AMD - Earnings call transcript Q4 2024 | Advanced Micro Devices, Inc.

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

Operator

Greetings, and welcome to the AMD Fourth Quarter and Full Year 2024 Conference Call. [Operator Instructions]. And as a reminder, this conference is being recorded. It is now my pleasure to introduce to you Matt Ramsay, Vice President of Investor Relations. Thank you, Matt.

begin. may You

Matt Ramsay

our welcome XXXX and should Fourth and copy Quarter had slides. and Thank Results to Conference Year of a review Financial opportunity you press Call. Full release the accompanying you, to AMD's earnings By have now,

found not be the you they have review materials, page the had If Investor to Relations chance these can on amd.com. of

and non-GAAP in non-GAAP refer measures during release full primarily today's The available are to financial posted website. to will our reconciliations We call. GAAP on slides press today's

on Participants our and in Vice Officer; Hu, Lisa a replayed today's conference our Officer call webcast and Executive This Treasurer. are and website. Chair Dr. Su, call President our Financial be live Executive and Chief is and Chief will Jean via

I that Before the on would will attend TMT note Stanley X. Monday, March like Hu to Morgan Jean Global Conference we begin,

such, speak to that for cause and results Today's of on materially statement today to discussion information beliefs, to refer results current current contains from as differ press our in could assumptions materially. cautionary risks more factors our differ forward-looking our our based expectations. as only cause that on actual involve Please and the actual expectations statements and release could uncertainties

that, Lisa? the With I call hand Lisa. will over to

Lisa Su

of to server set Thank our broad multibillion-dollar established successfully those and you, AI a listening launched PC and for good a AMD. We all and transformative share. Data Matt, products afternoon today. year gained market XXXX was franchise, leadership Center significant

than from income and we revenue, net for XX% the XXXX. doubled record cash more annual free result, a flow delivered As year grew

and processor with expanded and significantly customers. cloud, roughly as Importantly, the XX% annual adoption of revenue Data EPYC supercomputing Enterprise Instinct contributed segment Center

Data which billion, results. financial of quarterly Gaming Embedded by record Center billion revenue segment grew by $X.X grew percentage. Center led at basis, than our XX% a Client a Looking segment revenue record revenue double-digit XX% more Fourth nearly annual XX%, and and year to and quarter both Client segments. grew significant revenue to Data increased in full doubled as $XX.X On a year-over-year revenue declines our offsetting

segments. marks $X.X inflection fifth-gen our double-digit Turning driven share of major Center percentage strong, deployments gains increased strong point gen a Hyperscaler XXXX hyperscale server the EPYC billion. at the fourth Data by largest infrastructure XX% segment exited the record powering with In our growth another majority EPYC over business services. in to and for year-over-year sales. EPYC to as demand their we revenue XX% year-over-year customers. accelerated, Turin cloud, both internal online expanded XXXX driven of by was very compute for well CPUs share and ramp

a cloud Google, X,XXX. more with than and launched quarter includes memory by competitive offering new This that powered custom-built delivers demand Alibaba, HPC was on memory EPYC more XXXX instances in fourth of bandwidth EPYC Tencent Public AWS, strong number processor Xx in than the purpose alone. also to offerings. HBM a with AMD XX% XXX based general higher to leadership compared performance increasing Azure instances AI very Microsoft instances and

Enterprise instances strong We year momentum more from cloud in This businesses number growth the activated than consumption in XXXX. doubled nearly significant customers XXXX quarter. EPYC EPYC the EPYC as of cloud a off Forbes also tripled of capped global built using enterprise cloud the the with of from as prior

and year-over-year with Hitachi, EPYC on-prem by high-volume Turning CPU LG, sell-through ServiceNow, double-digit strong Akamai, deployments others. increased, closed to we Verizon, Visa as Enterprise a sales grew adoption. and percentage

XXX go-to-market and available EPYC investments. expanding than Dell, there from turn our production from and OEMs in Super platforms Micro more leading went quarter of increased number platforms pull fourth and platforms Cisco, We that growing the XXX available HPE, than Lenovo, are enterprise the server are the seeing XXXX, based Exiting others. into on EPYC ODMs, more including

clearly across benchmarks. Looking demand the safe are is of best processor broad AMD forward, industry processors the both more with for gen records a map consistent has performance we Turin At and server road range same execution world time, the standard third dependable EPYC seeing XXX as than our made in sustained and fourth the choice.

for the see on workloads and leadership range we in opportunities of of entire points. our and Data processors price based clear performance result, cloud growth system enterprise a EPYC full both across XXXX portfolio across Center As optimized

strong accelerated new to Instinct annual we our was year our uplifts significant greater an accelerators, customer built and the AI of our with and relationships Data delivered AI AI map inferencing outstanding Data revenue with XXXX deliver road ROCm leaders expanded business. hardware industry as to an Turning in $X for Center performance, than Center training of cadence year. billion key suite software

cloud the deployments designed used expanded Llama model serve to XXXB their meta.ai models partners. deep at OCP-compliant Meta quarter, platform production MIXXXX Instinct recommendation exclusively learning Teton large-scale on and and workloads. largest Grand added Looking inferencing for fourth to with our frontier GPUs MIXXXX its AI

have inference for power platforms IBM their and using and DigitalOcean, accelerators enable based being deployed than also this to and Vultr thousands AMD more and generative several number dozen a Instinct Copilot HPC MIXXXX services to other to is to CSPs new applications. for up multiple across enterprise-ready we deploying on in deploying watsonx.ai launched data GPUs platform Instinct are and AI announced expect grow of plans that currently AI scale begun globally, training AI-focused instances. for XXXX. GPT-X training and workloads. Microsoft flagship CSPs MIXXXX instances and IBM,

HPC XX began more and Face. Hugging largest part providing To also the Hub accelerate Instinct factory are adoption and the customers, suite adoption in and For their platforms, production is the containers platforms OEMs grew via with multiple series ODMs. of simplify AMD Enterprise than offering MIXXX a in as enterprise of MIXXXX enterprise solution ready-to-deploy AI on Dell Dell quarter.

most National Labs using and XX the debuted on compute fastest of than the system deliver of APUs list. X.X latest in world MIXXXA as Livermore AMD Lawrence more the Notably, Capitan the energy-efficient XXX exaflops El powers fastest now over to world's supercomputer, the the at top systems supercomputer X performance. XX XX,XXX of XX

at languages. scientific LLMs University Capitan, training Earlier computer the Hunter launched be uses of Stuttgart, XX El used month, supercomputer high-performance will the projects, center different foundational Hunter Like AI the this that advanced for also European both including research in and MIXXXA.

software all AI of the progress significant front, in ROCm stack made layers On the XXXX. we across

and choice innovation. million More as pace the AI, models accelerating industry than for software Our open the leading strategy providing of on establish is X AMD stack industry's ROCm with to greater developers now and serving AMD. like of leading solutions compilers platforms are Triton. vLLM and And Face supported like the out Hugging run the like box PyTorch in frameworks OpenAI on JAX, our

partners. production DeepSeek-VX. runtime the latest numerous faster lead ROCm runs versions state-of-the-art We customers support performance including for optimizations attention that hyperscale included algorithm for year like have with with successfully enabled X ROCm, that also large-scale ended release flash of the prior models including and SGLang than deployments using our Xx We the support ramped day up X.X that multiple to

has increased a these MIXXXX inferencing since X.Xx of result As enhancements, performance launch. latest

their improve out-of-the-box to software investments to diverse for workloads. our number to customers adopting growing a forward, experience AI of Looking we're the power accelerate continuing Instinct

a releases ready at feature began and For that provide example, adding source we new delivering in software continue faster updates performance community deploy we to packages, launch container more biweekly in enhancements resources to frequent the and that test enable to dedicated us open January, and build, pace.

production ramp the competitive On and market, compared front, to MIXXXX we product fourth to support progressing quarter. is well very well volume offerings. new wins. production is customer positioned of in TCO began in MIXXX significant performance advantages The delivering the

also have adopting customers with of progress made significant We Instinct. AMD number a

at customers example, with For the deploying several training recently their scale MIXXX time. are and for LighthouseAI inferencing at closed production large across environments and both we first that MIXXX Instinct wins

in We our performance ahead, in very X CDNA CDNA compute performance come CDNA series Silicon our deliver The to MIXXX well. a up the XXx increase X strong. history, will AI Looking generational next-generation our biggest is compared architecture in has really leap with AI featuring validation running X. feedback MIXXX preparation customer strong, engagements ahead hyperscale on progressing customers Silicon with within broader MIXXX were and receiving net both existing is in deeper LLMs for of and been driving large-scale deployments. schedule. customer XX The first has new series of hours looking at-scale work and

quarter and are and Silicon MIXXX Based plan this track on progress sample to the production to now in mid-year. we to the shipments strong accelerate customers lead early on customer interest series,

tightly integrate support AMD major at Next powerful excited map, at forward road CPU development that the GPU architecture progressing enabling and very we share Instinct solutions level scale. our another Data also As to solutions MIXXX look to Instinct scale rack that into Silicon leap series well. The CDNA multiyear is takes capabilities networking, I'm Center

to HPC CDNA leadership and designed deliver products. out Next capacity and of our and networking expanding and scale ecosystem an bandwidth supporting We up open advantages memory flops AI scale while

remain in in training MIXXX XXXX. large-scale to series track strong the seeing are deployments customer inference for interest on We and and launch

jurisdictions, and the OEM accelerator received to expertise Turning unconditional accelerate in been key Instinct for time Taiwan. to as to passed approvals Singapore quarter Systems future Cloud multiple regulatory our positive acquisition ZT of customer We including has milestones in the ZT Systems. very Japan, acquisition platforms. and and can response market

manufacturing business. We also significant have ZT's interest in received

to the for remains expect production track industry-leading capabilities the ZT's after on shortly first which successfully infrastructure half U.S.-based year. the divest closed Center We Data acquisition, we of

the our share record fourth and driven the client demand for revenue year-over-year XX% gained straight had by New holiday record multiple We to for numerous billion. fourth processors. segment. desktop a Ryzen in globally, to at others both at Turning over higher We MineFactory significantly mobile exceeding Revenue regions client increased channel period. Ryzen $X.X quarter, list AG, desktop dominated quarter the XX% CPU share sell-out Amazon, the as retailers many in best-selling and

sell-through believe mobile, in Ryzen we PC AI share fourth had ramp. notebooks we as Series record quarter the OEM XXX In a

focus Dell we to share first will offer wave powered Pro collaboration time notebooks full in The growing initial the addition of processors. half very by PC share. of In a they the existing we marks growing the full spring were as year portfolio commercial Ryzen commercial with ramping with PC portfolio the our announce to launch a strategic this Dell new commercial PCs excited planned is Ryzen-powered partners, of on with that second to

graphics our CES, that XX of the compute, launch deliver leadership processors new Ryzen At mobile with we expanded and AI capabilities. portfolio

portfolio never Ryzen Our compute performance across has the stack. with processor leadership been stronger

are gaming of and PCs, CPUs, mainstream only the enabling provider a ultrathin, that Copilot+ premium Windows experiences AI offers we on complete For commercial, portfolio notebooks.

strong design percentage TAM planning the PC grow revenue breadth CPU a we're by grow well the momentum, the client to of and believe win for of on XXXX, into leadership ahead segment Looking Based mid-single-digit client we portfolio our can we year-on-year. market.

reducing turning quarter. as our declined XX% segment. sales expected very million. in Sony declined on surpassing Overall, inventory. been by to Gaming fourth Microsoft shipments cumulative the this million and generation Revenue year-over-year console Highlighted unit strong. focused Now channel has to $XXX $XXX Semi-custom as

have will we believe forward, semi-custom Looking in patterns to historical XXXX. inventories now return sales normalized more and channel

sellout In revenue year-over-year as XXXX of our GPUs. channel for in preparation the series Gaming graphics, Radeon next-gen declined launch we accelerated

gaming Our when better March. this XK our of AI-powered on add and that Radeon delivers to portion market XXXX upscaling X significantly technology is will X series performance first the enthusiast architecture. GPUs in highest players for RDNA RDNA tracing address to gaming new early sale focus the volume support the with go bring generation high-quality mainstream rate with

Now strength slower remains million. the softness environment recovering emulation demand than with year-over-year industrial Fourth $XXX and the mixed, is market segment. by as offset test quarter in in decreased communications markets. to XX% turning Embedded aerospace our defense and The to and expected revenue overall and

differentiated in with portfolio We markets. for key solutions expanding computing adaptive continued our the quarter

our We launched and computing compute premium Gen X.X for compute performance PCI our Versal series introduced trading. ongoing industry's We we share adaptive based shipping believe well win leadership positioned CXL defense industry-leading devices aerospace series on the and Alveo for X share in RF with with supporting card Versal XXXX first momentum. and are Gen markets, for began gained performance as latency gains our our and design next-gen adaptive ultra-low X

portfolio summary, platforms year-over-year than we We we adoption closed embedded significant market road accelerated EPYC positioned and as processor up outgrowing and Data our our based in business $XX with XXXX, maps. In billion Instinct a full on new the strongest wins design computing customer in our adaptive Ryzen we won and grew multibillion-dollar Center industry-leading hardware our throughout year, both XX% record record continue more delivering having processor and the ROCm large and share software AI XXXX the CPU of quarterly and momentum, expands of are established history. gains well year revenue. ended designs. to We

For businesses strong and all deliver year-over-year. believe of the businesses. double-digit environment expect and Gaming to strengthen Client EPS backdrop, driving Center in can Embedded XXXX, strong our and growth Against we our revenue this in our we and growth demand modest percentage across businesses, we Data increases

Looking the highlight significant further AI every the Silicon AI to to the latest across of from of Stargate, and the models, layer Allen systems ahead, pace from algorithms of announcements model like infrastructure innovation recent breakthroughs investments incredibly Institute applications. and rapid DeepSeek and stack,

make while to pushing experiences we AI digital advances virtually accessible run to envelope device. on see of and industry and agentic in as software breakthrough capacity enable increased more on invests the every want types innovation compute generative the can AI that the exactly are These

All and products from of end of amounts is across businesses. software new devices. unprecedented power end expertise edge growth the and to the for massive create data and initiatives to client needed with these require AMD AMD compute breadth across only of center, AI our provider opportunities

of for the XXXX Center have coming needed we of AI of And the by billion outstanding of led AMD customer than annual building scaling trajectory our progress a technology tens this foundational years. market. relationships revenue in believe on made Data the $X of growth billions places of capture portion rapid product, over meaningful a franchise We dollars to long-term revenue more and steep this to

to like full provide additional and to some I'd call to turn the over on fourth our results. Jean Now quarter color Jean? year

Jean Hu

record XXXX. good and outlook I'll our growth our in and segment $XX.X margin segments. headwinds by gross fuel and a fiscal investing the afternoon, a very offset well our Data review Lisa, XX%, XXXX, basis Client AMD quarter billion, current Embedded with in and to of earnings first for and executed of of in than driven our everyone. Center revenue XX% results growth. XX% of delivering then you, achieved start aggressively Thank XX% up financial by per our growth which expanded while more in share in AI our Gaming segment, future XXX growth We points provide

year-over-year segments. Gaming the up growth and in in a due was For our the Center XXXX, impact Embedded Gaming margin points by embedded record driven Center revenues billion, the by by as favorable was growing with basis revenue and growth Data lower strong offset Center fourth was revenue Gaming lower $X.X partially of a Client, mix of segment partially XX% was quarter segments. in revenue and Revenue up revenue, the XX%, Client higher primarily XXX offset and lower the Revenue shift Data Data revenue. year-over-year Client of XX% Gross sequentially, to

to and representing invest was Operating operating our income billion, net $ expenses opportunities. XX% was activities marketing interest $X a year-over-year address Taxes, R&D million were of an billion, we other as growth XX% of Operating expense. and record X.X margin. increase $XXX in significant

an earnings first was the XX% share of our model. reflecting significant quarter of per For of operating increase diluted the XXXX business year-over-year, $X.XX, leverage

GPU turning was reportable up strong or growth a a both and Center $XXX CPU of segment. to segment ago. Now starting year EPYC record and revenue income the AMD $X.X year-over-year to Instinct AMD billion, or Data fourth XX% billion Revenue $X.X XX% in the of XX% generation million our compared Data operating with fifth by was driven segment. sales. Center the

year-over-year, was to AMD primarily revenue. mixed. a of million Client or $XXX higher $XXX down strong primarily income Gaming segment customer income million segment $XXX record of demand or revenue to demand a revenue from for market XX% compared operating million, in $XX or of of Client Ryzen revenue year-over-year segment compared X% revenue. year to a billion, operating income be Embedded up segment leverage by a year $X.X operating decrease revenue XX% Gaming to driven Embedded ago, was end $XX down of processors. or XX% million driven operating XX% $XXX or to revenue due million by revenue due year-over-year, was ago. was operating XX% was income XX% $XXX as XX% segment million compared of a semi million, $XXX ago. segment revenue or continues X% year million was

Turning sheet to balance the flow. cash and

of the end record in cash a equivalent free $XXX investments billion flow. short-term During X.X from quarter, cash generated In quarter, million we repurchased billion. increased billion. billion cash fourth quarter, to Inventory we the $X.X cash, shares and to of were $XXX and returned million $X.X $X.X sequentially the million the X.X by In $ shareholders. operations and

For and the million to year, returned shareholders. $XXX shares million we repurchased X.X

have We share repurchase billion $X.X our remaining a in authorization.

align with with Gaming to segment me segment turn Beginning to Center, Data We'll three our reporting. provide Embedded reporting Center, how let business. disclosures single we with our to current reporting. businesses and quarter fiscal Client Therefore, into We client one the gaming segments: cover and the Embedded. plan I consistent the manage Before combine and our statement financial first Data with our Gaming segment disclosures. continue reportable XXXX financial year Client, distinct the outlook, our for revenue financial to

Now quarter our outlook. turning to XXXX first of

XX% and offsetting in plus strong We the expect by approximately in revenue growth business. decline Center decline $XXX significant business. Embedded our our driven business a Client million, modest or More be in up than a and to minus Gaming $X.X year-over-year, Data billion our

revenue across We businesses. approximately primarily X%, sequentially to be driven down expect our seasonality by

tax operating $XX diluted share the gross income net approximately X.XX effective shares. and XX%. count to non-GAAP be to $X.X XX% be billion. rate to expected million, quarter expenses approximately first Non-GAAP be In is other to to be be addition, billion margin expect Non-GAAP non-GAAP we approximately

strong XXXX growth earnings expand for All innovation while to closing, a In growth. revenue fuel long-term a AMD, at faster execution and was investing disciplined our the AI deliver and than to revenue. in year demonstrating rate

will build ahead, double-digit Looking earnings revenue growth we beyond. percentage drive further and the to momentum on and XXXX accelerate in

turn to that, it Matt for Q&A With the I'll back session.

Matt Ramsay

the ready Jean. now very Q&A start much, you to Thank We're session.

pulls question remind ask to operator please Operator, questions, please poll we Thanks. the one for follow-up. questions. each for and brief a participant As

Operator

Aaron of Fargo. first the [Operator the Wells with Instructions]. question And line from comes Rakers

Aaron Rakers

ask just dynamics. I GPU we think about the GPU, which and about extremely you about the revenue, north color I'll of out you of guess I'm business, about progression framing revenue, half I second of $X is pace business any appreciate talked think think first we delivering XXXX? half revenue it curious as the the us -- of of versus gate the XXXX. right impressive of should can as kind product far provide if how of think we in cycle the kind as And about as Instinct as you the some we billion of

Lisa Su

Thanks Aaron. for Sure, the question.

we first terms business. very of we how XXXX Center pleased Data finished the GPU of all, look, with were So in

the throughout ramp executed steep year, the went as we was think I and well. the team

into on the maps. software as very and we're road with happy I we're in both making mentioned hardware road Going that progress the maps actually prepared remarks, XXXX, the the

launched the end So XXX side, then. on the fourth of the hardware started quarter, shipments at we

the XXX the half year. come And that of we'll and then is We in both series. on have XXX deploy the that news the first XXX new on big designs have

expected, come than had previously bring-up customer there's and that. demand the that has we stated we thought half the we in of would frankly, that launch strong we So And that better very for up year. second

which that ramp of are middle competitiveness. actually our we the So production year, into relative improves the to pull going

revenue will double very Data digits Center Center so be be And GPU as strong XXXX Center with relates XXXX. pleased XXX from product overall, second think would will So and overall well double the expect product than of Data business. for how Center grow Data in -- into the business we're to both the you line certainly, year just catalyst would then grow half full it shape a server the half, line Center I as the GPU stronger Data both the Data strong the trajectory that the first of business the as the will digits. given going But

Aaron Rakers

mentioned about Data And that down the know you're down business X% you seasonality a Yes. to Center Are thinking assuming the that do that quick across you. how guidance sequential, in seasonality? in follow-up, XQ? sequentially segments. overall frame to Thank as just And relative in I relative total I you

Lisa Su

Yes, sure, Aaron.

let on more give guide. color you So the me some QX

way We average, just guide breaks the as be segments would And to out business in assume business average. was so about than Jean that. X% the sequentially, embedded down expect So mentioned. down Data the corporate that each Center the the that of down QX more be that Client and would

that seasonality where than those then seasonality we're but And of atypical from coming inventory be And that. a inventory little expect would less we would Gaming that a given a off corporate business Just a let's for now year would And normalized, call than businesses. average. be we will bit normalization. little is less standpoint, was little down expect of the down when has there a that's lot a it,

Operator

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

Timothy Arcuri

think long the aggressive mid- are much that ask teens. become on as well. should you past super about or to more pricing server that Are Jean, grow they're is in your not bit aggressive seeing attempt already It shore as more getting that their share. business. sounds business they a CPU up going as And little you as you like I aggressive grow expecting pricing. core on a pricing I the competitor said wanted to have to high is count your to on roughly them

to you server line the in of kind teens. can that wondering with CPU think So high grow business core still if mid- that

Jean Hu

Yes.

generation, best provide our product TCO the CPU market we First, Milan, have for we customers competitive portfolio. but not even based Genoa currently and portfolio of on very only a then Turin the the we assume server is always best but lineup from

only quite share. perspective about not from to confident overall, CPU drive continue unit the are continuing server actually a businesses gain to So growing and perspective, we ASP

Timothy Arcuri

GPU can a flat then, where it willing be give sense or And range. came for to March? just Jean, in $X assumed to the thinking you for number I'm Center of is in December? you And then Data it's give probably a be up, us would billion down,

Jean Hu

Yes.

way I is to performance our Center think well. expectations. look at our the our did QX is Data overall Actually consistent with really business

Of Data performance. did course, when server GPU, we Center Data than it's the at the better and But look overall, GPU. Center server our very consistent with

Lisa Su

Yes. what Maybe I'll would your add, go on we just Tim, XXXX. as you as expect question to into

you should side. I businesses GPU the both that segment true half of side well And half the with will XXXX be of think first server that's the for on assume 'XX. as Center Data as Center Data second the consistent

Operator

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

Vivek Arya

Lisa, the Center business. questions GPU a few on Data

Day explicit specific you at I and think expectations. did it think if set you in versus assume not December, for beating plus meeting has last to This time, last that I to year, back what then grow a right, curious you that at had you a changed. AMD billion XX% long-term year. that the was forecast, Is $X fair year, your CAGR. 'XX, set I that can about go I'm or very have setting And Analyst and

the AMD contrast years trend that then whether and can grow XX% just at line. So X

Lisa Su

Sure.

So Vivek, question. thanks for the

with XXXX. give it actually at will billion. business level be clear more qualitative of first out the two into I long-term The believe between business, $X strong. over color as you're for about the the year And I wanted we certainly And that think Data absolutely as demand to at is now was going. some TAM segment $XXX right. upwards to we growth now compute going what's to progression we get Data talked the think of question -- Center we XXXX, And your mean is GPU I the at accelerator some a what AI we've time is by Center some scale, on businesses. go to billion at the about to as for relative as guidance I color our look rates,

points there. all that is out of would recent there think I strong suggest the demand data a

each a that XXXX, generation, tens into you this comments we we MIXXX significant start. and is business see confidence for in guiding years. getting is is to with have view in our of specific growing billions Without we as business of So particularly there couple get gives we that addresses around that's there. then on with the and of of the we right? set training. innovation road we going as next MIXXX XXX go we both number great one And broader just a made was rack-scale and as a including a can well map workloads the designs, through the do stronger as stronger of inference And series see traction series, excitement the and what that

long we're term, give as go yes, we'll we you through quarter each in progress the bullish on and certainly So XXXX.

Vivek Arya

news, that Like is few so done, at have forward? the that, going there have they how opportunity in the confident semiconductor And kind share opportunity. versus DeepSeek ASIC Have about Thank more right? for within something the disruptive make would over way should the news your you of do to then any those think the follow-up, that. my are views And is you you, And us that about in or more develops love I all? years? what that recently, on your from One Lisa. GPU heard perspective parts two overall once conservative reduces there evolved market next

Lisa Su

year. weeks for Yes, Great. Vivek. Yes. Thanks first it's question, of the been few the a exciting pretty

pace much Allen announcements, in about just talk how of the as I well AI as Stargate Institute the think DeepSeek world. happening announcements, innovation some of rate that's the the and

more ways adoption. So capabilities to AI the on that adoption. allows us continue in about because is infrastructure new are for DeepSeek. a and inference think Look, deploy fact is it broader AI to we The bring specifically good the and and thing application that relative models to innovation compute less algorithms to space training good there with actually

we're open standpoint, think like of much that from that pace source. amazing. source think, to is believers I in pretty there, is expect the adoption that at standpoint, And our and fact from open go. And very models, looking having also we I big rate things how we the

how the next-generation we overall some bullish So were call overall massive out, building I the and we it. to AI. and with feel about that for let's on should about question, infrastructure the it, OpenAI cycle. of infrastructure announced And feel similarly, of Stargate investments mean,

the as expect that say I think of curve. more a AI very a steep lot And those result, certainly all is on we part the should innovation.

then ASIC a compute of also think amount the right you address because for been there's need a place the good a that in right is me the workload. always And I've that where on let discussion. point, believer I

AI, that diversity be pieces. GPUs, going as is of that. still will very types it GPUs of models, medium I look programmability adjustments a so And of whereas specific and all really it in about when of ASICs in at models, you're given FPGAs. specific small models, training, will those AMD it, belief is, more, you're workloads, optimized, a algorithm these my ASIC Relative the workload includes compute. of there is CPUs, of and we've inference the or enable $XXX large always on is part But going significant And when TAM changes. piece models, much ASICs the AI our to talking all across need portfolio, all to still because given broad how time, plus models that going that to call But algorithms change smaller TAM with foundational out billion had

really And GPUs. want CPUs, overall also conversations well we're partner. of an as customers number have compute in to as a ASIC So involved

Operator

And of Buchalter TD with Cowen. from the Joshua comes question the next line

Joshua Buchalter

saw things see But potentially pulled playing more of changed clarify, ahead whether over half the Obviously, related MIXXXX or to of speak how if line, as with potential could last first good a wondering and pulled it this MIXXXX you consistent could shape ramps was pocket sort air half second first to wanted of few Data 'XX, it months not the out I you likely and was time this half MIXXX fully? consistent to was the midyear. you launch? that with is be GPU in Or and into said to I Center 'XX

Lisa Su

Yes. for the Data from Center we've standpoint, more ramp. gotten on Joshua. would No, say GPU the I question, XXXX our positive Thanks incrementally

mean expect the But XXX frame. end and with the XXX was in are have it always, it's up, first new think content. with an with half, in XXX some midyear larger be up second I more wins that inference series first more powerful series, include that half ASPs deployments in you and I statement, half And is a half deployed of we going year. training incremental GPU, design AI to pulling into time would mentioned, but as positive. I XXX go from we important a that and second

So is would to have the we shape similar what expected before.

Joshua Buchalter

drivers of how Could it clarify you notebook than believe the the I ask enterprise? year? how and given not, the a it market certainly inventory half the strength quarter, on higher think at outperformed are model. And to it particular, in about in about has number And broader desktop, I all or much the the and our you second we growth of should Obviously, buildup, going in fourth question to across PC mean Client. worried mean, was the your client XQ bluntly, revenue put

Lisa Su

the Yes. question. Thanks for

really are that. strong very a well and throughout couple a There business QX performed Client was quarter. XXXX for reasons of Our

we're adoption So what seeing We products. is believe we of should there's new substantial through very actually inventory don't some go that strong that. buildup. We think our

So years. we the many sell-out side, desktop highest our on in saw

As of catching as holiday have through the continued they CPUs, January up gaming with in been demand we've market, there. new shipping very month and through frankly, we strongly the launching some constrained we're season, went our the

So desktop launching new business very numbers slew was side, of new saw the strong. AI a at And with PCs also notebook that we number mobile CES. the announced part partners OEM of our on we

those PC top are well strongest PC products, into and very on to the our bottom the have We side portfolio playing mobile market. compatible with Copilot+

So was I QX think strong.

to that see in pull-ins there fourth whether tariffs. some I were don't know we didn't -- was commentary I relative there that quarter. about the

think, saw as we I sellout. I strong said,

mobile our But XXXX. the that expect grow first the is prospects in part Going I business seasonality. to overall, do seasonality of typical bullish think better on and seasonality very portion the But is, in desktop than we into the business. we're it, the of business more quarter, call performing portion is there. of the let's clients our

from all portfolio. PCs of product commercial drivers dynamics market as the the given Just portfolio some to of as new well our

Operator

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

Harlan Sur

follow-on pretty start server think, And On Enterprise a it step sequentially. share quarter XXXX? guys know increases CPU recovery on-prem For sequentially this based. grow of as the do your I third grow solutions. last sequential QX? as QX the double-digit second And was the Enterprise quarter, think, you was in up fourth service quarter I prospects did a that, consecutive year-over-year in sixth broad servers year, digits strongly see business did I quarter, the double Did a Lisa, see in you the how sequentially you basis, to in for to into overall your segment in of

Lisa Su

the for thanks Harlan, Yes, question.

Jean so I quarter, server So sequential a fourth growth cloud in double-digit the both think we We Enterprise. our saw in and as mentioned see did earlier, business. in that

more the cloud think continuing Genoa in We're footprint workloads business to been has server and grow our addition portfolio Turin well. to as extremely of strength have performing the Milan. with we the I

your Enterprise, some we're the traction And then on to question in do seeing I Enterprise. believe strong

number competitiveness XXXX, XXXX, more we've as that sales every in and in it Enterprise now to into into through sales deployments. throughout have converting is, invested are POCs there of price go-to-market strong X-month a go is off. workload. lot paying a a every across The And we've X- a volume us from portfolio investments point, invested helping that more we've every cycle are what's frankly, go-to-market significant think resources But a we standpoint, cycle. the very convert new we into as I count, often seen core

think GPUs. we So for see XXXX I strong server a

Harlan Sur

AI critical that. very important. yet that pace incumbent more Ethernet becoming is part a of here, infrastructure this Networking strong appreciate you're technology, misconception ecosystem. behind the curve the the of even AMD keeping I kind is installed leveraging There be you seems to

Ethernet Consortium. Autonet the spearheading are guys You

Infinity connectivity. got technology scale-up You've for Fabric your

as of innovations this networking AI the adoption AI on platforms. your be customer the the bringing team you What's to networking-related going overall So to your market feedback any been continue drive year? architectures? is to like And

Lisa Su

is Thanks, the of for of Yes. The our proof is are important Harlan, part bit and that we jointly. networking been think that MIXXX with an a question. it's and about No solution, MIXXX. spending effort AI that to our to increasing an quite actually as area to partners to way it and from points question. our investing have going customers Networking we're MIXXX MIXXX extremely

proof So of we're a in with those points, points. of each number increasing

clusters inference, I cost of Certainly, great total think performance people want to of more shown and ownership. on ours. see we've

We are are that important the putting training now on And board. networking. is part -- systems also of we have the a number there

very Ethernet. We the with We is this have technology that closely our partners worked for future. on believe right the

In mentioned look customers, AI addition that our actually solutions developing with team MIXXX at are both we're with as well standing the series. level in our in-house Pensando XXX as full rack our Advancing NIC to forward Forrest at the as networking also event. QX we solutions, AI working third-party we up our And own

of we all absolutely the we important. very is to And the with the are think addition road system of is to super level scaling deliver in yes, important our and So that net is it hardware it and map. I that, track work, on believe software

Operator

next of question line Curtis with comes And the Jefferies. Blayne from the

Blayne Curtis

growth want slows data business, for of to the next obviously, the your commentary year-over-year, follow for kind sequential the kind kind very X of but up strongly synergy it just I seems of quarters.

why. new to products you some have want just I So understand Obviously, the coming.

the look ASIC terms I inference. brain of kind just versus to story also So at a in there products. to maybe tinker when you tends of training want to the focus new it's shift to be on lines, just the on shift

So a customers initially workloads any you between of of lot shift your terms from demand just as in your perspective, I well? seeing inference were know the training Are our inference.

Lisa Su

Blayne. would build as Sure, saw Center GPU it I the up the Look, growth say we way throughout is tremendous a Yes. business we XXXX. Data

for deployments. new to important bring MIXXX is product are to a bit So new set. continuing the in we're is going the I through us seeing was the think for a do continuing Clearly, in year. first bringing of the little half series and the we're But very We're time frame customers. what on transition key of customer we very,

expect, lot is learned think, deployments. the MIXXX, this clean. landscape I come the of we've overall deployments we is given very And has So positive. the of on fact hardware And might that, that a as from initial

to the that second as big seen work each it as a all our really customers have like right? But said, one really that us that solution. People road I of a to consistent is map We've certainly, done then with that's haven't your kind conversation said strong as inference. And relates we've see say we've step-wise of show like conversation. what in solutions. And to we in a training the the want at to those of the question, I ASICs, it part relates to shift the AMD will following.

of networking. getting significantly into of just -- On talked more Harlan's We of this some thousands about about the And then up invested in tens those the software question side, centers earlier ramping is and sort we've data libraries. GPUs. of training

And good they I So strong of are the from sort toolbox. my very just there. want side, the I to standpoint, AMD also think part everyone, the I we're are -- reiterate progress I on ASICs to are a look, think very of remind but ASIC making want a part solution, of they

time. for a semi-custom long we've solutions So done

CPUs, as they customers all very We baseline like do they'd is GPUs and a all innovate of of like on would that. the And that that I top well discussions building involved our is well. networking of to that's put need think our together. to what really have ASIC as to solutions capability take are in IP we as And do with differentiates number of the you our what blocks the technologies of

Matt Ramsay

Operator, I have two callers, we time more please. for think

Operator

Rasgon next The from comes question Okay. Research. line with the Bernstein Stacy of

Stacy Rasgon

this explicitly. little to want I ask more a

I My QX. the even sequentially GPU your first up GPUs server guidance math that was that in double said sequentially. flattish suggests half could And business business digits in the down given your guess, meant for, So was strong down have wrong of both here? you'd my be and feel -- in doing am suggest Again, missing with math? versus I 'XX. in to second the does like something 'XX I what of sequentially math half not am QX Or you that QX like

Lisa Su

maybe we think perhaps, digits. said little Yes. Stacy, help there. color you don't strong bit give a me I double let of

I think digits. we said double

was Data X% segment Center perhaps bit that the a So Server more less Center sequentially. GPU that. little was so than that. Data a up -- than is was

you some be for out GPU Center light Data number. are the QX little there, might that bit the I in think models a of

But just that need 'XX, there to be done the think 'XX be to -- first your it, be I So some might if would half you second adjustments there. take let's But halves, incorrect. suggestion call We happen. what bit flattish, half we'll it on mean currently or kind is plus exactly it minus. goes. But little roughly we dependent to have I just see. to see of be a when how going deployments that's

Stacy Rasgon

'XX it. higher exit for you your that? Got on follow follow-up, my commit GPUs I exit to there, in your Are you rate is maybe willing than rate do to on in think 'XX. to guess And

Lisa Su

course. Absolutely. would Yes, grow double digits strong to hard otherwise, of right? be It

Operator

And Sachs. the of Toshiya final question Goldman comes from the line Hari with

Toshiya Hari

the the your prepared coming ramp can customers on how the of question business. you're And across thinking and that's would internal Enterprise, characterize to over and than at the delineate years? vis-a-vis remarks, intensity would all market In more a that share server And about then some question. year? helpful. quarters had the you custom to Silicon of major hyperscalers. you cloud the part talked expected about How be if curious I that those between really this XX% you Lisa, having CPU kind I'm B competitive you

Lisa Su

Toshiya. Sure,

in the investments server the into think optimizing 'XX. and cloud of, for AI. say, would a a second part in me CPU let's see let of pick little up we half year side was that were XXXX, I 'XX, into market saw between on investments we we I bit of to call people good that as of The there think go So less We and -- went as into sort it, Enterprise. early sort expect of we 'XX, as the the look

of are number data now, cloud let's So refresh call And are coming centers. there again. cycles are that vendors Enterprise their some a the reupdating it, certainly, of in

second question your relates as then it And was, to...

Toshiya Hari

custom Competitive Silicon.

Lisa Su

it expanded right Toshiya, generation. Yes. right versus is do What much always X would Look, the I about the say, competitive product think about it's same. workload. custom line I have of Silicon each less product look, about the it's We've the It's seen Zen we've product you xXX. design is for our the market you've as done, more a market. And line. Zen in for and our well X What as the core points in server

core then core perf have low cloud highest we performance, best count, dollar. per and So that high we optimized have enterprise native count,

as amount -- trust delivery our And that's up we've across winning win those it's and think, Enterprise. do look, We always number competitive. as And years. our over I of think very very, with the of very both our portfolio. continuing share every good of a are for share. of capability there's strength I franchise -- I design last And built things, also, take think product grow the but And cloud we're to fair the we I seriously, sort think we our

Toshiya Hari

quick follow-up, great. That's maybe one And a for then Jean. as

major of first into we curious the second the are So takes the importantly, should more in QX margin and you're and half. going guiding the some to what cognizant of be quarter. and puts things XX% I'm gross that

Center the expect run if the improve the in But half, would Data to kind I of margins your commentary to minuses, helpful. you Given be would can more and second yes, pluses through second really that skewed half.

Jean Hu

question. the for Thanks Yes.

You're right.

revenue our Our mix. gross margin is primarily driven by

grow I not year-over-year. think is when QX same to client you At the year-over-year. XXXX, also significantly look at time, guide, to growing business Center only -- Data continued looking

with revenue the quite consistent is the overall, So QX. mix

XX%. So guide the is gross margin

gross driver the the our going the with Data and second the half, fastest for up level, think second gross to we But this growth do the do will margin company at will is mix XX%. is if the feel into step in believe for margin first half, I revenue be drive Center we that consistent half.

Matt Ramsay

that, in you I operator. right. All the thank today we now, are call say your listened interest participated Thank that close to everybody wanted With for very ready you to AMD. much. I and to and just in think

Operator

for you. conclude does Thank And teleconference. gentlemen, thank your and you We today's ladies that participation.

this time. at lines your disconnect may You

Advanced Micro Devices (AMD) 4 Feb 252024 Q4 Earnings call transcript