AMD(AM D.O) released its second quarter 2023 financial report (as of June 2023) after the U.S. stock market on the morning of August 2, 2023, Beijing time. The main points of the call are as follows:! 1. $ AMD(AMD.US) Call Increment **1) Data Center: Annual growth may be closer to single digits * *. This is very strong growth. The third quarter may mean strong double-digit growth, and the fourth quarter will continue to grow strongly. * * 2) Gross profit margin : The gross profit margin outlook for the third quarter is 51%. Despite the great resistance brought by the rise of embedded business in the third quarter, the gross profit margin has increased by more than one percentage point in a row. * * The fourth quarter will continue to improve * *; * 3) Inventory cycle: **Longer manufacturing cycle due to advanced process technology (5 nm). So in the long run, the company's inventory cycle will be extended ; 4) The scale of generative AI: It may grow at a positive or negative CAGR of 50% in the next three to four years. By 2027, our market size will reach $150 billion. Original text of the 2. call 2.1 Management Overview We performed well in the second quarter, launching multiple leading products, significantly expanding our AI engagement, and accelerating our latest investment in epic and emerging product lines. Second-quarter revenue fell 18% year-over-year to $5.4 billion. Sales were flat on a sequential basis as growth in our customer and data center segments was offset by lower expectations in our gaming and embedded segments. As multiple customers initiated or expanded projects to support future large-scale deployments of Instinct, MI250, and MI300 hardware and software, the number of AI customer engagements increased more than 7-fold in a row. Data center segment 1.3 billion revenue declined 11% year-over-year and grew 2% quarter-over-quarter, although market demand remained mixed, with adoption of fourth-generation EPYCCPU accelerating in the quarter, with revenue doubling almost continuously as cloud providers expanded deployments to support their on-premises infrastructure and public instance offerings. In terms of cloud computing, 30 new AMD instances were launched in the second quarter, with AWS, Alibaba, Microsoft and Oracle announcing multiple Genoa instances. Genoa's performance for enterprise and cloud applications is 1.9 times higher than the competition, and its performance per watt is 1.8 times higher than the competition, making it by far the highest performance and most efficient server processor in the industry. AWS has released M7 Universal Instances, the highest performing and most cost-effective log server since they were offered. In total, there are now more than 670 AMD-powered cloud instances publicly available, and we expect that number to grow 30% to nearly 900 by the end of this year, driven primarily by new Genoa deployments. We also expanded our Zen4 server portfolio during the quarter with the launch of Bergamo and GenoaX. Microsoft Azure announces the first Genoaxhpc instances that deliver more than five times better performance for technical compute workloads than the previous generation. With Bergamo, we deliver more than twice the performance for cloud-native applications while providing full x86 software compatibility. We are thrilled to have meta join us at the launch as they announce plans to deploy Bergamo widely across their global data center infrastructure to support applications including Facebook,Instagram and WhatsApp. Looking ahead, large server providers such as Dell, Lenovo and Super Micro are expected to start launching their new Bergamo promotion platforms in the third quarter. On the enterprise side, although macroeconomic uncertainty has led to weak year-on-year customer demand, sales of EPYC processors for enterprise servers are also increasing as we approach the end of many victories with large energy, technology, financial services and healthcare companies. The overall appeal from large enterprises continues to grow. **We expect EPYC's revenue to achieve double-digit sequential growth in the third quarter, led by the growth of the fourth-generation EPYC CPU. **In addition, Sienna, our first EPYC processor optimized for leadership edge servers and telco infrastructure, is also expected to be available this quarter. Data Center On the network side, the largest cloud providers expanded their adoption of Pensando DPUs during the quarter, which was highlighted by the new deployments of Alibaba and Oracle Cloud. When it comes to supercomputing, EPYC and instinct processors continue to be the solution of choice for the most powerful supercomputers, with the 121 fastest systems on the latest Top 500 list using EPYC and instinct processors. On AI, we made strong progress in the second quarter as we reached key hardware and software milestones to address growing customer demand for our data center AI solutions. Our AI strategy focuses on three areas. First, a broad portfolio and multi-generation roadmap of leading GPU, CPU and adaptive computing solutions for AI inference and training. Second, extend the open and proven software platform we 've already built so that our AI hardware can be widely and easily deployed and used. Third, expand our deep partnerships across the ecosystem to accelerate the large-scale deployment of AMD-based AI solutions. These are the key factors that affect the performance of AI reasoning. Customer interest in our instinctive MI300A and MI300Xgpu is very high. Partnering with top cloud providers, large enterprises, and many leading AI companies expanded significantly during the quarter. **We are now providing early system access and product samples to our key AI, high-performance computing and cloud customers, and are on schedule to launch and phase-in production in the fourth quarter * *. Client Customer segment, revenue declined 54% YoY to 1 billion. Ryzen 7000 series CPU sales increased significantly and client revenue increased 35% sequentially as the largest original equipment manufacturer introduced new laptops. We also introduced new commercial products, including our first rise and Pro laptop and desktop processors, powered by our leadership Zen4 framework. HP, Lenovo and other leading OEMs will launch more than 100 AMD-powered commercial PC platforms this year. As we develop this important part of our customer business, we expect that our customer segments will experience strong seasonal growth in the second half of the year based on the strength of our product portfolio and our increasing adoption rate of 7000 CPUs, including our increasing 7040 mobile CPUs, which provide leading performance and energy efficiency, and are the industry's first x86 processors with dedicated AI engines. Going forward, as Microsoft and other large software providers incorporate generative AI into their products, we believe AI will be an important driver of PC demand. We are executing on a multi-generational rise roadmap for AI processors that, together with our ecosystem partners, will fundamentally change the PC experience. Gaming Gaming segment, revenue decreased 4% year-over-year to $1.6 billion, as higher semi-custom revenue was offset by a decline in game graphics sales. Subsequently, segment revenue fell by 10%. Semi-custom SOC sales were strong in the quarter as Microsoft and Sony saw healthy demand for consoles on the back of improved global retail availability and new 3A game launches. In terms of game graphics, we expanded our Radeon7000GPU series in the second quarter with the introduction of mainstream RX7600 graphics cards for 1080p games. We will launch new hobbyist radeon7000 series graphics cards in the third quarter, further expanding our RDNA3 GPU offering. Embedded Embedded segment, revenue increased 16% YoY to $1.5 billion. Weak communications customers offset strong demand for industrial vision and healthcare, automotive and broadcast customers as some operators slowed infrastructure upgrades, with revenue down 7 percent. During the quarter, we expanded our leading adaptive computing portfolio with the launch of the new general-purpose Premium VP19O2 Adaptive SOC in an advanced chip package, the industry's largest and highest-performing solution for simulating and validating next-generation ASICs and SOCs. At the low end, we announced the Spartan Ultrascale plus FPGA family to address a range of new cost-optimized industrial, computer vision, healthcare, and robotics applications. We have also released enhanced versions of the Vivato and Vitas software platforms, making it easier for customers to develop high-performance applications for our universal adaptive Soc. Embedded CPU sales increased during the quarter, driven by new AMD security, storage, and networking solutions from HPE, Fortinet, and other leading vendors. Looking ahead to the second half, after six quarters of very strong year-over-year growth, **we expect embedded segment revenue to decline in the second half as lead times return to normal and some customers reduce inventory levels. **We continue to be very pleased with the winning momentum of embedded design, especially based on our adaptive and embedded processing product portfolio, we see growing revenue synergy opportunities. In summary, in line with our strategic priorities, we executed well this quarter. Looking ahead to the second half of the year, we expect seasonal growth in the PC market and more normalized inventory levels across the supply chain. In the data center market, with the expansion of AI deployment, we see a hybrid environment. However, cloud customers continue to optimize their data center computing, and enterprise customers remain cautious about new deployments. * * 3Q performance outlook: * * expected revenue of $54-6 billion, up 2.5 percent year-over-year. The client and data center segments are expected to grow at double-digit percentages, with the data center slightly flat compared to the client, while the gaming and embedded segments will decline. Non-GAAP gross profit margin is about 51%, Non-GAAP operating expenses are 1.65 billion US dollars, Non-GAAP effective tax rate is 13%, diluted shares are 1.63 billion shares * * 2.2Q & A **** Q1: Regarding the data center business, where is the driving force for H2 to increase 50% over H1? What is the follow-up outlook? **A1: On the positive side, we certainly see an acceleration in AI demand. We look at this problem from several aspects. We 've had a lot of design successes in AI deployment, and as you know, CPUs are used with GPUs and other accelerators. So in the head node, we see the positive side. We have also seen some strong interest in our MI250, which is currently scheduled for shipment. We believe that the MI300 accelerator, which starts production in the fourth quarter, will receive a very strong pull. These are positive market dynamics going into the second half of the year. We 've also seen a reduction in cloud spending outside of AI as some cloud providers are optimizing their capex. What I want to say is that the enterprise side is still in a weak position. With all of this in place, **we expect the data center business to grow significantly in the second half of the year and tilt towards the fourth quarter. **We are still looking for a 50% growth in the second half to the first half. So, this is a big slope, but when we look at all the components, I think the customer appeal is definitely there, and it is exciting to be in this part of the industry. Q2: What is the customer feedback on the performance of the hardware itself? How will the software work done in the next year translate into deployments for other customers? A2: There is a very strong interest in how customer engagement is going on in our AI solutions. This includes, let's call it, multiple level hyper scalers in which we are involved. It includes some large enterprises, but also some new categories of AI-centric companies that are very forward-looking in how they deploy and build AI solutions. From this perspective, we have made a lot of progress on the Rockham software stack. In fact, there is still a lot of work to be done, but I would like to say that the progress we have made is very important. We get a lot of feedback from those major customers. We see the benefits of optimization. So we're also working on higher-level model frameworks, and we're working with the Pytorch Foundation, with Triton. The point is, we're getting important real-time feedback from some of our major customers. So we learn very fast. As far as performance feedback is concerned, many companies are now able to adapt the MI250 to a wide range of workloads. This is a good response, and the MI300 feedback is also quite positive. We let customers sample in our lab system, they can access the hardware, or sample in their lab. I would say, so far, very positive. While there is still a lot of work to do, **we are very pleased with the progress of our overall AI solution for the data center. **** Q3: How do you look at the data center business's implied growth expectations in 2H20? **A3: Enterprise market continues to be weak and cloud optimization is still in progress. **So overall, we think the annual growth rate may be closer to single digits * *. This is very strong growth. The third quarter could mean strong double-digit growth. Of course, we will also see continued strong growth in the fourth quarter. **A4: We have been investing in our supply chain, and the development of data centers is very strategic for us, which has become part of our strategy. If you look at all aspects of the supply chain, from wafers to back-end capabilities, to some specific components, you need to do some MI300 levels. We work with the entire supply chain. We believe we have sufficient supply to deliver positive growth in Q4 and 2024. But that's certainly one of the areas where we 've spent quite a bit of time making sure that we do have confidence. Q5: Of the implied growth rate of data centers in the fourth quarter, can you talk about what percentage of it is over-calculated? How the contribution is split? A5: Entering the third quarter, we expect data centers to have double-digit continuous growth. It's basically an epic increase. This is mainly due to the Zen 4 (we call it a combination of general Bergamo) as we continue to grow into the fourth quarter, bringing in an implied significant increase in revenue. I think there are multiple components to this. Server CPUs will continue to rise, as we 've seen with Zen 4 growth. The tailwind of overcalculation is mainly in the fourth quarter, and there will be a little in the first quarter. Then we will have MI300x contributions to large AI customers as they start their initial ramp, and MI250, which these customers have now seen as a very, very good choice for some workloads that are not necessarily the big language model or the biggest parameters, let's call other AI workloads. So these are all components of the implied growth in the fourth quarter. There are many aspects, but it is clear that a large part of them are brought about by MI 300. **Q6: On the customer side, operating margins remain well below 21 and 22 year levels. Can you talk about the competitive landscape of the customer business? Is there a way back to 20%, 30% operating margins? **A6: The PC business has been very volatile over the past few quarters, and we 've been dealing with some inventory digestion since the peak of the epidemic until now. I'm happy to say that the growth we saw in the second quarter and the growth we saw in the second half of the year is the strength of our product portfolio. I think Ryzen7000 series did well. There is a good customer attraction. I think the business has been competitive from a competitive dynamics standpoint, but we feel good about the most important point-the little drag on operating margins is due to low revenue and destocking, so we're normalizing inventory levels in the supply chain. As we overcome these difficulties, I think what we see is that the customer business continues to grow. We believe customer growth will also go to 2024. In terms of some cost initiatives, we have been optimizing. On the Opex side, the team has done a good job of really optimizing the customer's investment in the market segment to make it more efficient, patient and effective. If you * * look at the overall company level, our operating costs are basically flat, but we are investing in artificial intelligence, data centers and strategic priorities, and the return on investment in these areas is much higher * *. We are reasonably comfortable with this level of operating expenses and can continue to invest in our customer space. This model will generate profits, and we should be able to return to a 20% profit margin. Q7: Do you expect the enterprise sector to contribute to the strong growth of the data center business in the second half of the year? A7: The enterprise business is very strategic for us. We feel we are underrepresented and that's where we put more resources. When we looked at Genoa's value proposition across the entire Zen 4 portfolio, we thought it worked well in the business. We are pleased to see the growth in the second quarter. We do believe that we are on a long-term path of sustained growth. The key here is also to invest in some of the activities that go to market. So the investment is more of a business developer who can visit these enterprise customers directly with our OEM partners and make sure our value proposition is well understood. Q8: There is still a lot of unmet demand for accelerated computing in China, have you considered developing a China-specific version for their MI250 or new MI300 platform? A8: China is a very important market for us, and in our portfolio, we will definitely consider the accelerator market. Of course, our plan is to fully comply with U.S. export controls, but we do believe that we have an opportunity to develop products for Chinese customers who are looking for AI solutions and will continue to work in that direction. Q9: Is it reasonable to assume that your GPU accelerator sales are about 0.5 billion this year, or 7.8 percent of your data center sales? If that number is correct, does that mean that your server CPU sales are actually the same as last year? A9: What we're saying is that **GPU sales in the first half of the year were very low because we're in a product transition period. As we move into the second half, especially the fourth quarter, the MI300 will have a slope. **I think your numbers may be a bit high in terms of GPU sales. But overall, I think our expectation is that the data center business is growing in the high single digits year over year. Compared to the first half, we saw a much better second half. I think the product portfolio and the development of Genoa and Bergamo, as well as the development of MI300, are key components of our development in the second half of the year. Q10: What is AMD's product position, value proposition and strategy regarding the market for AI accelerators? A10: Our investments in AI are very broad. I know there are a lot of people interested in data centers, but I don't want us to lose our investment in edges and clients. To your question, what is our value proposition in the data center, I think we have shown that, as you mentioned, we have strong capabilities in supercomputing. When you look at artificial intelligence, there are many different types of artificial intelligence. If you look at training and reasoning, the biggest language models and what drives the performance of some of them. When we look at MI 300, MI 300 is actually designed as a highly flexible product family that covers all these different market segments. In particular, we see a lot of interest in large-scale language model reasoning. The mi300x has the highest memory bandwidth and the highest memory capacity. If you look at the reasoning workload, it's actually very, very dependent on these things. Having said that, we also believe we have a strong value proposition in training as well. When you look at the amount of work and the investments we're making, not just today, but our next generation of MI 400 series and so on. We strongly believe that our hardware roadmap is very competitive and capable. Frankly, I think the discussion about AMD has always been about the software roadmap. We did see some changes in software, and we put a lot of resources into it. So by bringing together our former Xilink software team and AMD's software team, we 've greatly increased our resources and now the focus is on optimizing on these higher-level models. If you think about the frameworks of Pytorch, Triton, and Onix, I think a lot of AI-centric new companies are actually optimizing on a different level, and they're working closely with us in the very exciting area of AI. I think there will be multiple winners and we will start by saying there are multiple winners. But we think our product portfolio is actually quite unique, because we do have cpu, gpu, we have rise and AI accelerator technology on the PC side, and we also have embedded technology in our Xilinx product portfolio. So I think it's a pretty broad and capable portfolio. Q11: Regarding the Q4 data center guidance, what is the proportion of MI300 and CPU in data center spending? Given the instability of EI Capitan, what does this mean for the potential seasonality of the first quarter? A11: I think the largest part of the * * fourth quarter growth is the upgrade of MI300, but there is also an important component, that is, the upgrade of processors-Zen 4 product portfolio. **Regarding the instability of income and the situation until 2024, let me tell you a few parts. **I think there was a question about how much of the MI300's revenue is AI-centric, or is it supercomputing-the bigger part is supercomputing, but it's a meaningful revenue contribution. **In terms of artificial intelligence, our expectation by 2024 is that customer interest in the MI300X is very high. Many customers want to deploy as soon as possible. So as we go into the first half of 2024, we expect early deployments and then we expect more numbers in the second half of 2024. As you know, these things are perfectly qualified. So we're going to have some difficulties in the next few quarters. But our visibility is such that many customers want to deploy as soon as possible. We are working closely with them to carry out the necessary co-engineering to get them into production. **The volume of EI Capitan is in the hundreds of millions of dollars. **** Q12: What is the outlook for follow-on gross margin? A12: Gross margin is the main driver we have discussed in the past, if you look at our guidance or outlook for the third quarter, gross margin was 51%, despite the very strong headwinds from the rise in the embedded business in the third quarter, gross margin improved by more than a percentage point in a row. As a result, the data center and client business is expected to grow by double digits in a row and have a positive impact on gross margin, which actually outweighs the negative impact from the embedded business. Looking at Q4 again, we are not guiding Q4 it depends on the mix. What I'm saying is you're going to have similar dynamics, right? The data center growth is very significant. At the same time, we will also face the adverse impact of the continuous decline in embedded business. **Overall, we do expect gross margin to continue to improve from current levels * *. Q13: About the decline in the embedded business in the second half of the year and what are the drivers? A13: When I look at the embedded business, I think we should remember that we are experiencing six quarters of strong growth. I mean, this business has performed very well and is very satisfied with the overall momentum of the business. As for what you just asked about what we see in the market, we actually see that the core market is holding up very well. So let's call it aerospace and defense, industrial vision and medical, testing, simulation strong. We see weaknesses in communications. So that was the main driver of the weakness in the second half. As well as some inventory optimization factors, you might think, because our turnaround time has dropped in the past few months. I would say, think of it as a straight double-digit decline in the third quarter, and that's our current view. But overall, our business is very strong. So I think it's kind of an expected decline as we come out of the economic cycle. Q14: What is the follow-up guidance on inventory days levels? Q14: **As we add these product lines in the third and fourth quarters, you will see inventory decline first in the third and fourth quarters. Again, I think the number of days in stock is probably 110 to 120 days * *. The point is, if you look at many of our products, they are like advanced process technology, and the 5-nanometer manufacturing cycle is often very long. So **in the long run, you should expect our inventory to go from 100 days to 120 days instead of the traditional 80 or 75 days. This is too short for state-of-the-art process technology. **** Q15: Will supply be a limiting factor next year? **A15: I won't comment on specific units, but what I want to say is that we have been watching the supply chain of MI300 for some time. It's tight. There is no doubt that the industry is cash-strapped. However, we are committed to capacity throughout the supply chain. Cowos is one part, high-bandwidth memory is the other. Then there are the general capacity requirements. Our goal is to make it an important growth driver for AMD. I think it's a good market opportunity. We like the contact with our customers, and our responsibility is to provide supply for demand. That's what we 've been working. Q16: What might the timing look like with respect to the associated multi-billion dollar growth opportunity presented by the MI300? What are the design arrangements between different versions of the MI300? A16: I don't think we're ready to talk about the time of the revenue numbers. What we're saying is that we believe this is a multi-billion dollar opportunity. I think 2024 is a very important year for us. In the next few quarters, the promotion of MI300 among multiple customers is very important. I think I mentioned in the previous Q & A that the interests of customers are actually diverse, which is good. It includes the head hyperscalers you 'd expect, and these new types of AI-focused companies have been working closely with us, and some large enterprises are also considering ramping up their efforts. The performance we are seeing is strong. I think we did a lot of big language model work on the MI250. We have seen very good results. This includes both training and reasoning. I think the early results were strong in our study of MI300. For AI applications, we are now looking at the MI300X, so we call it the GPU-only version, which is the most common version of AI customer conventions. But in the MI300A, we actually bring the CPU and GPU closer together, and that's what we're interested in. So I think the key is that we have built a platform that allows people to choose what best suits the model and workload they want to implement. This is what we are working. Q17: How much telecom market share can Sienna help to gain? A17: We are very excited about Sienna. I think Sienna fits, as you said, this is a telecommunications market that we haven't looked at before. I think our interaction with telecoms providers is the desire for Sienna to be part of their portfolio. Sienna is also what we will use for other edge applications, or what we call low-end applications, which require the performance of Zen4, but may not be the heavy-duty platform we have on general and Bergamo. So we do think we're starting from a very low starting point. Therefore, in the next few years, we have an opportunity to gain market share, and we will focus on that. Q18: What is the impact of MI series growth on AMD's gross margin? A18: Both MI300A and MI300X will be part of the growth, especially in the fourth quarter and next year. For AI-specific applications, we prefer MI300X, just given where the software is written. As for your question about corporate-level gross margins, you would expect our AI business to increase corporate-level gross margins. Obviously, when you start climbing, you need to learn something. But **overall, we expect this to increase our company's gross margin. **** Q19: Most of the revenue comes from LCAP? How about working with other customers? **A19: If your question is, in addition to LCAP, do we have other customers dedicated to MI 300? The answer is yes. What I would say is that the difference in AI deployment is that I think customers are willing to go very fast. There is an eagerness and flexibility because we all want to accelerate the amount of AI computing. So the speed of customer engagement and the speed of customer decision-making is actually faster than they would in a normal environment. This is good. I think, as I 've said before, it helps us learn, improve the software, and be ready for major growth next year. Q20: With regard to server sharing, is there a theoretical limit to what AMD can get? Are there any large vendors in your server business that have a market share of significantly more than 30%, say 40% or even 50% ? A20: I think in the server business, the most important thing for customers is that we have a strong roadmap that they can trust. Over the past four or five years, we have been building this working model, roadmap and trust. Therefore, **I think there is no theoretical upper limit to AMD's share * *. I would say, if we look at it today, there are a lot of customers who have deployed more than 50 percent of our share in their data centers. From our point of view, we may be slightly underrepresented in the corporate space. It's just a matter of the breadth of enterprise customers and the breadth of enterprise software. So we believe we are leading now and we will be very focused on making sure we continue to lead in the market. With these, Microsoft has the opportunity to continue to gain share in the server market. Q21: What are the metrics for spending on generative AI and what is the outlook? A21: It's a bit like a crystal ball predicting what will happen in the next four or five years. There is no doubt that the demand for generative AI solutions is very high and requires a lot of computing power. **Our assessment of the market size is that it may grow at a positive or negative CAGR of 50% in the next three to four years. By 2027, our market size will reach $150 billion. **Now, these are data center accelerators. This includes GPUs, including other ASICs and accelerators. But I think we have an opportunity to address a large part of this market. So this is a very clear priority for us. This is our number one strategic priority, and we will continue to work closely with our customers to optimize between CPUs and GPUs. Risk Disclosure and Statement for this article: Dolphin Investment Research Disclaimer and General Disclosures
