Superior Micro Gadgets (AMD) Q2 2024 Earnings Name Transcript

Superior Micro Gadgets (AMD -0.94%)
Q2 2024 Earnings Name
Jul 30, 2024, 5:00 p.m. ET

Contents:

• Ready Remarks
• Questions and Solutions
• Name Contributors

Ready Remarks:

Operator

Greetings, and welcome to the AMD second quarter 2024 convention name. Presently, all individuals are in a listen-only mode. A quick question-and-answer session will observe the formal presentation. [Operator instructions] As a reminder, this convention is being recorded.

It’s now my pleasure to introduce to you Mitch Haws, vp, investor relations. Thanks, Mitch. You could start.

Mitch Haws — Vice President, Investor Relations

Thanks, and welcome to AMD’s second quarter 2024 monetary outcomes convention name. By now, you must have had the chance to overview a duplicate of our earnings press launch and the accompanying slides. When you’ve got not had the possibility to overview these supplies, they are often discovered on the Investor Relations web page of amd.com. We are going to refer primarily to non-GAAP monetary measures throughout at present’s name.

The total non-GAAP-to-GAAP reconciliations can be found in at present’s press launch and the slides posted on our web site. Contributors on at present’s convention name are Dr. Lisa Su, our chair and chief govt officer; and Jean Hu, our govt vp, chief monetary officer, and treasurer. It is a stay name and can be replayed by way of webcast on our web site.

Earlier than we start, I want to notice that Dr. Lisa Su will attend the Goldman Sachs Know-how Communacopia and Know-how Convention on Monday, September 9; and Mark Papermaster, govt vp and chief know-how officer, will attend the Deutsche Financial institution Know-how Convention on Wednesday, August 28. Right now’s dialogue incorporates forward-looking statements based mostly on present beliefs, assumptions, and expectations, converse solely as of at present and, as such, contain dangers and uncertainties that might trigger precise outcomes to vary materially from our present expectations. Please confer with the cautionary assertion in our press launch for extra data on elements that might trigger precise outcomes to vary materially.

With that, I will hand the decision over to Lisa.

Lisa T. Su — President and Chief Government Officer

Thanks, Mitch, and good afternoon to all these listening at present. We delivered sturdy second-quarter monetary outcomes, with income coming in above the midpoint of steerage and profitability rising by a double-digit share pushed by higher-than-expected gross sales of our Intuition, Ryzen, and EPYC processors. We continued accelerating our AI traction as main cloud and enterprise suppliers expanded availability of Intuition MI300X options, and we additionally noticed constructive demand indicators for general-purpose compute in each our consumer and server processor companies. Because of this, second-quarter income elevated 9% yr over yr to $5.8 billion as considerably greater gross sales of our knowledge middle and consumer processors greater than offset declines in gaming and embedded product gross sales.

We additionally expanded gross margin by greater than three share factors and grew EPS 19% as knowledge middle product gross sales accounted for practically 50% of general gross sales within the quarter. Turning to the segments. Information middle phase income elevated 115% yr over yr to a file $2.8 billion, pushed by the steep ramp of Intuition MI300 GPU shipments and a robust double-digit share improve in EPYC CPU gross sales. Cloud adoption stays sturdy as hyperscalers deploy fourth-gen EPYC CPUs to energy extra of their inside workloads and public cases.

We’re seeing hyperscalers choose EPYC processors to energy a bigger portion of their functions and workloads, displacing incumbent choices throughout their infrastructure with AMD options that supply clear efficiency and effectivity benefits. The variety of AMD-powered cloud cases out there from the most important suppliers has elevated 34% from a yr in the past to greater than 900. We’re seeing sturdy pull for these cases with each enterprise and cloud-first companies. For example, Netflix and Uber each not too long ago chosen fourth-gen EPYC public cloud cases as one of many key options to energy their mission-critical customer-facing workloads.

Within the enterprise, sell-through elevated by a robust double-digit share sequentially. We closed a number of giant wins within the quarter with monetary providers, know-how, healthcare, retail, manufacturing, and transportation prospects, together with Adobe, Boeing, Industrial Gentle & Magic, Optiver, and Siemens. Importantly, greater than one-third of our enterprise server wins within the first half of the yr have been with companies deploying EPYC of their knowledge facilities for the primary time, highlighting our success attracting new prospects whereas additionally persevering with to increase our footprint with current prospects. Wanting forward, our next-generation Turin household that includes our new Zen 5 core is wanting very sturdy.

Zen 5 is a grounds-up new core design optimized for management efficiency and effectivity. Turin will lengthen our TCO management by providing as much as 192 cores and 384 threads, assist for the newest reminiscence and I/O applied sciences, and the flexibility to drop into current fourth-gen EPYC platforms. We publicly previewed Turin for the primary time in June, demonstrating our important efficiency benefits in a number of compute-intensive workloads. We additionally handed a serious milestone within the second quarter as we began Turin manufacturing shipments to steer cloud prospects.

Manufacturing is ramping now forward of launch, and we anticipate broad OEM and cloud availability later this yr. Turning to our knowledge middle AI enterprise. We delivered our third straight quarter of file knowledge middle GPU income with MI300 quarterly income exceeding $1 billion for the primary time. Microsoft expanded their use of MI300X accelerators to energy GPT-4 Turbo and a number of copilot providers, together with Microsoft 365 Chat, Phrase, and Groups.

Microsoft additionally turned the primary giant hyperscaler to announce basic availability of public MI300X cases within the quarter. The brand new Azure VMs leverage the industry-leading compute efficiency and reminiscence capability of MI300X at the side of the newest ROCm software program to ship management inferencing value efficiency when working the newest frontier fashions, together with GPT-4. Hugging Face was one of many first prospects to undertake the brand new Azure cases, enabling enterprise and AI prospects to deploy tons of of 1000’s of fashions on MI300X GPUs with one click on. Our enterprise and cloud AI buyer pipeline grew within the quarter, and we’re working very intently with our system and cloud companions to ramp availability of MI300 options to deal with rising buyer demand.

Dell, HP, Lenovo, and Tremendous Micro all have Intuition platforms in manufacturing, and a number of hyperscale and Tier 2 cloud suppliers are on observe to launch MI300 cases this quarter. On the AI software program entrance, we made important progress enhancing assist and options throughout our software program stack, making it simpler to deploy high-performance AI options on our platforms. We additionally continued to work with the open-source neighborhood to allow prospects to implement the newest AI algorithms. For example, AMD assist for Flash Consideration 2 algorithm was upstreamed, offering out-of-the-box assist for AMD {hardware} within the well-liked library that may improve coaching and inference efficiency on giant transformer fashions.

Our work with the mannequin neighborhood additionally continued accelerating, highlighted by the launches of latest fashions and frameworks with Day 1 assist for AMD {hardware}. At Computex, I used to be joined by the co-CEO of Steady Diffusion to announce that MI300 is the primary GPU to assist their newest SD 3.0 Picture Era LLM. Final week, we have been proud to notice that a number of companions use ROCm and MI300X to announce assist for the newest Llama 3.1 fashions, together with their 405 billion parameter model that’s the {industry}’s first frontier-level open-source AI mannequin. Llama 3.1 runs seamlessly on MI300 accelerators.

And due to our management reminiscence capability, we’re additionally in a position to run the FP16 model of the Llama 3.1 405B mannequin in a single server, simplifying deployment and fine-tuning of the industry-leading mannequin and offering important TCO benefits. Earlier this month, we introduced our settlement to accumulate Silo AI, Europe’s largest non-public AI lab with in depth expertise growing tailor-made AI options for a number of enterprise and embedded prospects, together with Allianz, Ericsson, Finnair, Karber, Nokia, Philips, T-Cell, and Unilever. The Silo staff considerably expands {our capability} to service giant enterprise prospects seeking to optimize their AI options for AMD {hardware}. Silo additionally brings deep experience in giant language mannequin growth, which can assist speed up optimization of AMD inference and coaching options.

Along with our acquisitions of Silo AI, Mipsology, and Nod.ai, we have now invested over $125 million throughout a dozen AI corporations within the final 12 months to increase the AMD AI ecosystem, assist companions, and advance management AMD computing platforms. Wanting forward, from a street map perspective, we’re accelerating and increasing our Intuition street map to ship an annual cadence of AI accelerators, beginning with the launch of MI325X later this yr. MI325X leverages the identical infrastructure as MI300 and extends our generative AI efficiency management by providing twice the reminiscence capability and 1.3 occasions extra peak compute efficiency than aggressive choices. We plan to observe MI325X with the MI350 sequence in 2025 based mostly on the brand new CDNA 4 structure, which is on observe to ship a 35x improve in efficiency in comparison with CDNA 3.

And our MI400 sequence powered by the CDNA Subsequent structure is making nice progress in growth and is scheduled to launch in 2026. Turning to our AI options work. Broadcom, Cisco, HP Enterprise, Intel, Google, Meta, and Microsoft, all joined us to announce Extremely Accelerator Hyperlink, an industry-standard know-how to attach tons of of AI accelerators that’s based mostly on AMD’s confirmed Infinity Material know-how. By combining UALink with the broadly supported Extremely Ethernet consortium specification, the {industry} is coming collectively to ascertain a standardized strategy for constructing the subsequent era of high-performance knowledge middle AI options at scale.

In abstract, buyer response to our multiyear Intuition and ROCm street maps is overwhelmingly constructive and we’re very happy with the momentum we’re constructing. Because of this, we now anticipate knowledge middle GPU income to exceed $4.5 billion in 2024, up from the $4 billion we guided in April. Turning to our consumer phase. Income was $1.5 billion, a rise of 49% yr over yr, pushed by sturdy demand for our prior-generation Ryzen processors and preliminary shipments of our next-generation Zen 5 processors.

In PC functions, Zen 5 delivers a median of 16% extra directions per clock than our industry-leading earlier era of Ryzen processors. For desktops, our upcoming Ryzen 9000 sequence processors dropped into current AM5 motherboards and extends our efficiency and power effectivity management throughout productiveness, gaming, and content material creation workloads. For notebooks, we introduced our Ryzen AI 300 sequence that extends our industry-leading CPU and GPU efficiency and introduces the {industry}’s quickest NPU with 50 tops of AI compute efficiency for Copilot+ PCs. The primary Ryzen AI 300 Collection notebooks went on sale over the weekend to sturdy critiques, and greater than 100 Ryzen AI 300 Collection premium, gaming, and industrial platforms are on observe to launch from Acer, ASUS, HP, Lenovo, and others over the approaching quarters.

Buyer pleasure for our new Ryzen processors may be very sturdy, and we’re properly positioned for ongoing income share positive factors based mostly on the power of our management portfolio and design win momentum. Now, turning to our gaming phase. Income declined 59% yr over yr to $648 million as semi-custom SoC gross sales declined in step with our projections. Semi-custom demand stays smooth as we at the moment are within the fifth yr of the console cycle, and we anticipate gross sales to be decrease within the second half of the yr in comparison with the primary half.

In gaming graphics, income elevated yr over yr, pushed by improved gross sales of our Radeon 6000 and 7000 Collection GPUs within the channel. Turning to our embedded phase. Income decreased 41% yr over yr to $861 million. The primary quarter marked the underside for our embedded phase income.

Though second-quarter income was flattish sequentially, we noticed early indicators of order patterns enhancing and anticipate embedded income to progressively get better within the second half of the yr. Long term, we’re constructing sturdy design win momentum for our expanded embedded portfolio. Design wins within the first half of the yr elevated by greater than 40% from the prior yr to better than $7 billion, together with a number of triple-digit million-dollar wins, combining our Adaptive and x86 compute merchandise. We introduced our Alveo V80 accelerators that ship management capabilities in memory-intensive workloads and entered early entry on next-generation edge AI options with greater than 30 key companions on our upcoming second-gen Versal Adaptive SoCs.

Final week, we additionally introduced Victor Peng, President of AMD, would retire on the finish of August. Victor has made important contributions to Xilinx and AMD, together with serving to scale our embedded enterprise and main our cross-company AI technique. On a private notice, Victor has been an excellent accomplice to me, guaranteeing the success of our Xilinx acquisition and integration. On behalf of the entire AMD workers and board, I wish to thank Victor for all of his contributions to AMD’s success and want him all the most effective in his retirement.

In abstract, we delivered sturdy second-quarter outcomes and are well-positioned to develop income considerably within the second half of the yr, pushed by our knowledge middle and consumer segments. Our knowledge middle GPU enterprise is on a steep development trajectory as shipments ramp throughout an increasing set of consumers. We’re additionally seeing sturdy demand for our next-generation Zen 5, EPYC, and Ryzen processors that ship management efficiency and effectivity in each knowledge middle and consumer workloads. Wanting forward, the fast advances in generative AI and growth of extra succesful fashions are driving demand for extra compute throughout all markets.

Beneath this backdrop, we see sturdy development alternatives over the approaching years and are considerably rising {hardware}, software program, and options investments with a laser deal with delivering an annual cadence of management knowledge middle GPU {hardware}, integrating industry-leading AI capabilities throughout our complete product portfolio, enabling full-stack software program capabilities, amplifying our ROCm growth with the size and velocity of the open supply neighborhood, and offering prospects with turnkey options that speed up the time to marketplace for AMD-based AI programs. We’re excited concerning the unprecedented alternatives in entrance of us and are well-positioned to drive our subsequent section of great development. Now, I might like to show the decision over to Jean to supply some extra colour on our second-quarter outcomes. Jean?

Jean Hu — Government Vice President, Chief Monetary Officer

Thanks, Lisa, and good afternoon, everybody. I will begin with a overview of our monetary outcomes after which present our present outlook for the third quarter. We’re very happy with our general second quarter monetary outcomes that got here in above expectations. On a year-over-year foundation, knowledge middle phase income greater than doubled, consumer phase income grew considerably, and we expanded the gross margin by 340 foundation factors.

For the second quarter of 2024, income was $5.8 billion, up 9% yr over yr as income development within the knowledge middle and the consumer segments was partially offset by decrease income in our gaming and embedded phase. Income elevated 7% sequentially, primarily pushed by development within the knowledge middle and consumer segments income. Gross margin was 53%, up 340 foundation factors yr over yr, primarily pushed by greater knowledge middle income. Working bills have been $1.8 billion, a rise of 15% yr over yr as we proceed to spend money on R&D to deal with the numerous AI development alternatives forward of us and enhanced go-to-market actions.

Working revenue was $1.3 billion, representing a 22% working margin. Taxes, curiosity expense, and different was $138 million. Diluted earnings per share was $0.69, a rise of 19% yr over yr. Now, turning to our reportable segments.

Beginning with knowledge middle, knowledge middle delivered file quarterly phase income of $2.8 billion, up 115%, a $1.5 billion improve yr over yr. The information middle phase accounted for practically 50% of whole income, led primarily by the steep ramp of AMD Intuition GPUs and a robust double-digit share EPYC server income development. On a sequential foundation, income elevated 21%, pushed primarily by sturdy momentum in AMD Intuition GPUs. Information middle phase working revenue was $743 million or 26% of income in comparison with $147 million or 11% a yr in the past.

Working revenue was up greater than 5 occasions from the prior yr, pushed by greater income and working leverage whilst we considerably elevated our funding in R&D. Shopper phase income was $1.5 billion, up 49% yr over yr and 9% sequentially, pushed primarily by AMD Ryzen processor gross sales. Shopper phase working revenue was $89 million or 6% of income in comparison with an working lack of $69 million a yr in the past. Gaming phase income was $648 million, down 59% yr over yr and 30% sequentially.

The lower in income was primarily because of semi-custom stock digestion and the lower-end market demand. Gaming phase working revenue was $77 million or 12% of income in comparison with $225 million or 14% a yr in the past. Embedded phase income was $861 million, down 41% yr over yr as prospects continued to normalize their stock ranges. On a sequential foundation, embedded phase income was up 2%.

Embedded phase working revenue was $345 million or 40% of income in comparison with $757 million or 52% a yr in the past. Turning to the steadiness sheet and money movement. Through the quarter, we generated $593 million in money from operations, and free money movement was $439 million. Stock elevated sequentially by $339 million to $5 billion, primarily to assist the continued ramp of information middle GPU merchandise.

On the finish of the quarter, money, money equivalents, and short-term investments have been $5.3 billion. Within the second quarter, we returned $352 million to shareholders, repurchasing 2.3 million shares, and we have now $5.2 billion of authorization remaining. Through the quarter, we retired $750 million of debt that matured this previous June, using current money. Now, turning to our third quarter 2024 outlook.

We anticipate income to be roughly $6.7 billion, plus or minus $300 million. Sequentially, we anticipate income to develop roughly 15%, primarily pushed by sturdy development within the knowledge middle and consumer phase. We anticipate embedded phase income to be up and the gaming phase to say no by double-digit share. 12 months over yr, we anticipate income to develop roughly 16%, pushed by the steep ramp of our AMD Intuition processors and powerful server and consumer income development to greater than offset the declines within the gaming and the embedded phase.

As well as, we anticipate third quarter non-GAAP gross margin to be roughly 53.5%, non-GAAP working bills to be roughly $1.9 billion, non-GAAP efficient tax charge to be 13%, and the diluted share depend is anticipated to be roughly 1.64 billion shares. Additionally throughout the third quarter, we anticipate to shut the acquisition of Silo AI for about $665 million in money. In closing, we made important progress throughout the quarter towards attaining our monetary targets. We delivered file MI300 income that exceeded $1 billion and demonstrated stable traction with our next-gen Ryzen and EPYC product.

We expanded gross margin considerably and drove earnings development whereas rising funding in AI. Wanting ahead, the alternatives forward of us are unprecedented. We’ll stay targeted on executing to our long-term development technique whereas driving monetary self-discipline and operational excellence. With that, I will flip it again to Mitch for the Q&A session.

Mitch Haws — Vice President, Investor Relations

Thanks, Jane. John, we’re completely satisfied to ballot the viewers for questions.

Questions & Solutions:

Operator

Thanks, Mitch. We are going to now be conducting the question-and-answer session. [Operator instructions] One second, please, whereas we ballot for questions. And the primary query comes from the road of Ben Reitzes with Melius Analysis.

Please proceed together with your query.

Ben Reitzes — Melius Analysis — Analyst

Hey, thanks lots, and congratulations on these outcomes. Lisa, I needed to ask you about MI300, the way you see it enjoying out sequentially for the remainder of the yr. I suppose there’s about $2.8 billion left to hit your annual goal. So, I am questioning when you see issues choosing up within the fourth quarter and the way that is going sequentially.

And when you do not thoughts, I needed to additionally ask about subsequent yr when you see potential for fast development. You are in all probability conscious of a few of the chatter on the market, and I simply was questioning when you’re already seeing indicators which you could develop considerably, given your street map for subsequent yr. Thanks a lot.

Lisa T. Su — President and Chief Government Officer

Yeah. Nice, Ben. Thanks for the query. So, to begin with, on form of MI300 and the shopper evolution, we’re very proud of how MI300 has progressed.

After we began the yr, I believe the important thing level for us was to get our merchandise into our prospects’ knowledge facilities, to have them qualify their workloads, to actually ramp in manufacturing, after which see what the manufacturing capabilities are, particularly efficiency and all of these issues. And I can say now being form of greater than midway by means of the yr, we have seen nice progress throughout the board. As we glance into the second half of the yr, I believe we might anticipate that MI300 income would proceed to ramp within the third quarter and the fourth quarter. And we’re persevering with to increase each present deployments with our current prospects, in addition to we have now a big pipeline of consumers that we’re working by means of which are getting accustomed to our structure and software program and all that stuff.

So, I’d say general, very happy with the progress, and actually persevering with proper on observe to what we anticipated from the capabilities of the product. As we go into subsequent yr, I imply, one of many vital issues that we introduced at Computex was rising and increasing our street map. I believe we really feel actually good about our street map. We’re on observe to launch MI325 later this yr, after which subsequent yr, our MI350 Collection, which can be very aggressive with Blackwell options.

After which we’re properly on our approach to our CDNA Subsequent as properly. So, I believe, general, we stay fairly bullish on the general AI market. I believe the market continues to wish extra compute. And we additionally really feel excellent that our {hardware} and software program options are getting good traction, and we’re persevering with to increase that pipeline.

Ben Reitzes — Melius Analysis — Analyst

Thanks.

Operator

And the subsequent query comes from the road of Aaron Rakers with Wells Fargo. Please proceed together with your query.

Aaron Rakers — Analyst

Yeah. Thanks. Thanks for taking the query, and congrats on the quarter as properly. I suppose sticking on the information middle aspect, as we glance ahead and you consider the total yr, I am curious of the way you’re presently interested by the EPYC server CPU development expectations as we go ahead.

And any form of up to date ideas in your capability to form of proceed to realize share within the server market? Simply form of simply replace us on the way you see the server market enjoying out over the subsequent couple of quarters.

Lisa T. Su — President and Chief Government Officer

Yeah, certain, Aaron. Thanks for the query. So, you realize, we’re very happy with the progress that we have made with EPYC. I believe a few issues.

To begin with, when it comes to aggressive positioning and simply the traction out there, our fourth-gen EPYC between Gen 1 Bergamo is de facto doing very properly. We have seen broad adoption throughout cloud. After which we have been very targeted on enterprise in addition to third-party cloud cases. And as I stated within the ready remarks, we’re beginning to see very good traction in enterprise with each new prospects in addition to current prospects, after which for third-party cloud adoption, additionally pickup there as properly.

So, I believe general, I believe our EPYC portfolio has achieved properly. Going into the second half of the yr, I believe we additionally be ok with it. There are a few positives there. We see — to begin with, the market appears prefer it’s enhancing, so we have now seen some return to spending in each enterprise and cloud.

And so, I believe these are constructive market developments. After which along with that, we’re within the technique of launching Turin. So, we began manufacturing right here within the second quarter and we’re on observe to launch broadly within the second half of the yr. We’ll see some income of Turin within the second half of the yr contributing as properly.

So, general, I believe the server market and our capability to proceed to develop share within the server market is among the issues that we see within the second half of the yr.

Operator

And the subsequent query comes from the road of Timothy Arcuri with UBS. Please proceed together with your query.

Timothy Arcuri — Analyst

Thanks lots. Lisa, I needed to ask concerning the knowledge middle GPU street map. As you stated, 325 launching later this yr, so I suppose I had two questions. Does the better than $4.5 billion, does that embrace any income from 325? And might you speak somewhat bit extra about 350? Clearly, we’re seeing a giant rack scale or shift towards rack-scale programs for the competitors’s product.

And I am questioning if that is what 350 goes to appear like. Is it going to have liquid cooling and is it going to have a rack scale side to it? Thanks.

Lisa T. Su — President and Chief Government Officer

Yeah, completely. So, you realize, let me begin together with your unique query. I imply, I believe 325X, we’re on observe to launch later this yr. From a income standpoint, there can be a small contribution within the fourth quarter, but it surely actually continues to be principally the MI300 capabilities.

And 325 will begin within the fourth quarter after which ramp extra within the first half of subsequent yr. After which as we take a look at the 350 Collection, what we’re seeing and the rationale we name it a sequence is as a result of there can be a number of SKUs in that sequence that we’ll undergo the vary of, let’s name it, air-cooled to liquid-cooled. In spending time with our prospects. I believe there are individuals who actually need extra rack-level options, and we’re actually doing rather more when it comes to system-level integration for our merchandise.

You may see us make investments extra in system-level integration. However we even have many purchasers who wish to use their present infrastructure. I believe the fantastic thing about the MI350 sequence is it truly suits into the identical infrastructure because the MI300 sequence. And so, it might lend itself to, let’s name it, a fairly quick ramp when you’ve already invested in 300 or 325.

So, we see the vary of choices, and that is a part of the growth of the street map that we’re planning.

Operator

And the subsequent query comes from the road of Ross Seymore with Deutsche Financial institution. Please proceed together with your query.

Ross Seymore — Analyst

Hello. Thanks for letting me ask a query, and congrats on the sturdy outcomes. Effectively, knowledge middle is clearly crucial. I simply wish to pivot to the consumer aspect.

Lisa, are you able to speak concerning the AI PC aspect of issues, the way you imagine AMD is positioned? Are you seeing any aggressive depth altering with the emergence of ARM-based programs? Simply needed to see the way you’re anticipating that to roll out and what it means to second-half seasonality.

Lisa T. Su — President and Chief Government Officer

Yeah. Positive, Ross. So, first, you realize, we’re very happy with our consumer enterprise outcomes. I believe we have now a really sturdy street map, so I am very happy with the street map.

The Zen 5-based merchandise, we’re launching each pocket book and desktop on this — in the course of this yr. What we have seen is definitely very constructive suggestions on the product. So, we simply truly launched the primary Strix-based notebooks over the weekend. They went on sale.

You’ll have seen a few of the critiques. The critiques are very constructive. Our view of that is the AI PC is a crucial add to the general PC class. As we go into the second half of the yr, I believe we have now higher seasonality usually, and we expect we will do, let’s name it, above-typical seasonality, given the power of our product launches and after we’re launching.

After which into 2025, you are going to see AI PCs throughout kind of a bigger set of value factors, which may even open up extra alternatives. So, general, I’d say the PC market is an effective income development alternative for us. The enterprise is performing properly. The merchandise are sturdy.

And we’re working very intently with each the ecosystem companions in addition to our OEM companions to have sturdy launches right here into the second half of the yr.

Ross Seymore — Analyst

And is the ARM aspect altering something or not likely?

Lisa T. Su — President and Chief Government Officer

You understand, look, I believe at this level, the PC market is a giant market, and we’re underrepresented out there. I’d say that we take all of our competitors very significantly. That being the case, I believe our merchandise are very well-positioned.

Operator

And the subsequent query comes from the road of Matt Ramsay with Cowen. Please proceed together with your query.

Matt Ramsay — Analyst

Thanks very a lot. Good afternoon. Lisa, I needed to perhaps draw a parallel between the Intuition portfolio that your organization is rolling out now and what you guys did 5 – 6 years in the past with EPYC. And I bear in mind when the Naples product launched, there was lots of, I’d say, response positively and negatively and form of sentiment round the place your street map would possibly go to comparatively small perturbations in what the volumes have been.

Tremendous early, but when I bear in mind again to that, what was a very powerful was that was the toehold into the marketplace for long-term engagement, each on the software program aspect and the {hardware} aspect together with your prospects two, three, 4 generations ahead. So, is that an correct parallel to the place you guys are with MI300? And perhaps you could possibly speak concerning the degree of engagement, the depth of engagement, the breadth of it throughout the shopper base with 350 and 400. Thanks.

Lisa T. Su — President and Chief Government Officer

Yeah, completely, Matt. So, look, as I stated earlier, we’re very happy with the progress that we’re making on the Intuition street map. That is completely a long-term play so completely, you are appropriate. It has lots of parallels to the EPYC journey, the place you actually need to — you achieve extra alternatives, broader workloads, bigger deployments as you go from era to era.

So, we’re enjoying the lengthy sport right here. Our conversations with our prospects, so I’d begin with first, within the close to time period, we had some very key milestones that we needed to go this yr. And as I stated, they associated to getting {hardware} in quantity in a number of hyperscalers in addition to giant Tier 2 prospects. We have achieved that.

We have now seen our software program in lots of completely different environments, and it is matured considerably. ROCm is in very, from a standpoint of options, features, out-of-box efficiency, attending to efficiency with prospects, we have gained lots of confidence and realized lots in that complete course of. The networking points of constructing out the rack scale and the system-level objects are areas that we’re persevering with to spend money on. After which the purpose of getting long-term conversations throughout a number of generations can be actually vital.

So, I believe all of these issues have progressed properly. We view this as excellent progress for MI300, however we have now much more to do. And I believe the assorted street maps will assist us open up these alternatives over the subsequent couple of years.

Matt Ramsay — Analyst

Recognize it. Thanks.

Lisa T. Su — President and Chief Government Officer

Thanks, Matt.

Operator

And the subsequent query comes from the road of Vivek Arya with Financial institution of America Securities. Please proceed together with your query.

Vivek Arya — Analyst

Thanks for taking my query. Lisa, there appears to be this ongoing {industry} debate about AI monetization and whether or not your prospects are getting the fitting ROI on their capex. And at present, they’ve these three choices, proper? They will purchase GPUs out of your largest competitor with all of the software program bells and whistles and incumbency, or they will do {custom} chips, or they will purchase from AMD. So, how do you suppose this performs out subsequent yr? Do you suppose your prospects, given all this concern round monetization, does it make them consolidate their capex round simply the opposite two suppliers? How is your visibility going into subsequent yr, given this {industry} debate? And the way will AMD proceed to form of carve a place between these two different aggressive decisions which are on the market? Thanks.

Lisa T. Su — President and Chief Government Officer

Yeah, certain, Vivek. Effectively, I imply, I believe you speak to lots of the identical those who we speak to. I believe the general view on AI funding is we have now to speculate. I imply, the {industry} has to speculate.

The potential of AI is so giant to impression the best way enterprises function and all that stuff. So, I believe the funding cycle will proceed to be sturdy. After which relative to the assorted decisions for the dimensions of the market, I firmly imagine that there can be a number of options, whether or not you are speaking about GPUs otherwise you’re speaking about {custom} chips or ASICs, there can be a number of options. In our case, I believe we have demonstrated a extremely sturdy street map and the flexibility to accomplice properly with our prospects.

And from the standpoint of that deep engagement, {hardware}, software program co-optimization is so vital in that. And for big language fashions, GPUs are nonetheless the structure of selection. So, I believe the chance may be very giant. And I believe our piece of that’s actually sturdy know-how with sturdy partnerships with the important thing AI market makers.

Vivek Arya — Analyst

Thanks, Lisa.

Lisa T. Su — President and Chief Government Officer

Thanks, Vivek.

Operator

And the subsequent query comes from the road of Joe Moore with Morgan Stanley. Please proceed together with your query.

Joe Moore — Analyst

Nice. Thanks. I additionally needed to ask about MI300. I ponder when you may discuss coaching versus inference.

Do you could have a way — I do know that lots of the preliminary focus was inference, however do you could have traction on the coaching aspect? And any sense of what that cut up could appear like over time?

Lisa T. Su — President and Chief Government Officer

Yeah, certain. Thanks for the query, Joe. So, as we stated on MI300, there are many nice traits about it. One in all it — one among them is our reminiscence bandwidth and reminiscence capability is main the {industry}.

From that standpoint, the early deployments have largely been inference generally, and we have seen incredible efficiency from an inference standpoint. We even have prospects which are doing coaching. We have additionally seen that from a coaching standpoint, we have optimized fairly a bit our ROCm software program stack to make it simpler for individuals to coach on AMD. And I do anticipate that we’ll proceed to ramp coaching over time.

As we go ahead, I believe you will see — the assumption is that inference can be bigger than coaching from a market standpoint. However from an AMD standpoint, I’d anticipate each inference and coaching to be development alternatives for us.

Joe Moore — Analyst

Nice. Thanks.

Operator

And the subsequent query comes from the road of Toshiya Hari with Goldman Sachs. Please proceed together with your query.

Toshiya Hari — Analyst

Hello. Thanks a lot for taking the query. I had a query on the MI300 as properly. Curiously, when you’re presently delivery to demand or if the up to date annual forecast of $4.5 billion is in some form or kind provide constrained, suppose final quarter, you gave some feedback on HBM and CoAs.

Curious when you may present an replace there. After which my Half B to my query is on profitability for MI300. I believe prior to now, you’ve got talked concerning the enterprise being accretive and enhancing additional over time as you form of work by means of the kinks, if you’ll. Has that view advanced or modified in any respect given form of the aggressive depth and your want to speculate, whether or not it’s by means of natural R&D or a few of the acquisitions you’ve got made? Or are you continue to assured that revenue margins within the enterprise proceed to increase? Thanks.

Lisa T. Su — President and Chief Government Officer

Yeah. Positive, Toshiya. Thanks for the query. So, on the provision aspect, let me make a few feedback after which perhaps I will let Jean touch upon form of the trajectory for the enterprise.

So, on the provision aspect, we made nice progress within the second quarter. We ramped up provide considerably exceeding $1 billion within the quarter. I believe the staff has executed very well. We proceed to see line of sight to proceed rising provide as we undergo the second half of the yr.

However I’ll say that the general provide chain is tight and can stay tight by means of 2025. So, underneath that backdrop, we have now nice partnerships throughout the provision chain. We have been constructing extra capability and functionality there. And so, we anticipate to proceed to ramp as we undergo the yr.

And we’ll proceed to work each provide in addition to demand alternatives, and actually that is accelerating our buyer adoption general, and we’ll see how issues play out as we go into the second half of this yr.

Jean Hu — Government Vice President, Chief Monetary Officer

Yeah. In your second query concerning the profitability, first, our staff has achieved an incredible job to ramp the product MI300. It is a very advanced product so we ramped it efficiently. On the identical time, the staff additionally began to implement operational optimization to proceed to enhance gross margin.

So, we proceed to see the gross margin enchancment. Over time, in the long run, we do imagine gross margin can be accretive to company common. From a profitability perspective, AMD at all times invests in platforms. If you happen to take a look at our knowledge middle platform, particularly each the server and the information middle GPU aspect, we’re ramping the income.

The enterprise mannequin can leverage very considerably even from GPU aspect. As a result of the income ramp has been fairly important, the working margin continued to increase. We undoubtedly wish to proceed to speculate as the chance is large. On the identical time, it’s a worthwhile enterprise already.

Toshiya Hari — Analyst

Thanks very a lot.

Operator

And the subsequent query comes from the road of Stacy Rasgon with Bernstein Analysis. Please proceed together with your query.

Stacy Rasgon — Analyst

Hello, guys. Thanks for taking my query. I needed to dig into the Q3 steerage somewhat bit if I may. So, with gaming down double digits, it in all probability means you’ve got obtained near $1 billion of development income throughout knowledge middle, consumer, and embedded.

I used to be questioning when you may give us some colour on how that $1 billion-ish splits out throughout these three companies. Like if I had 70% of it going to knowledge middle and 20% going to consumer and 10% going to embedded, like would that be like approach off? Or how ought to we take into consideration that apportioning out throughout the segments?

Lisa T. Su — President and Chief Government Officer

Yeah. Possibly Stacy, let me provide the following colour. So, the gaming enterprise is down double digit as you state. Consider it as the information middle is the most important piece of it, consumer, subsequent.

After which on the embedded aspect, consider it as single-digit sequential development.

Stacy Rasgon — Analyst

Acquired it. So, I imply, inside that knowledge middle piece then, how does that cut up out? I imply, is the majority of the information development Intuition? Or is it form of equally weighted between Intuition and EPYC? Or like once more, how does it — once more, when you obtained, I do not know, $400 million to $600 million of sequential knowledge middle development or one thing like that, how does it cut up up?

Lisa T. Su — President and Chief Government Officer

Yeah. So, once more, with out being that granular, we’ll see each — actually, the Intuition GPUs will develop, and we’ll see additionally very good development on the server aspect.

Operator

And the subsequent query comes from the road of Harsh Kumar with Piper Sandler. Please proceed together with your query.

Harsh Kumar — Analyst

Yeah. Hey, Lisa, from my rudimentary understanding, the big distinction between your Intuition merchandise and the adoption versus your nearest competitor is form of rack degree efficiency and that rack degree is the construction that you could be be missing. You talked somewhat bit about UALink. I used to be questioning when you may increase on that and provides us some extra colour on when that may — when that hole may be closed.

Or is that this a serious step for the {industry} to shut that hole? Simply any colour could be appreciated.

Lisa T. Su — President and Chief Government Officer

Yeah. So, Harsh, general, perhaps if I take a step again and simply discuss how the programs are evolving, there is not any query that the programs are getting extra advanced, particularly as you go into giant coaching clusters, and our prospects need assistance to place these collectively. And that features the form of Infinity Material-type options which are the idea for the UALink issues in addition to simply basic rack-level system integration. I believe what you must anticipate, Harsh is, to begin with, we’re very happy with the entire companions which have come collectively for UALink.

We predict that is an vital functionality. However we have now the entire items of this already inside form of the AMD umbrella with our Infinity Material, with the work with our networking functionality by means of the acquisition of Pensando. And you then’ll see us make investments extra on this space. So, that is a part of how we assist prospects get to market sooner is by investing in the entire parts, so the CPUs, the GPUs, the networking functionality in addition to system-level options.

Harsh Kumar — Analyst

Thanks, Lisa.

Lisa T. Su — President and Chief Government Officer

Thanks, Harsh.

Operator

And the subsequent query comes from the road of Blayne Curtis with Jefferies. Please proceed together with your query.

Blayne Curtis — Analyst

Hey, good afternoon. Thanks for taking my query. I simply wish to ask one other query on MI300. Simply curious when you can form of characterize the make-up of the purchasers within the first half.

I do know you had, finish of final yr, a authorities buyer. Is there nonetheless a authorities contingency? And form of the second a part of it’s actually you’ve got invested in all these software program property. Form of curious the problem of ramping the subsequent wave of consumers. I do know there’s been lots of speak on some {hardware} challenges, reminiscence points, and such, however you then’re investing in software program.

I am certain that is a giant problem, too. Simply form of curious what the most important hurdle is so that you can form of get that subsequent wave of consumers ramp.

Lisa T. Su — President and Chief Government Officer

Yeah. So, Blayne, lots of items to that query so let me attempt to deal with them. First, in your query about, I believe you are mainly asking concerning the supercomputing piece. That was primarily This autumn and a bit in Q1.

So, if you consider our Q2 income, give it some thought as virtually all AI. So, it is MI300X, it is for big AI, hyperscalers in addition to OEM prospects going to enterprise and Tier 2 knowledge facilities. So, that is the make-up of the shopper set. After which when it comes to the assorted items of what we’re doing, I believe first in your query about reminiscence, I believe there’s lots of noise within the system.

I would not actually take note of all that noise within the system. I imply, this has been an unimaginable ramp. And I am truly actually pleased with what the staff has achieved when it comes to simply undoubtedly quickest product ramp that we have ever achieved to $1 billion right here within the — over $1 billion within the second quarter after which ramping every quarter in Q3 and This autumn. When it comes to reminiscence, we have now a number of suppliers that we have certified on HBM3.

And it is a tough — reminiscence is a difficult enterprise, however I believe we have achieved it very properly and that is there. After which we’re additionally qualifying HBM3E for future merchandise with a number of reminiscence suppliers as properly. So, to your overarching query of what are the issues that we’re doing, the thrilling a part of that is that the ROCm functionality has actually gotten considerably higher as a result of so many purchasers have been utilizing it. And with that, what we take a look at is out-of-box efficiency, how lengthy does it take a buyer to stand up and working on MI300? And we have seen, relying on the software program that corporations are utilizing, notably when you’re based mostly on a few of the higher-level frameworks like PyTorch, and so forth., we will be out-of-the-box working very properly in a really brief period of time, like, let’s name it, very small variety of weeks.

And that is nice as a result of that is increasing the general portfolio. We’re going to proceed to spend money on software program, and that was one of many causes that we did the Silo AI acquisition. It is an excellent acquisition for us. 300 scientists and engineers.

These are engineers which have expertise with AMD {hardware} and are very, excellent at serving to prospects stand up and working on AMD {hardware}. And that is — so we view this as the chance to increase the shopper base with expertise like Silo AI, like Nod.ai which introduced lots of compiler expertise. After which we proceed to rent fairly a bit organically. So, I believe Jean stated earlier that we see leverage within the mannequin, however we’ll proceed to speculate as a result of this chance is large, and we have now the entire items.

That is nearly constructing out scale.

Blayne Curtis — Analyst

Thanks a lot.

Lisa T. Su — President and Chief Government Officer

Thanks.

Operator

And the subsequent query comes from the road of Tom O’Malley with Barclays. Please proceed together with your query.

Tom O’Malley — Barclays — Analyst

Hey, Lisa. Thanks for taking my query. I will offer you a breather from the MI300 for a second, however simply to deal with consumer within the second half. No drawback.

Targeted on consumer within the second half, you form of stated above seasonal for September, December. You are clearly launching a brand new pocket book, desktop product, however you are additionally speaking about AI PC. Might you simply break down the place you are seeing these above-seasonal developments? Is it the ASP uplift you are getting from the brand new merchandise? Is it a unit assumption that is coming with AI PC? Simply any form of breakdown between these two and why you are seeing it somewhat bit higher. Thanks.

Lisa T. Su — President and Chief Government Officer

Positive, Tom. So, I believe you truly stated it properly. We’re launching Zen 5 desktops and notebooks with quantity ramping within the third quarter. And that is the first purpose that we see above seasonal.

The AI PC component is actually one component of that, however there’s simply the general refresh. Normally, desktop launches going into a 3rd quarter are good for us, and we really feel that the merchandise are very well-positioned. So, these are the first causes.

Operator

And our remaining query comes from the road of Chris Danely with Citi. Please proceed together with your query.

Christopher Danely — Analyst

Hey, gang, thanks for sneaking me in. Only a query on gross margin. So, if we take a look at your steerage, it looks as if the incremental gross margin is dropping somewhat bit for Q3. Why is that occuring? After which only a follow-up on one other a part of the gross margin angle.

Have you ever modified your gross margin expectations for the MI300? Has the accretion level moved out somewhat bit?

Jean Hu — Government Vice President, Chief Monetary Officer

Yeah, Chris, thanks for the query. I believe, first, we have now made lots of progress, as you talked about, this yr to increase our gross margin from 2023 at a 50 share level to — you realize, we truly guided 53.5% for Q3. The first driver is de facto the sooner knowledge middle enterprise development. If you happen to take a look at the information middle enterprise as a share of income from 37% in This autumn final yr to now near 50%.

That sooner growth actually helped us with the gross margin. If you take a look at the second half, we’ll proceed to see knowledge middle to be the main driver of our top-line income development, will assist with the margin growth. However there are another places and takes. I believe Lisa talked about the PC enterprise truly goes to do higher in second half, particularly usually, seasonally, it tends to be extra consumer-focused.

So, that actually is somewhat bit completely different dynamics there. Secondly, I’d say embedded enterprise, we’re going to see embedded enterprise to be up sequentially every quarter. However the restoration, as we talked about earlier, is extra gradual. So, whenever you take a look at the steadiness of the image, that is why we see the gross margin — the tempo of the gross margin modified somewhat bit, however we do see continued gross margin growth.

So far as MI300, we’re fairly assured over the long run, it will likely be accretive to our company common. We really feel fairly good concerning the general knowledge middle enterprise to proceed to be completely the motive force of gross margin growth.

Christopher Danely — Analyst

Thanks.

Operator

I want to flip the ground again over to Mitch for any closing feedback.

Mitch Haws — Vice President, Investor Relations

Nice. That concludes at present’s name. Because of all of you for becoming a member of us at present.

Operator

[Operator signoff]

Length: 0 minutes

Name individuals:

Mitch Haws — Vice President, Investor Relations

Lisa T. Su — President and Chief Government Officer

Jean Hu — Government Vice President, Chief Monetary Officer

Ben Reitzes — Melius Analysis — Analyst

Lisa Su — President and Chief Government Officer

Aaron Rakers — Analyst

Timothy Arcuri — Analyst

Ross Seymore — Analyst

Matt Ramsay — Analyst

Vivek Arya — Analyst

Joe Moore — Analyst

Toshiya Hari — Analyst

Stacy Rasgon — Analyst

Harsh Kumar — Analyst

Blayne Curtis — Analyst

Tom O’Malley — Barclays — Analyst

Christopher Danely — Analyst

Chris Danely — Analyst

Extra AMD evaluation

All earnings name transcripts