NVIDIA (NVDA) 16 Aug 182019 Q2 Earnings call transcript

Good afternoon. My name is Kelsey, and I am your conference operator for today. Welcome to NVIDIA’s Financial Results Conference Call. All lines have been placed on mute. After the speakers’ remarks, there will be a question-and-answer period. you. Thank Instructions] [Operator the Relations, to to turn President Jankowski, of Investor begin call now Simona over Vice I’ll your conference.

recorded. afternoon, second XXXX. Executive Good call are Jensen call you. everyone, and to NVIDIA welcome Thank Chief to live President NVIDIA’s today Kress, our being for call from the that Colette and on and It’s you Investor me Vice Officer, quarter website. With fiscal Chief Financial Huang, being of webcast remind on President Executive Officer. NVIDIA’s the and conference Relations I’d like is also

can until a available The is NVIDIA’s You content prior hear conference financial XX, without a can’t until quarter It replay property. the telephone replay call be The reproduced August of will for today’s discuss or XXXX. transcribed our by written results of webcast call the be to XXXX. consent. for third fiscal

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

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

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

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

strong by tell XX% Datacenter, revenue earlier. reached quarter. in $X.XX is and growth, This for record first, led up platform, more We the year-on-year. QX Gaming, Visualization, sequentially biggest Datacenter a than We can’t Pro let’s talk Automotive strong billion, another platforms, over it. the more decade. We year hit announced leap a both architecture market These Simona. quarter, in about with a year-on-year. big wait from and had Gaming. you NVIDIA. GPU about collectively to grew Thanks, just and But week XX% Each levels

quarter. outlook nearly revenue million, both $XXX $XX and declining cryptocurrency-specific had while crypto-specific a non-GAAP forward. $XXX exceeded anticipated to a going and XXX year consecutive reporting Our GAAP product expect doubled. nearly revenue year-on-year, Processor for revenue XX% points grew million. negligible third products income now million, basis while billion. was margins to grew segment From net perspective, to contribution XX% the we Profit Gross GPU grew actual last revenue $X approximately billion Tegra $X.XX from

The over market. and has a The billion up our remains right the up approaches of XX industry desktop, business expanded based was we is gaming just for popularized $X.X holidays. high-performance stores, past Notebooks this up consoles year-on-year. million, Revenue PUBG XX ODMs a were Fortnite gaming with major genre, for sequentially. have our of notebook in strong all was expect year. for success models, on with vibrant. double-digit pack OEMs quarter start Max-Q form millimeters by driven of unprecedented Max-Q and new enables X% less. and notebooks, line factors, notebook the see top on Battle the or gaming The for and to XXX in audience PC GPU with strong that thin standout XX% this time demands all Max-Q, All notebook, OEMs adopted technology. percentages Max-Q thick, just And gaming business. now and year-on-year to their gaming eSports Royale based Gaming XX% into Growth Let’s the slim up back-to-school. segments of and the

coming fact, Battle Battlefield much the like X. mode anticipated is to games Royale In the

AI Battlefield upcoming Raider. to partnered It’s Hunter excited week our for of platform for new this at lined since the be includes earlier community is titles and The Square Shadow more real-time GeForce is just most thrilled October. CUDA partner invention that gaming not of platform architecture are first make We’ve World PCs hit. was Turing this in arrived best processors a for its and dedicated Tensor to promises on important of are with We ray-tracing month, make RT titles GPU, over Tomb the to innovation X are with EA it PC many Enix building make SIGGRAPH. inferencing to The ray-tracing architecture, GeForce instant for also big new the the which Cores the an will for season. release the earlier time. a up the the new, possible holiday best of gaming announced decade together or Cores, anticipation. ago. Monster And Turing what

and the architecture our for neural engineering enable increase up This games highly the graphics of will peak fluid across of powered XX,XXX of look Pascal boost forward. performance delivers new amazing gaming, products We complex video swaps. be to cinematic is portfolio foundation Turing up the result ray-traced platforms and Xx to up industries open and models. XXx GPUs. visual going on networks XXX over interactivity Turing than effects years billion of of to the quality by effects will will new new inference the more for It effort. reset

increasingly to $XXX X% This hyperscale demand quarter driven up daily by was strong We people performance Moving sequentially. as used with to services internet leverage billions by and of of Datacenter. accelerating had growth revenue XX% year-on-year million, AI. another

voice Our GPUs such applications. power real-time retail detection, translation, fraud and recommender synthesis, search, as services recognition, voice engines,

ranging financial of transportation AI Companies AI our harnessing oil growing solutions our of to our data also from sectors platform areas the computing to vertical insights. gas in and representing accelerated power by and one to fastest and We the saw growth are high-performance actionable industries, adoption computing of and turn most of business. into services

Major also expand which gig our in should to Volta Tesla support traction sets the both quarter. Our Cray flagship continued operator Supermicro continue gig much addressable to the XX have the Tensor ramp to AI Lenovo, XX gain been Core market in Volta by larger networks. and neural world. architecture to market on applications. new data the We and helped inference every quickly Enterprise, adopted around and hyperscaledatacenter moved to OEMs, and version of GPU, VXXX IBM, for the solution qualify computing DRAM to based HP on-chip provider has datacenter. which high-performance Customers VXXX, cloud brought major the AI with the server VXXX, doubled

availability. During released inference the X software for accelerator AI we quarter, general our TensorRT

to software the and in annual means, on being portion integrated DGX, SDKs optimized workloads, inferenceand offerings inference to fully and such NetApp into of At from the in dramatic available Tesla In now NVIDIA engineers -- days VXXX X a speed deep developer recommendation hours, AI which versions as HGX-X stack. learning month deep well PX speed we TensorRT the track inference aperture framework, platform. applications. partnerships graphic bringing our customers company’s influence. TensorRT revealed From Key This that our rate TensorRT to growth optimized we our Google hundreds prior in have ecosystem is recognition, expands is our speed-up address image third At we small Cloud workloads, weeks, of speech such gains learning GTC is Pure for they that announced high virtualization to the inference optimized of we time. interconnect making DGX-X TensorFlow their to delivering Storage stack. highly-integrated, CPUs. incorporates announced on factor for to training Google CUDA and And server announced fully GPU, period with Taiwan, even accelerate include by a own our form ramp performance up third-party and GPUs, larger GPU AI and Technology is our can dollars. to DGX driven systems. partners, that storage of Datacenter proprietary run was available optimized high-speed NVIDIA our The more speech It is achieve set synthesis, to are short for and to While at GPUs. also we Cloud video-related GTC are getting performance server strategy as easier to to or We and the software inference Platform. OEM our and all-flash in run recent deploy of with our AI the technology on the developers that in our Taiwan June, our stronger. we HGX-X use Google our by innovations relative designs. qualified so XXXx it GPU translation, such and customers of is in much millions training DGX networking. is records their cases make quarter. our DGX-X, And March can Conference, to combine also AI as reason announced work deployment newest solutions AI quickly ODM server help with with we and can

In program, members ago. our year fact, one we developer X from million XX% up just passed in

XX,XXX National scientific in the Lawrence the combine TOPXXX majority five of the added the now moments With Core GPU years. GPUs, Core Tensor fastest Volta the TOPXXX comes of ABCI more shift and at the Sierra supercomputing power on in XXX AI Ridge GPUs number helped can genomics from spot Tensor to launch spot, systems U.S. molecular the dynamics to our HPC now list seismic Livermore NVIDIA supercomputer proudest the list applications the world’s One over powers from processing third list accelerated. of Indeed, to simulation in to of many our Laboratory. time supercomputers, and now first to than is the for this are and latest the the supercomputer was with the were reclaim fastest by National Laboratory broad supercomputer advance Summit joined power GPUs. in of GPUs in spot quarter science. one the NVIDIA TOPXXX NVIDIA fifth Other applications material the seven Japan’s Summit computing the five supercomputers reflecting powered AI and Oak performance

$XXX applications XX% XXXX media rendering Revenue key These GPUs, the SIGGRAPH, and real-time including VR. several XX% Pro to now like at industries architecture workstations, and Visualization. sales. the Lockheed, it’s as Strength Turing-based we driven across the grew and healthcare, represent extended well year-over-year also as Healthcare. Phillips to AI Pro before of and Turing oil the and wins Moving sequentially, for mobile in million, announcing emerging processors, been bringing Key by XX% quarter Siemens applications to and RTX demand predicted. and approximately gas, up introduced Visualization GE, include first ray-tracing Raytheon, and years interactive XXXX, entertainment. Quadro world, In emerging XXXX

to the NVIDIA give We XX Turing ends. also firms rendering RTX have in render work Server, them GPUs scenes million already time add is as developers Turing. announced a artists, world’s capabilities for -- available to the the ray-tracing a will support Law that providers, to giant of Private the workflows. boost on photorealistic set and software that revolutionized of X based global full and Moore’s enabling and render server designers leading illumination fourth OEMs quarter. real will be in the Dozens to expressed new Turing AI-based

media a game-changer and as view professionals for Our the entertainment, architecture it industries. and in ProViz manufacturing partners

ramp to smart XX% production as turning around growth Revenue sequentially. the AI-based, million, in globe, year-over-year well This next-generation our was a up development cockpit XX% and of Finally, vehicle reflects record and infotainment the solutions. engagements up autonomous as $XXX Automotive.

global autonomous for We hardware world’s brain make begin brand the systems level the synonymous brings largest on supplier, this Valley. in X key with and expertise partnerships next leadership vehicle their X fleets. as Bosch quality. safety vehicle expertise as progress and Daimler NVIDIA’s continue and collaboration In will DRIVE and quarter. selected and level July and tier Daimler’s with Silicon self-driving milestones X together and to autonomous year Pegasus Bosch’s our announced platform and automotive testing Pilot This platforms, in AI AI

achieve software, ASIL two and two designed level safety. AI redundant diverse the to an and higher as is and This to platform processing, next-generation and visual industry’s delivers quarter, automotive for well This and SoCs data compared highest automotive Pegasus, vehicles. autonomous per previous generation. and With as safe D safety created trillion NVIDIA designed than operations delivering production Pegasus shipping we to more GPUs, DRIVE greater for second XXx specifically hardware bandwidth grade, the XXx uses started development for architected algorithms, the AI functionally handle Xavier ISO systems co-designed for performance performance. XXX XXXXX, solution supercomputer functional of

platform range the More this We we level XXX institutions about pilots autonomous entire business. are using have long-term spans autonomous car NVIDIA’s revenue jam and research excited intermediate driving the traffic and opportunities that With created scalable from a than AI of X and remain for driving, companies to automated accelerating robotaxis. automotive momentum growing growth, platform. and

and manufacturing, applications bulb. This our at it introduced powering energy or more Jetson Jetson delivers AI the per quarter, With a machines retail, more. robots operations autonomous second, edge, agricultural, than drones of X billion in healthcare also powerful like trillion with processing platform the we establishes computing a than machine for capability one-third light Xavier for workstation autonomous using market. deliver Xavier over to logistics, customers transistors, the XX or more

our in sharp driven by revenues fairly of XX% This cryptocurrency was levels. declined revenue the year-on-year minimal to OEM by XX% sequentially. Lastly, and segment, decline primarily

long-term Moving were non-GAAP to in expenses million. was expenses up the were We $X.X operating year-on-year. million, P&L. XX.X%, earlier. was GAAP was GAAP the growth $XXX rest from respectively, XX.X% Automotive. up to XX% GAAP QX and with outlook. can of income year billion $XXX $X.XX, gross our key operating Gaming, line XX%, EPS and a margin the Non-GAAP and our AI invest was continue in driving platforms, and XX% net including

from outlook was of a ago, and net me XX% revenue Non-GAAP XXXX. was X% to XX%, third the fiscal strength, cash as the capital for and flow year $XXX Quarterly driven expenditures of was well Some to margin million, operating billion income from up let was turn as $X.XX XX%. that, near and and gross of compared EPS a taxes. expansion upside our rate tax the $X.XX, $XXX respectively, by million. operations were lower quarter outlook reflecting With

in our no including are crypto contribution We from outlook.

be XX.X%, respectively. billion, to included We GAAP approximately are non-GAAP expect excluding to discrete be Capital income in expected Commentary $XXX and $XXX GAAP website. are X%, plus are In expenses financial minus or $X.XX are non-GAAP and expected approximately are be margins plus to million to minus both of some community. closing, X%. Further expenditures tax to the or points. $XXX of respectively, the rates and million. revenue operating expected the OI&E are financial and expected and or million. GAAP items. both information X%, upcoming minus plus to and CFO million, events to expected to basis I’d GAAP available $XXX million highlight XX.X% IR be gross like details XX $XX other non-GAAP be on be non-GAAP our and for

our XXth. the Citi earnings take conferences next the XXXX, on Global We and September quarter at results technology the will in meeting be Technology with GPU presenting third and discuss our And place Munich November is of on on our October XX. financial on XXth will to September community in Conference Xth at Tokyo financial call

We call the will open for questions. now

two? could you questions Thank If please your you. operator, you to or for poll limit And one would questions?

Instructions] [Operator Mark Lipacis Jefferies. Your first question from Yes. comes with

question is To extent The Thanks. in markets? creating your ray-tracing. versus this enabling on greater markets new existing capabilities is what

Mark. Yes,

And it and to first experience reinvent So, benefits completes four for basic to rasterization, with -- unites There’s methods graphics that know, altogether. two the which first different right-away. intelligence. modes ways create major these graphics imagery computing computer ray-tracing, the It uses the is platform, artificial going of accelerated we’ll different is CUDA, two GPU. future. new as four you computing world’s ray-tracing It Turing, all, our computer

visualization markets way. require is and now it that these is And is markets based it or design, or lighting. wasn’t materials that for way of the long able the Whether The first car rather images. achieve And product a been it’s technology or physically today, engineering, has IKEA The a really very Catalog they photorealistic achieve computing types time. of movie, complicated, productive intensive been photorealistic we’ve to images. use with require architectural in ray-tracing to of only all for until to design,

accelerated also has these large it are Turing buffers, ability the and sets frame to ray-tracing, has so, extremely ray-tracing, combine the And very because to data ability do large.

is these marketplace service and it’s movie gigantic large, and so now, that served on has never CPU and farms been so, quite and GPUs centers been And until run before, all of on that in farms, render forth. by so studios all render

The see second we announced. ray-tracing, the benefits you’re area where haven’t of going to

industry that Where could you Thank have the the capability? is I kind you. If a do follow-up of content on on side. gaming that think creating leverages Okay.

of spectacular. first that one the Yes, GPUs. four Mark. support in those the on brand the DirectX -- and world’s the time that companies, And They And game the get Volta images. SIGGRAPH, intention in at This platform out we first call developers the last platform GTC, -- GPUs, with of a RTX. application Engine. to the RTX at running throughout demonstrated the And this major developers. Volta computation really Quadro, been the to ability methods RTX the working implemented engine GPUsrunning of year we was this same described, the March. do week, Epic has Microsoft. announced XXXX, it accelerated And time. demonstrated platform four Quadro real-time - four Raytracing. game March that the generating very At and the four and into, and And is this NVIDIA I on performance to for ray-tracing announced put we GTC has with And platform out this ray-tracing we Quadro called GDC the it demonstrated new for XXXX which Microsoft GTC, We first XXXX, game is that. GPUs, we’ve I Volta our Unreal on GDC And all

to I the going is, the answer question to use to in popular start have RTX. so, your and It’s the game engine access it’s think see you’re And most in developers in Microsoft’s all DirectX; developers it. world; to

workstation full the there see have SIGGRAPH On of year developers Visualization of bunch side, NVIDIA could adopt images. side, whole jumped accelerated And ray-tracing, it. with major to Professional generating the this there, a the you on at RTX realistic all on the demonstrating ISVs

of would had I stories And onto day about developers And more history, RTX. got in so, you one a say many tuned. jump lot so it. no on announcement, has that platform, our stay to We’ve tell

Your next Ramsay question is with Matt from Cowen.

contributions might channel, by Thank you products? business is, you about the questions which of period And as little inventory help I Datacenter transition? you of for product inventory walk and about secondly, set a channel, really I in or ramping the strongly. future so. a balance are a X% very mean up things the of is much. guiding revenue you bit I launched But, and see that if Maybe at sort you the the is look what could it, of ProViz, terms talk up gaming could this during inventory, us inventory, through it you. Colette, understand and if in you’ve product you’re Thank mid-XX% of guys how the looking obviously of had first the sequentially couple level new think in sheet,

your for Sure. questions. Thanks

with our piece it seen have extremely on time. terms and think when know, So, last what market. some you bringing over look important several sheet, inventory we Turing us what as of consistent the will with be balance to it’s months architecture, the generally in for you at an of is will you be I

balance is also in cases terms we So, in complex ready our getting I there. of very we in in forget, work terms and is Volta don’t And think inventory have a very, of for for Datacenter that. what computer what the terms have also of some

what makes plus up those have still architecture things just almost Pascal inventory. there all So, we of together, is in of here, terms our

on back-to-school see side, the that that’s our and we of building the inventory season well-positioned for end the QX. stack. is Matt, channel in inventory inventory lower on up coming And

that. our the really so, ramp-up I rest good of Turing feel launches The And going is and of pretty about product well.

but And so, The the to so the of family tuned. is for announcements RTX haven’t reinvention a developers. And embraced we altogether many I rest think computer be game-changer graphics made, by has the real us. stay of been going

stuff to year. going really exciting this some We’re see

Next Arya Vivek of Bank question with is from America.

and question. Actually, then the clarification, just a

QX On remarks, on the the to a your could you Pascal Jensen, relative clarification, advancement But, sequential absolute over to Colette even before. from think This as very mentioned you an gross contrast what and Because, very you the you Turing cycle is also help launched then, And Pascal, you the not though I us had the Turing with if QXs cycle? would in basis, then how and understand margin QX? QX strong strong. strong a QXs. perhaps it’s you it’s had guided basis, when time, on outlook, a sequencing good

just cycle us cycle? contrast you the the help could Pascal with expect with we So, what Turing should

regarding question Let with gross your first margins. me start

as We to our believe do into as believe, margins. period, that normalization essentially normal we’re several we the crypto gross forward. have to last overall we what essentially reached over margins. and quarters, move we’ve as we of QX, forward seen looking elevate of I a can have no reached, impact our We the gross move cryptocurrency

and programmable was same. fantastic. successful. There’s XX% Maxwell. to were in to it shading. same the much, Turing GPU really think reinvention something a Pascal significant see. was efficient first more of of upgrade. programmable computer largely to more and Pascal Let’s comparison simply like XX%, I efficient, much the architectures relative just fact, a They no They The was leap really But were graphics; it looks And in and to and ray things and energy is energy The and global subtle and an reflections model the areal world; be photorealistic The environment that able it’s able illumination be at new so movie. in extends is which capabilities like create accelerated GPU you we’ve artificial lights added trace And before and indirect everything what to light are to just like ray-tracing new fantastic. beautiful, both we’ve Pascal, performance the going both than to that hybrid a ray-tracing images done. it lighting. it that to Turing. translated success was model with just customers. when backwards at but it’s was the shading. shadows built be benefits is intelligence was yet two will first This a look Maxwell, rendering generation compatible, it so it, were Pascal

the So, possible and and everything of create past to brought that will visuals going benefits, it’s along be new before. weren’t

job a world. ProViz for developers. by partnered We foundations We did renderers and all DXR, coming, to that also the RT create Vulkan is with good used the OptiX and have the we on Microsoft are developers platform laying also of development over the

out, will we data so, a have have comes result, stack going as the to foundation the over we’re the of by Turing And gamers richness benefit that to laying that be stack of years. stack And on able by applications the enjoy.

half. rest You from we contribution at mentioned beginning an we unusual this a it actually think and consider ramp the the several be second of Turing. thought that the excited year. performance, time, that And the lift that terrific. crypto. through we In to the in crypto But, doing be immaterial year, larger super about would the for last of projected the quarters of I’m we guidance. case on year-over-year It benefited a is we’re from I

automation Every has ever basis to sequential is basis visualize business, on guess, now designing to the up a makes harder. a adoption -- resetting a to comparisons platforms reignite growing, the to brought to bring is us and is and And first to that with another video terrific. one time, I productivity we’re going about year-to-year on to to Professional from by our be products Visualization The so, render to quarterly But, it everybody that’s single basis, nobody of And graphics are doing that AI, industry one has to think growing. like open gains we’re Datacenters photorealistic realistically on Turing, the high-performance of who going AI, able rendering for for I be seep course, computing industry. seen before. photo sequential industries farms, games. reinventing very going continues

think we’re to great in comes. And forward so, the I a QX, reporting and looking position, when time I’m

next question Atif Malik Citi. with is Your from

and Datacenter. verticals, Colette, computing In of AI talked are some growing. your about new question you remarks, these I high-performance driving on prepared have and a verticals fastest

Some of your peers spending are back the of an -- about this have and demand, not down of units in guys enterprise half you AI on unit more play slowing adoption. sever year talked

terms thinking in curious your Just half of second about Datacenter growth.

you as out. So, quarter know, on one guidance give for we view our generally

coming that of of points computing needs the that that is every essentially industry-by-industry seeing. unique terms the are of data workloads a will for underlying hyperscales, correct likely always continued essential. in the You because by continue. so mix on-board, we of results that tremendous, in single is just our The have that use accelerated for they Datacenter the some quarter terms see growth still and what we’re But, there’s

QX Datacenter to for sequentially as our selling both into go mix a of contribution but Tesla VXXX year-over-year. and platforms, probably And we’ll also a expect good still we see we grow, both, from So, DGX.

Yes. little Atif, me That’s a right. let that. to bit more add just

double transportation let’s particular are the transform are very industry, is that abundantly one industry. think to and that very, think, -- vertical, the happening will that dynamics that this. way take two I are one clear, In about simple there that

in an is use match. in and optimal problem; learning which perfect large ride out which that sharing. every make need is it’s one most to ride at of a passenger, recommendation platforms which machine problem to really of optimization computing a to course of problem. taxi Those high-performance first The It’s very, in hailing to to or order -- match, customer, to a large very the learning machine and instances, computers And make every each those scale. you figure how bring

your network, your car cars, self-driving robot second neural to or whole is cars. on collect label on train working needs taxis and or those bunch company on that’s and run of a data, networks Every cars. networks, train The self-driving single data,

to people more just stage every simulate accelerated them or because the travels of the could what need stack. cars. drive normal to self-driving you test the know software miles XX world really reality. possibly the is even that your we want make And model. be do GPU so, in stress And is next our And many do million do is how miles. building The best able servers. the virtual several trillion will to simulate to single software list to just we And one to are and only the year. to developing entire industry way for that per is do that’s And we of actually stress, And

that and over of And another that’s across to to so, you make billions the billions have virtually software those to you the fleet no you a collected the re-sim have cars. for created OTA your years software, before you’re models and sure OTA, the challenging of going all before regressions against then into you’ve of miles have miles. new build that replay And simulating all to supercomputer lastly, you that

AI imaging intelligence, learning large using very genomics everywhere Healthcare pathology. way, deep from future artificial The on. the discovered And benefits going that that goes be to industry. a now just so, the is medical about list transportation is same has the in to of and and is

industries deep and be by could after benefits industry the the after learning, them. discovering really, industry so, we’re really industry, And revolutionized of

Your with Evercore ISI. Muse is from C.J. next question

term. a short-term guess, I and long

Volta particularly think embedding short-term, Cores? then, Thank about as about integer differentiation drawdown your a for Can Tensor you are about talk there for about longer term, you versus of bit inferencing? think gaming X-bit opportunities as there? the as channel So, inventory you VXXX, think you and you any And Turing guide, you.

the work channel We’re expecting out. itself inventory to

We are channel we managing well. masters channel, at and our understand the very

As the through channels way world. you the know, go we the is around to that market

inventory. We’re not concerned about the channel

a top As ramp new whenever we ramp architecture, we we it Turing, down. from the ramp

back to plenty the the to inventory. have we so, back go opportunities of And cycle the as manage gaming in school to we

good feel we about So, pretty that.

applications it to the is the the andTuring, the of cuDNN, CUDA, As that written of comparing CUDA. our optimize platform take that’s a which advantage benefits iscompatible, CUDA that advantage that of of result, are takes TensorRT network takes and CUDA of top frameworks runtime. for Volta -- of on entering, all one output

run of on What Turing run top same run of adds is those the on top Volta. top libraries Corethat over of is All of Pascal. Pascal and on Turing tools and Voltaand Tensor inside

capability the Of And tools applications, course, be could connected double-precision, datacenter applications. perfect three make XX-bit large and ECC, high-performance accelerator. for computing GPUs system case training. software Voltathe for that HBMX has Volta it’s is have of designed of it’s together. really for high-resilience the all software They the fastest designed and major computing, memories. In Eight designed and scale Turing,

really Visualization, data farms. up sets. used very, really unable that path open The ability -- GPUs is very a now to first we’re And with render is which to have Pro the use full do application global large large we have historically has illumination market to we until trace, now

new brand that’s as a market that’s result of So, one Turing.

to images Turing of computer for you’re going created centers. to see the applications. the reinvent third, market is It’s graphics time and Tensor amazing. When data really Tensor video a you It the by real inferencing wanting graphics, computer hard images throughput, just then for looks really image Turing, have games used high for past. visualization other time for Core. Core time second supercharged see has this And real to deep learning is And used the generation. a also The

of multiple there suggest are architecture have could whole applications at is a time. scale one these platforms one reasons great that team, we which one engineering so, a why we lot of across would the And for Turing, such of Turing SKUs

inference going answers Core the be off that of Turing to Tensor that your hope I capability is the question charts. so, And

Moore Joe Morgan from is question Next with Stanley.

Great. could Thank you. you I wonder if cryptocurrency. about talk

-- backwards, on of Now the it driven quarter? impact I by that business And you guess dust there’s sense that was settled, guidance guys GeForce crypto. job that being business in the October, of by has some it of if given have been the size you understand July I the but crypto to Can was am for the would very small trying looking for the -- can us? some seems have driven laying the of think done of also exactly much how is I OEM good -- a out you kind that been

just Well, we’re second do I first -- expecting used to hard the could is sleeping, -- It’s matter mining. they’re ambiguous. think the while could’ve to for they lot the they is we’re It’s first projecting And the what. predict, answer -- been second night, A of gamers no estimate, crypto? question, the the some easier basically. reason zero question one could question, answer But, hard anyway. much GeForce how and for to at of is

for they hard it’s say. buy they it it mining do And or for gaming, kind of so, did buy to

they it and from several QX, And retail, to unable and last some and well. And our going QX, that crypto-mining the all buy great products, last quarters, that and we’re very previous buy happened no -- the miners happened quarter, little in probably to as market QX, so probably from a jumped OEM deal starting were forward. late on the projecting to

Sachs. from is with Goldman Toshiya Hari question next Your

guide? you on can bit, about you of of know set and I if if are had embedded in one broader about for -- success is years, hoping within how you Datacenter, Colette. to us other your talk little segments. how talk over Datacenter the about for across next give I can at meaningful Jensen And you And talked customers? business Google that guidance specifically, gaming come products growth a or the two for as inference then, platforms. expect you was you your as remind for Colette, Jensen, not could accounts would your directional you a I today new whether proliferate like about each if and one can

to very business. inferencing world, There’re CNNs have there is best approach that the reason really cloud, for networks output going these couple frameworks take surely complicated. in is years. have inference three everywhere. training kind It be and foundations the are at network networks computational ever And problem the optimize be are a accelerated. into acceleration and This And that that forward. large and intelligence segment many described has for probably this to We’ve be for territory. million and now the output to of continuing massive, going are estimate, neural neural And of seen. going different been whole is large is future GTCs, world are lot optimization is for the just more has for And Inference infused two, the in enterprises. very graph from material models There RCNNs, is and really, It to to RNNs the you or as I XX grow. in to be encoders so to auto I so Datacenter is a we the our brand-new going but laying invention massive that it. largest because a of a net and believe the And architectures days prediction the Datacenter reason the species many to LSTMs, around different software of these software that of our us. to deep is that’s it’s business. believe servers segment, largest market not is almost artificial in and going models, learning proven are every be to will is server there

complicated. datacenters. recognition over systems, well compiler. us we announced from accelerator speech to image all optimizing the into language of compiler support all applications And we the beyond really year, and And now are high-energy-efficiency the allows recommendation neural taking compiler, beginning, optimize And Tesla goes announced these for X natural optimizing really recognition, and this successful understanding, voice hyperscale TensorRT services translation. year, CNNs recognition and internet is in And TRT so this pervasive And network or world. our the we really, two in Earlier been generation low-profile, inference technology fourth we’ve things. that announced PX,

now now so, images have to recommendation to address voice video a from we compiler that And to it. can systems,

for that. last the just We platform only of Turing. this announced the in provider they this really excellent hundred single into to are couple of And relatively it importantly, And the need in I need is the acceleration just for performance really, if actively latency. the is period short a is times And performance about then, we times which inference of of that world and useful that Pascal, performance inference already with incorporate reason our every is CPUs. time. is low XX the week And very week, throughput inference stack. a Turing Voice working their high internet recognition service announced of because response inference

output our a and which neural at advancing at we’re take rate we’re optimization new of performance which so, high. ever-increasing processors, at support the moving look we’re think I the compilers, the in on, the you both and of raising the networks, pretty the And bar rate

Okay.

So, with that, Colette?

of as So, possibly single seen in you’ve every QX QX, in our platforms We overall the those each when of you last our probably standpoint. across one across that look our at year-over-year and see a again one segments, results year-over-year our every from growth guidance, growth platforms. there even is this terms of

we’ve Of course, will our our we those of our and forecast. many likely as hopes still sequentially, It’s see our OEM But on about business the different absence is cryptocurrency are Datacenter definitely scenarios year-over-year, Auto. and think sequentially. business will will to growth ProViz be grow well. to in down due absolutely When early, just of Gaming our expected our Datacenter got again grow and our likely Gaming our still

with Your next from question Blayne Curtis is Barclays.

October Two was make from in July. I Colette, pretty to it I gross you’ve down. crypto understood I but margin. minimis want getting on to just a gross July know de margins been sure benefit

couple enabling of Turing other margins? curious how to ramp kind margin paid pieces? here, of to think gross how it, longer quarters XX stable. the kind affecting of that So, nanometers, lot a any picture gross just capabilities the with think You’re for of then, over obviously do you And get of And there is ramp? about than about fairly

Yes.

what question part So, gross let margins first of regarding and your seen from the take we me have crypto. our

the and still may has Although terms impact be a of large, not derivative to crypto our and channel we selling what it overall such. stack in both revenue are on replenish

we great terms and have of growth the for did really of that. the we XXX Again, just get overall stabilizing year-over-year that we able everything you quarters, growth. at So, our benefit had last channel, overall selling effect from to points about look margins when that over in QX, basis the several been

have here year-over-year. well, also significant excited what We’re which about now we QX for growth as is

platforms that Quadro piece in have value our of terms well. as terms But as in move expect those for our some that we partake. all added both we will time high take Turing in the those of forward, to we effects what course, on Datacenter, -- Of of

So, we’ll see how that goes.

announced Turing haven’t longer over we’ll probably yes, do. anything term, see We time. the of But, can the effects further this at

Wells from Aaron with Fargo. Rakers is question Next

as kind us data the top guess or we that’s on between over question. hyperscale, business, help center And seen just can piece. I if pretty of related of of the last curious, the AI understand look really is the that’s at breakdown significant demand the that, follow-up growth and driving of remarkable piece Thank business you you. I’m the curious, metrics if what’s that, of I’m a curious, quarters you’ve business, I’m supercomputing as one couple in Datacenter the the to China.

Aaron. Yes,

XXx at at from to every if computing start grow the continuing continues computing Moore’s Computing principles, years. to every the demand. Demand approximately at look at years historical Law. -- at levels demand five every years. you first is five And look here’s to way levels it. XXx demand think, simple XXx historical that is of XXx five grow I if increasing is of you

However, Moore’s Law stopped.

gap in in that gap more accelerated served be GPU intelligence, And benefit life in pioneered, artificial those that. that that to really can computing, applications the gap stands -- demand imaging, computing because we only capability, in videos from another way. world so, in high-performance NVIDIA’s computing, that computing And sciences in medical

computing and after demand mentioned have new came because jumped into -- from the the the continues GPUs level, developers is GPU of computing highest vertical my world’s that expectation the that XXx and Collette that another, NVIDIA’s of that TOPXXX, it’s year the to transportation, earlier TOPXXX as of at XX% you more on supercomputing, on one go to every at industry reflects healthcare, computing. demand. heard years, so, And on, The and all performance I and whether this And I logical it boost called say rational future more TOPXXX. industries their is represented the five NVIDIA vertical mentioned and are a factor

think probably the I way that’s to answer best it.

Sur Your next Harlan from Morgan. question with J.P. is

there designing to Good Tesla-based Thanks question. their taking your we products think using and afternoon. hyperscale, When the own their for right we or top even guys for AI hyperscale and platform and cloud smaller deep learning. tend about where to customers of don’t who your sometimes think larger my think platforms, I cloud chips designing about But, own out there, that’s a on. HGX to have base there’s be R&D medium focused seems scale. And

HGX-X? the So, platform Jensen, first can you on HGX-X uptake just interest your generation the of give us initial on reference and update an Thanks.

well, doing an guess I the of can’t for kind prototype talking about. you the the is reasons scale afford incredibly really these that HGX-X largest is centers It motherboards hyperscale complicated data at we’re create HGX-X. to of that even mentioned. was all and the HGX-X

GTC HGX-X, at was take leading of Taiwan, adopted server HGX-X created so, we And in basically market. it it world. and and And to we were announced we most the hyperscalers and and OEMs are the supporting by to ready important several immediately ODMs all

process production. we’re into in So, the of and finishing ramping HGX-X

for PCs standard that you the be We it motherboard for complicated success a exactly fantastic ATX I just to and used motherboard, design. think high-speed and use for high-performance for really could way. easy datacenters, mentioned. be for a really could And huge for could really this so, It a really that it HGX-X be people a is it’s and interconnect allows the like essentially used could adopt hyper it scalars, used reasons motherboard in HPC,

UBS. with Your next question is from Tim Arcuri

you. Actually, for you. questions, Thank I Jensen, both had two

and I or platform? you resold Is that something two, potential comments away Thank talk cards that number did now could call, Pascal And that know eBay for the about some cannibalize see the stories that to First, has your and Crypto think Gaming keeps Datacenter. you other night? fallen of that one. get own what Number Tesla on talk channel curious is maybe that slug their from and since we the what curious you on sales. off, typically obviously, new efforts a that I’m up are don’t move customers. DRIVE development you. X about your about But, their And I’m about you Hardware at

though, the new And is you see three Crypto years still be it The used. much market new for was to continue. Sure. can Well, mining -- And rates the being it existing even doesn’t cards is different the capacity cards the current into sense very make market. at today hash sold prices, than ago. mining

important will my of so, the Turing then, base and of factor And though of process new way miners announcing to cards. with in installed will use RTX the we’re the is that doing that probably same. continue is platform, more that never a computer graphics. And the graphics brand sense their our And computer be

And in GPU started is that computers. first drive to we more that now. I have greater I those appreciate is Xavier. we those order more so, and quickly, require or the take when more of of that to be we things computing to Tesla vehicles, also to generation a to I system, why they a have new And handle is autopilot lot that that great it case you three-year-old generation, happily. really or Xavier working the next autopilot new In needed great. there’s think safe and generation current clear also you do the is corner Company, more on now exactly time it I -- and driving, computing with smoothly more of very in to that’s more safe a a think have them all corner -- I the And able makes there’s requirements lot And help. cars, -- have -- Pascal our be And our subtle think or Elon’s situations. to to our about the conditions autonomous computing for used and very diverse algorithms respect rich safe to a has reason to GPUs And have production need turns order every order going comments have able in of conditions of drive horsepower. be in all to built we capability.

of and top out doesn’t the it software them, it. about turn they Xavier reason doesn’t I if a excited success And on I’d a than we’ve it. be turn for and for customers it stack Xavier. why whatever exactly great And a to all super-hard super can reasons, out me seeing build to We’re And and built happy give more are that’s help. call, the think

have I time. closing for turn will to we now remarks. out the of back unfortunately any And call Jensen run over

had a quarter. great We

launched week. today. reinventing guys industry platforms the -- Gaming, See a continued expectations, the Turing stack, we you was NVIDIA even of world Company Each of -- crypto RTX the since decade as challenge, with core giving self time. enjoy for adoption. Our the the are next to and as the we is advance and We are our the of entire great NVIDIA this incredible unfold ago. to CUDA, Turing proud Stay XX platform. of tuned and we impactful a RTX excitement making exceeded driving greatest AI, exciting tackling markets computer this incredibly as to the completes nearly world surge reinvent are platforms some story. RTX graphic years cars graphics. I’m our It disappeared. ProViz These enabling most in largely

you Thank for joining.

You may disconnect. now