event where so of to handle -- fumigate to dreamed remote in people with dangerous of either forward applications environments to environments, in autonomously, are as autonomous an circumstances of case interact And the to we society to All right to help capability have need now

permanently think single going the then to strong home. best working to a home, a to lastly, work people work increase it's are from when are people more which going home home. games. larger people percentage of from I video There's is and And entertainment, support more And movement clearly that from of going companies

to is entertainment think going of represent segment larger society. a much of overall games budget video I the

I trends, changes AI. And so good I of I are computational are would from these dynamics cloud would are these for us. say. national defense computing, say would a some system, that here say to structural And are labs, stay. be the to robotics going really and working home

Operator

Toshiya question Sachs. from Goldman with comes Hari Your next

Toshiya Hari

Colette Jensen one may. come gross the Colette, I well, and to for as I had for question. one then if back to wanted I margin

could gross And we of indirectly. scientific to that of Jensen. a may about sort guided guiding I what between sort in given non-ops quick July the role talked given you contribute But way outside coming at see one margins is, you efforts current mixed was I claim directly what of I in flat a that play I'm related where the in that the question quarter? the sequentially. the guess on feel current trade on guys better despite people guys landscape same can essentially guys about the pulling of in the hoping that, offsets be going critical and compensation some I time, You're some the down impacted down sequentially, with should you automotive XX% you thinking computing, for comment material the nor guessing scenario U.S. to pull one then is cadence you like and and X-month in be the And shouldn't forward, are are China. how U.S. the OpEx side?

that, kind on speak -- be of that if that So helpful. will can you to speak

Colette Kress

Thanks, Toshiya, for your question.

up record terms including quarter Ampere overall terms even sequentially we the our what our in second achieve we with a guide quarter, guidance overall that next our of in to architecture. gross with This So even level of regarding is margins QX. from had second record in at XX% hope is our

So a the tend typically, new over bit when we lower up architectures, little transition to can and somewhat but trend time. kind of be move on margins up to a onset

those other overall, than also, may to of growth our Additionally, offset going our nothing in transition. as But and in see there's its mix highlight you we're consoles, business which is really of some lower. structural articulated, platforms But automotive the ramp our gaming Ampere two. as such and in to

Jensen Huang

as been high-performance on longer. see, the Toshiya. living Chinese the for electronics are up tension. background based China's this know, And trade in a in gotten And trade has you been the environment some anyhow. computing for Let's time, We've largely probably tension coming on systems year,

-- And change that's condition won't think so going I forward. materially our

Colette Kress

that our which me is into was second felt respond it to current the the OpEx was for let question me, overall So our to something We year. forward Toshiya, decision had COVID-XX. your This regarding we've done later to QX. normally local our pull that in prudent you and during

Although and QX. family wanted than to couple upfront. We cash it our their about safe X members the months It just are opportunity had is for also safe. and guidance employees that have about is normal, of to in earlier a quite months, were make our sure incorporated

Operator

with comes question Lipacis from Jefferies. next Mark Your

Mark Lipacis

how guess if Is of applications. solutions trying evolution the evolution a and inferencing. to set the that A and I'm kind to the which you And AXXX. -- a the I solves it it over demand this training is question into of efficiently. understanding solution, the you going more solution coming your of for my understand about back based. were back, think you sounds fits now largely kind both to targeted I inferencing then that And training solution, and introduced it had of time have

framework inferencing-specific application? to thinking see then -- be the an years, of I'm computing product? Or about for the X the is down is this the solution Jensen, a XX where Ampere market kind an general would Ampere-class And framework Ampere you that guess solutions years purpose like talked you we of this or past, I expect in kind years, Ampere helpful. acceleration line, is still and you X in had of I still so part and -- If wondering then is think for class could different should a context, that would provide about in those solutions training-specific

Jensen Huang

call, the Mark. for Good Question. Thanks Yes.

take all -- last kind our in accelerating in came back, X data years last classes we probably and of X think starting X, ago, a current of in but way in I years into really of the step are years, back, currently centers, into the data our couple the the if you over they centers, way setup workloads, time. accelerating learned the then and it. acceleration There really X from

the a is up. major It's for a like -- the VXXX The training. or learning class first ideal learning today one on it's prior architecture And a scale architecture. discovered that setup to was It's workload we like workload prior training. was of the with that Ampere Deep deep and to is GPUs supercomputer -- simulation that like NVLink, weather called architecture. X SXM Ampere board, yesterday

computing trying the you called largest build operating can possible You're scale node for one up. to system

GPUs. larger have for many learning it, accelerated you that along nodes, they ability started world the we to that way And for actually PCI lightly cloud PCI needed Express, to access because they scale but way to Express grow bit developing more algorithms. learned want and degree have an a thing that platform their the was second the to And model the to perfect different SXM, then offer computing not a around up machine into the of VXXX and was to because to allows little to GPU the X budget, a get up get researchers all

in versatility, TensorRT and for really seventh of more inference, get performance but And Express, was on generation we PCI Cloud VXXX flexible started not TensorRT, so scalable And then it ideal. much as SXMs, as rentals. that the VXXX into renting quite our X.X. we're to was

the day. you precision ideal could where for scale that reduced inference today, up, able that quite which learning the and graph and And millions small even accelerate one days with is queries each electronics job something a versus training And largely deep application, It -- are more. job for that we've been more have computational have of X accelerate that's and every reduced a that not for because sometimes and there. GPU has the was Along be for inference is to called the focused inference we where is scale-out X [indiscernible], training is more out weeks. precision, and running you way, queries, of running

of application on of so each hundreds machine. period like Scale could and hundreds user done And out, it very time of query milliseconds a a one them for in is that for that get milliseconds. large X it a users, very millions last query you of long to for it's scale-up where have of uses small and ideally,

has And a so notice, out I've said X in center scale Voltas today. centers storage today servers, CPU has service has it servers has cloud acceleration different server, with data then and with computing, scaled VXXX. architecture data a GeForce it Most and servers based has flexible scale-up on

utilization workload systems and therefore, these spiky. so will so hard, the is ability the to And be predict of

architecture an created X we allows for so And things. that

one, training it -- generation increased number mean characteristics just forward. Ampere it And we XXx. throughput it's and -- where are: is training don't in inference, and remember things I gigantic I a the a the a leap of greatest generational history. is So inference X by for for for leap gigantic

big it's the We could is over the of unified. Ampere architecture world all in use engine, it's the Spark computation single The the to the this which data world, comes but many used engine into From to important used now the XX,XXX of computational It most -- processing, is you engine data probably customers. second, data. data Hadoop, called know, be first Spark. for it's ETO; computational that today the it's the accelerates the the center. moment

accelerate Ampere machine the deep have training, way finally also deep good to learning, all well the that. boost And ability learning, as learning, out as inference. for then We -- is it's to extra

third that's thing we And acceleration a ever, have platform the first now is then And ever GPU workload. the so entire acceleration for unified elastic. first the platform the it's

it. reconfigure could You

XX a for this for You or more want configure scale becomes partitions, of really mean computation powerful node GPUs. that using you configure scale Each than When it it's you're and it it I NVLink, those each Volta. you of GPU. it gigantic GPUs X you is can When could quite to small for same creates XX up, gaining extraordinary. whole bunch out. scale scale together up X one out, either configure it

And have for and the then unifies performance so fact breakthrough on three, all can now and workload, these Ampere number you a cluster; for elastic. acceleration of fronts it X that is the it's

use you in could you use for training. cloud, for use a could You could it the inference,

most that serves you That's of excited needs. is could about. Ampere that helpful. now the acceleration your versatility of cluster thing very I'm have so And X And all the

Operator

Arcuri Your with next from Timothy question comes UBS.

Timothy Arcuri

two, I for guess, Actually first I you. had Jensen,

strong pretty but there's that the things like have visibility you always on good sounds recently. CapEx, business, pulling concerns center data Obviously, have very customers Just it been are into in your July. wasn't attrition kind would immune But low was but I that time, really any into thought last that folks the that be of guess you case. digestion, so most also

you quarter, then if CapEx with maybe second things Colette, XXX I time, that side? how and like wondering, are my the it So digestion I'm a this comp implies to AXXX through I XXX now to guess handicap guess, question, had a different stock and goes get And running like the quarter. on guidance is been whatnot, your but question ability do any

up XXXX? you a retention? goes all Is that is it right as sort that So lot. the look executive level And of into

Jensen Huang

to want that then the... handle first? And do you did Colette, I'll

Colette Kress

Sure.

so GAAP delta non-GAAP help and the me let GAAP between OpEx overall our the adjustments, OpEx. So on our out you

we probably just than If mind, what in have more $X.XX year you billion about there for and associated there our it Keep we level expenses. stock-based is look GAAP at in full with the compensation. guided,

the incorporated those and and with deal good costs the for of accounting fees really do associated are will of also onetime terms amortization in and overall the that have items. intangibles of portion a acquisition-related we We also Mellanox, and costs

So Mellanox. the most of overall three But just is year need which also the our the NVIDIA stock-based with some retention for the of working onboarding compensation. onboarding Mellanox. for in part, and for what to them is stock-based we compensation includes is There quarters, the it influencing

Jensen Huang

condition today. then condition there Tim, our our several are differences and between

the of diversity -- we our inference. inference, then, were in still the accelerate. early Back and deep for now our early learning. we most center the We were So data was first is difference of workload first acceleration in used the

of versatility unsafe learning today, model from detecting, types being processing from of to a different data trained stands is video. number so training AI the learning And number of models deep video, training that's the the growing to for deep and tremendously -- from detecting from

language natural recommender deep now to conversational the The understanding systems. gigantic AI movement towards a

are that hundreds. number the The the models are models so of are And training Recommendation data being different growing. trained The is gigantic. size They're gigantic. systems models are on of hundreds of hundreds -- take recommender Terabytes, needed to the of it of systems. terabytes. of these sizes, servers is train data all to that would tens And hold hundreds terabytes. of that

are Euronet speech which the probably And all the important to most all today. of models the model elements -- of different influence to inference models all recurrent, diversity analytics from didn't a data Text language models, so recurring training We models. models. different at the is time,

those early models were time. us so And for at the

So one the of is number diversity workload.

of second The the is computing. to cloud -- acceleration

today, acceleration accelerated I a a it's movement to are IT that transition. going of be industry only $XXX the movement to billion we segment cloud that is decade-long think that about It's computing if time, is and not From where going our largely today, multiyear industry. someday, a where deep diversified half. went And much the only trillion We're clouds. At hyperscale $X just learning. of for out the be more starting. is cloud also represents was

And and cloud, including mentioned just and AI that diversified of of with edge iceberg. And we've not we're lot but out so out with of Walmart that's working I and the inference. cloud, And a earlier, do industries. of the significantly BMW USPS, vertical tip has and to as

a was we've great strongest what is demand demand data Ampere. ramp different. best then centers. it had significant, fantastic. I quite multiyear we've few think we're say had. was we've is the ramp. mean It a conditions is ever so weeks. And though would in is ever the And lastly of I a And starting The ramp. I little The to Even ramped -- are are the ramp we it a

are so some -- those the And differences. of

the are very different. I think conditions

Operator

with Your from Sur question comes JPMorgan. next Harlan

Harlan Sur

implementation, Volta custom chip has adding other team Jensen, importance you you interface pretty guys you now networking, some the the from compute how and example, and network a performance But your data to of about as acquisition, to Volta more OEMs, improvement of partners? guys guys in for like, scaling the Mellanox large whom does GPU. per fabric Ethernet among by from into out DGX-X those platforms center showed long-term move things, cards, networking switching significant skilled switching have a by think acquisition didn't then And fit when the GPU networking does provided chipset. fabric differ acceleration when been we networking the change strategy how Mellanox Mellanox' networking well? and DGX-X, But the other drove Cumulus

Jensen Huang

appreciate Thanks Great. Harlan. the Yes. I a question. lot,

most AI our you third-generation this DGX, It's the instrument DGX advanced So know it. world. really People successful. and it's in is love the

speed your If NICs, AXXXs And X highest is there there are instrument. they you're NICs serious and X DGX, researcher, AI the Mellanox are this a the in have.

reason how great because go and fit so it to no High-performance high-performance the And computer, computing one matter high-performance big a that is problems we're networking. And in no we is. hand-in-hand. networking solve have appreciation trying for the for longer

And has be to so distributed. it

is so one why and you the so of reason such valuable. really communications scale, one And There's overhead is of which this computational and the bottlenecks, of a a when intense its becomes precious reason greatest company Mellanox kind. is workload jewel why a distribute

architecture in so that chip not second of world-class. software is of design. they're mean deep have of the is speed. we understanding supercomputers. of electronics link it's world's they're the the computing supercomputers. I And none. why software. combination, reason design, a distributed to mostly. in combination XXX% just and in a And large-scale why And And not just Mellanox and It's That's AI the XX% and their -- appreciation for about It's just that's

the and you're in just up. of talked absolutely I world And -- Now scale right. about

the reason out? is scale the why question that Now is for And this.

microservice these orchestrated well. world know be one, PyTorch; X application data This and would two over probably the Spark; towards important center. and center. you desire. you number And would across hyperscale across into of estimation across brand-new And workload X my TensorFlow applications they're center an it could all broken doing and a distributed applications, where however be small in -- is case could would and in the broken rank today, most be microservice disaggregated with up in movement so the you the reason entire up Kubernetes. why pieces it's number entire number three small are very the of a and data be it's And the Kubernetes, containers that application architecture a so orchestrated application the containers, distributed large where That The and pieces data type because orchestrate the is communications the networking the increasing so doing nodes reason center, center the of network between a of data that's offloading why the tremendously. well. becomes the increase they're CPUs, performance the you By the again. bottleneck compute throughput the data And by

they the been a had reason the It's year. And so per quarter. why reason they've last it's XX% quarter record growing why

back, stock whether companies, even low cloud incredibly their Ethernet their or makes latency, And them it's link they're unique, low into their latency in was really they're both hyperscale of integration InfiniBand the cases.

And from incredible. an a is so new the a data In forward, a incredible. utterly just like stack. software -- moving going It's The compute the fantastic that would world integrate we computing computing is writing as away data in developer And future, the lastly, -- just entire It's software node. of unit center. it's is fundamental would to runs the entire center. computing, so CPU the a innovate they're world on really unit, to piece this where then a we Cumulus, like

could application, application, an an an like it entire able And data would literally X write end innovate write from to in and from would networking end, You activate security. center. be to we storage, to that human world,

down secured can the Everything surface to to so attack reduce has be future nothing. practically we in that the

all And storage, networking way through compute, the all and performance we'd switch. then able incredibly all all is the And be low way to like to the second security entire the completely all incredibly high the across stack. so offloaded, are latency, to thing innovate

markets. the obsessed software supremely about stacks. that just know NVIDIA And for reason You software is because creates that is

it's or You or we're like can't create new care or edge or AI. whether health about, autonomous or computational markets driving conversational robotic AI talking recommender systems

All of that markets. requires create software stacks. It takes software to

They lot And platforms Cumulus they so data for and software obsession our into plays And and networking perfectly about our all pioneered, open open -- that. the they pioneered the partners, developer stack. creating of software-defined ways, a of in centers. ecosystem are

ability we top of in then end data center the end have the innovate bottom to super, so from from super to excited about the now and And And team. a entire stack. to world we're scale

Operator

with William from comes question Stein next Your SunTrust.

William Stein

focus to you on I'd Jensen, said. like something

where that. I your talk we of almost of of in where you of we of many to architecture and entirely, it accelerators CPUs evolution how where of now much Maybe are that in REXs are and position day terms interpret sort said say, responses, the sort part we relative the centers about number are goes architectural of you future curious and longer term, some about it with shift of you was we REX to earlier something If think those. going think us from where CPUs very sort to in can more think compute about that? maybe one let's to accelerated data a significant I'm smaller have a of GPUs.

Jensen Huang

computer Yes. I of And them that architecture and history who geeks there two computing are so the far, only has of is well history, for which it xXX. in architectures made entire follow you appreciate the is one question. that this, know people time,

computer ARM And And no there's in in computing company. you get as an it. And address. ARM any along. only platform program every came you one's reasonable an way. an xXX get every we're you accelerated The platform in in until we We're accelerated the computing if computer, really computer, other that today, We're fact, cloud. could you thing can such largely country. in computing every -- We're

years. XX-some-odd single a and this we then journey graphics CUDA. workstations on accelerate beginning since general it and long general at computing accelerated We on time. reached -- to is and with And company. learning focused computing to endeavor accelerated developers made And a C been to world, of scientific company, to an be applications accelerated XX-some-odd beyond Cg, been supported conclusion video appears years. I of was This year the a purpose The it. games we it's became Inside really computing starting million took every And and machine computer And have cost to growing quite our we've the to size, from developers made the We working that robotics. our X the computing purpose. point, we've that around support foregone graphics, a has number that almost for think this it well point we tipping exponentially. took for it's what

computing is data in It's think in pipeline the designing your put is accelerated much? do computing how you are what computing it. is people accelerated sense, use? centers established. The it? now How and so established. date do much NVIDIA-accelerated And And I accelerated common well computing well part who expect of to question you do

know of three and abundance the deep ignited -- the algorithm modern recognized the bang course, breakthrough, the as big And data. is now, one NVIDIA AI. AI, pillar, that the gigantic the well course, and two is And the big of of pillars other of learning modern we of

over so or book in today use you training -- largely web way because machine website could recommender or most or It's app And expanded the and shopping was news music came we learning for and the It's And the a most a earlier, or have computing video you a NVIDIA these or to video mentioned way ingredients app model it's important Internet lot -- of three time, as a a I I the I that is models. the single use only the important could together model and people training. system. use accelerated more that reasonable any way. anything. only But

Internet the the perspective. it engine consumer's from is And so of the

it could company they is the make way possibly no the there's On recommender the commerce. money. engine of Without perspective, system,

in core preferences their do. And accuracy everything predicting to so is user they

and RAPIDS, we NVLink focused spent the engine. And really did that that list just it, the big is all on gigantic You the just years a And X and data down analytics. which gigantic. of on Spark that Spark the is and reason on possible every why and data go stuff went and up made It work engine which company. all we from is of processing was part

the and The second learning models of inference. is of all training the deep the

importance grown the tremendously most are these tremendously, of of companies. of applications has and the So grown number because the its applications, footprint, has the computing applications accelerated the important

think that, I when world's is. now solidly that is so major And important said of workloads, acceleration the -- important I acceleration. the still growing, I the are require -- it But mentioned most workloads when most companies

really all that And so I'm to forward exciting just ramp for I mentioned. looking a for reasons the Ampere

Operator

John Crédit comes Pitzer Your Suisse. question next from with

John Pitzer

plus as ask sort think ones. minuses of I quick the guide, as two something also just like And raises. there's year COVID it but that. hate to you of I mundane unpack. mean in have Colette, Just probably some or given talked lot I some about you the full to OpEx, a

is just there's half you there's sort these to when it for of shelter a revenue thinking is markets? you're a certain for of and And hurt with of advantage year kind Or care. kind extra the well? opportunities? how there and all come actually the the of that makes curious, use then, of can as something in even thinking in I sense the those as just demand vertical structural that guess pent-up It of this you. then, maybe to sort how going to guide, me in And like about OpEx that? hyperscale take week understand COVID as we you're year of of try on activity the in of about accelerating COVID as one do could And think other look and guidepost place How out verticals kind health back quick we well. in we verticals, at Mellanox Jensen, But sort an I has in some course, I engagement? full asset. year I'm investing of HPC

Colette Kress

Thanks, the question. Okay. John, for

approximately incorporates the Let's $X.X at X OpEx full on Mellanox, first quarters non-GAAP We've employees. start guided and for from Yes, year. of perspective billion for the the overall that year. the the Mellanox its

to X,XXX about coming have board. close We on Mellanox employees

You are correct.

quarter pulled some has our employees. And We opportunities. this our -- a that excuse see forward is though, XXrd several We great focal as then months me, this this and We expect been outlined in to quarter, our investing lastly of -- you SEC are a should take in have little care filings, in well. week not by year. in bit that And we business. order

You've from to seen some investment, our and results more there's good do.

hiring and in investing are businesses. We those

are onset I there's this different of think, we'll long-term structurally, So together, nothing but just Mellanox investing great and results. produce

Jensen Huang

then these that as John, AI. computing know into X you opportunities, investing is usual, edge IT and cloud And AI. after largest opportunities the industry's we're And

hint? opportunities. that. thoughtful rate fairway usual, excellent well-managed. pretty to with question? are could But continue you I one And haven't... looking to of about we're was and what give on we're you NVIDIA's And and so count the we're do leadership Simona, me the as extraordinary down Could John's investment, team some managers, us just

John Pitzer

shelter place. hampered... levels Has engagement with of idea in the just in Just verticals, that

Jensen Huang

affected. Manufacturing of the a right. right. halt. -- Yes, industries have Yes, largely guidance. some saw automotive The our been mentioned Right. been to automotive few stopped. yes. Oh A has already industry you in We hasn't that grounded industry. And

way. automotive right the hunkered the back in they they'll question how that come come long, what expect the going And remain that nobody to, going automotive in It's future be moves. or back. it's is down now, not It's They extinct. not possible to to to capability but future everything down mind you not to of have autonomous going knows going my level but there's It's to XX% in vehicles. they're And autonomous absolutely that invest have no industry, of We to come quarter-to-quarter. be and to back. will

and drive responsible safety down could of that course, mostly a -- driving you. but mind. joy so be benefit, comfort to you're car it mind completely and the And and like Not without But coasting still road, the your for seems just your of because reading That it and a helping just you. seems just it nice of too. is what reading

I allows think vehicles companies going enjoying. certainty. And is to going software-defined have to love that The to economic future recognize technology their And is the that go so enjoy autonomous the after they of car the companies Tesla, companies. incredible a the People economics an future And of it. is pioneer, them be the to would are fleets. and base installed industry

And so they're going it. to go after

to -- industries And And so is going come back. this have aren't us. back. are retail impacted. is large sector confidence see, energy -- the -- to I'm those let's has come certain The is this going been well, that have And I for been sectors every impacted. There's

factories. focus the has way towards for rapidly figure industries in example, so into The BMW accelerating Like, their their to moving they're a the one in to impacted is the why out on manufacturing, They get impact hand, in robotics. some robotics is of which robotics. have reason obviously

So thing same retail. with

to robotic more you're to in in a in a robotic support retail, see logistics. lot going You're going see lot support warehouses, more

is so -- when leaders into really the market investing lean industry into disrupted the And future. the to it time, impacted, during the market when allows this and

And so when stronger back back, ever. they'll be coming they come than

Operator

next your Matt Cowen. And comes question Ramsay with from

Matthew Ramsay

of a It's all, different product. congrats Ampere. heck a on Jensen. first Two topics, Well, of

question... first The

Jensen Huang

so Thank I'm you, proud. Matt.

Matthew Ramsay

EX might bit the stack The to the and industry At moved Mellanox obviously one having now this of the around. Computex to on customize innovate first closed, deal been standard. second question And little about little talk talk is balance stay-at-home time, about but the an is and a canceled, hard it bit topic, the when demand. you have it's was maybe pending same opportunities that there's could a gaming

you helpful. there product, really launch And would gaming comments logistics? any think Just about how on be

Jensen Huang

Yes. a because backwards kind go Thanks Matt. your it's lot, appreciate questions. I cool. of I'll

face papers to each GTC so one I frankly, the developers analyst doing hundreds and miss It's hand, the really have. do the learned all the work And is much much their I engage that On just that fun. enjoyed we face. meeting of so time. are can't presented, we that I

million there's keynote been stuff from here's had right the kitchen thing. -- And of it kitchen amazing kitchen the incredible. so We video times. I GTC physical that just I the me, almost keynote. behind and GTC, And my did is viewed the almost about but missed kinds all the has X

our reach quite could think be I so great. And is --

great team, find know to people. we our to next, something that are And not going be you -- it. and reach going so amazing too to I'm and an just a going way marketing to to are got -- our gamers And gamers. we've we've see whenever They're -- markets going excited to launch really we our end be are customers got be

I'm so the Matt, I do before? what And we're never fine. very to going just should confident backwards. was question that do

Matthew Ramsay

customization versus standard industry of opportunity. Mellanox the Just

Jensen Huang

I see.

products Okay. closely architectural and as we of in synergies And Yes. synergies working together. mentioned. DGX include that that day you we Matt, that are development development Mellanox with announced product in. product GTC, -- the incredible, went a lot really that the There's see comes with years. X went so the lot The number a we could of I over of the and have work software on Mellanox,

frameworks the has All And before has every of accelerating center, ship the -- AI all data to our -- to to customers, in it. that From Isaac single at Clara ship AI of in because look to our the in you engineers. we our and Metropolis to it, ship to right? products to we Merlin If our it is reason to we for Jarvis AI own DGXs DRIVE we're industry. product company for

comes Tensor Ampere is with TFXX industry now PyTorch get to a brand-new format Tensor with with medium format Flow standard the in performance a NVIDIA we with incredible. it and Float come standard the -- And is medical comes And numerical Tensor TFXX. just frameworks. and called And And TFXX, integrated TF with fantastic -- standard had Float XX.

own so large-scale need And data center. we our

ship the ourselves. was started could we so customer quickly shipping as we And as customers. all we then the first And of to to

our our in You saw that in data supercomputer. center,

We Mellanox brand-new have state-of-the-art, is kilometers of mix. cables. almost gigabit XXX switches. And XXX one Ampere. it's And on of that per based Mellanox -- XX today, X,XXX supercomputers most second powerful cables, the fiber the -- world And XXX in optic and

single a together. of And is so we deal we announced a It's integrates call It virtualized. and it incredible Mellanox' telcos edge security XG work us there and for our AXXX. computing. integrates we it have it's between EGX and has designed of this great that first edge computer in did card, Dx, route Ampere trust, which and the We and Mellanox CX-X new

center cloud, so quite which box. it's remarkable. hyperconverged made this a a a The And AXXX xXX that inside incredible now really native, technology in AXXX, server processing, it's The capability secure, a turns of state-of-the-art cloud you is EGX a has EGX basically, -- cloud, into box. of entire is X standard that when into we completely AI it put it cloud server, computer

could you on. acceleration we together. working list piece synergies large-scale called done could the on Mellanox. goes Spark infrastructure the and possible the I so of computing. nickel software have and in the worked worked collaboration We different all on that made for have with We without together. networking distributor this see And They UCX. product on on It mean just

-- the the chute, of them. great possible culture great The out are we so all the saw synergies chemistry. I fit. working great And two right of product teams love because have you made of culture it's the -- And with a combination. that

Operator

all time call the remarks. is for back we questions. I'll That turn for the to Jensen closing have Huang

Jensen Huang

coming. It's

thank our initiatives. announcements, With several great me highlighted you. We and we a just busy quarter. Let -- had

data center hand gives scale of computing hand. networking The end Mellanox in innovate and to go expertise moving deep us from where to scale acquisition end. to is computing First, and

force the is technology AI of powerful time. most Second, our

is analytics, learning, leap deep XXx and data inference; throughput; the inference for Our Ampere in platform elastic. the fast, scale-out for going be data re-architect applications can it's and scale-up like that inference. universal accelerator center. machine largest learning, the generation acceleration offers several training breakthroughs. is applications it's like elastic modern ever first training and and generational It It's to unified training to Ampere configured first the

for up for robotics, forward and Merlin finally, of AI, DRIVE on intelligence, engines important have progress are conversational to recommender built RTX we from then center computing next multiple And systems. the graphics, for scale IoT, we our and look large data with artificial I new to as you care, such Thanks, computer Third, growth. health for opening framework, application accelerated software Isaac computing with markets edge. for cloud everybody. quarter. updating very AI autonomous edge for Metropolis Clara vehicles, Jarvis AERIAL XG

Operator

conference call. concludes This today's

disconnect. now may You

NVIDIA (NVDA) 22 May 202021 Q1 Earnings call transcript