AMD (AMD.O) released its third-quarter financial report for 2024 on the morning of October 30, 2024, Eastern Time. The key points are as follows:
The following is a summary of the AMD 2024 Q3 earnings call, for a detailed interpretation of the financial report, please refer to " AMD: "Timid" guidance, does AI landing mean AI downhill? "
I. $AMD(AMD.US) Financial Highlights:
II. Detailed content of AMD financial report conference call
2.1. Key points from the executive team:
- Business Progress
Data Center Business
Revenue: Increased by 122% year-on-year, reaching $3.5 billion
AI Accelerator Applications: MI300X is widely adopted in AI inference workloads of major customers such as Microsoft and Oracle, providing leading performance and TCO advantages.
ROCm Ecosystem: Since the release of ROCm 6.2, MI300X inference performance has improved by 2.4 times, transaction performance has increased by 80%EPYC Processors: The fifth-generation EPYC processors have started development with major OEMs and ODMs globally, expanding into a wide range of business applications.
Key Customers and Expansion:
- Meta has deployed over 1.5 million EPYC CPUs in its global data centers
- Cloud service providers such as Microsoft and AWS have increased the number of EPYC instances
- Google and OCI plan to launch the fifth-generation EPYC instances in early 2025
- Cloudflare has chosen the Genoa-X processor for high-performance and energy-efficient next-generation servers.
- Adoption of the fourth-generation EPYC platform has significantly increased, with customers like Dell, HPE, and Lenovo expanding by 50%
- New customers from the aviation, finance, and automotive industries have been added, including Airbus, Daimler Trucks, FedEx, and Siemens, among others.
ZT Systems Acquisition: The acquisition of ZT Systems to enhance AI infrastructure capabilities is expected to be completed in the first half of 2025. GPU Product Development Plan: Introducing the MI325X GPU, with a 20% performance improvement over the H200, production is scheduled to start in Q1 2025; accelerating the next generation MI350 series, aiming for a generational leap in AI performance by the second half of 2025;Based on the CDNA Next architecture, the MI400 series may be released in 2026.
Future Outlook: Data center GPU revenue is expected to exceed $5 billion for the full year, surpassing the guidance provided by the company in July and at the beginning of the year; AMD expects the AI accelerator market to grow at an average annual rate of 60%, reaching $500 billion by 2028.
Client Business
Revenue: Increased by 29% year-on-year to $1.9 billion
Product Iteration: Launched the Ryzen AI 300 and Pro 300 series, incorporating enterprise-level AI capabilities for the first time, expected to cover over 100 commercial platforms by 2025.
Future Product Plan: Planning to launch Ryzen 9000 X3D in November.
Gaming Business
Revenue: Affected by inventory and product transition, gaming revenue decreased by 69% year-on-year to $462 million.
Future Outlook: Preparing for the transition to RDNA 4, planning to launch a new generation GPU in early 2025.
Embedded Business
Revenue: Decreased by 25% year-on-year to $927 million, mainly due to a slowdown in industrial market demand.
New Product Applications: Versal products are used by aerospace customers such as SpaceX for satellites; the second-generation Versal product Telluride, released last quarter, can provide 10 times or more computing power.
Future Outlook: The embedded business is expected to expand at an average annual growth rate of 20% over the next few years.
Financial Performance
Overall Financial Performance
Total revenue was $6.8 billion, an 18% year-on-year increase, and a 17% increase quarter-on-quarter.
Gross margin reached 54%, an increase of 250 basis points year-on-year.
Operating income was $1.7 billion, with an operating profit margin of 25%.
Diluted earnings per share were $0.92, a 31% year-on-year increase.
Data Center Business
Data center revenue was $3.5 billion, a 122% year-on-year increase, and a 25% increase quarter-on-quarter, benefiting from strong market demand.
Data center business revenue accounted for 52% of total revenue.
Operating income was $1 billion, accounting for 29% of revenue.
Client Business
Client business revenue was $1.9 billion, a 29% year-on-year increase, and a 26% increase quarter-on-quarter.
Client business operating income was $276 million, accounting for 15% of revenue.
Gaming Business
Gaming business revenue was $462 million, a 69% year-on-year decrease, and a 29% decrease quarter-on-quarter.
Gaming business operating income was $12 million, accounting for 2% of revenue.
Embedded Business
Embedded business revenue was $927 million, a 25% year-on-year decrease, and an 8% increase quarter-on-quarter.
Operating income was $372 million, accounting for 40% of revenue
Q4 Guidance
It is expected that Q4 revenue will be approximately $7.5 billion (±$300 million), with a year-on-year growth of 22% and a quarter-on-quarter growth of 10%, mainly driven by data center and client businesses to offset the decline in gaming and embedded businesses.
Non-GAAP gross margin is expected to be 54%.
Non-GAAP operating expenses are expected to be $2.05 billion.
Non-GAAP other net income is $17 million.
2.2 Q&A Analyst Questions
Q: In terms of the data center GPU business, the company has raised the outlook for 2024 by $500 million. What are the main driving factors? Additionally, looking ahead to 2025, how do you view the growth potential of Instinct products between large cloud customers and enterprise customers? Especially in terms of the possibility of acquiring new customers in the cloud business with Instinct products?
A: Firstly, the data center business performed strongly in Q3, especially the Instinct product portfolio. We achieved several important customer milestones that exceeded initial expectations, driving quarterly growth in data center GPUs. As a result, we have raised the full-year revenue guidance from over $4.5 billion to over $5 billion, expecting a strong growth trajectory by the end of the year.
Regarding the overall outlook for 2025, based on market feedback, we are confident in future developments. Customers have significant investment needs in infrastructure for AI workloads. Our product portfolio is also steadily strengthening, with the upcoming launch of MI325 later this year and the model MI350 next year.
At the customer level, our customer base is expanding widely, with large cloud customers expanding their workloads on the Instinct platform.
Q: The gross margin guidance for the fourth quarter is relatively stable. Can you provide detailed explanations on the influencing factors? Looking ahead to 2025, you mentioned the continued growth of the data center business. Will the embedded business gradually recover, and will enterprise demand for server CPUs and other businesses accelerate growth from a low base? What potential resistance factors should we also pay attention to?
A: We are very satisfied with the gross margin performance in Q3, reaching 53.6%, and the guidance for Q4 gross margin is around 54%. Looking back at 2024, the improvement in gross margin was mainly driven by the business portfolio, especially the data center business. The data center business, as our core growth driver, accounts for over 50% of total revenue, playing a key role in improving the gross margin.
Looking ahead to 2025, we will not provide specific guidance, but among the main factors that will continue to drive the gross margin, the data center business remains the largest growth driver, including CPU and GPU products. Additionally, the expansion of enterprise-side business will also be a tailwind for the gross margin.
Furthermore, the gradual recovery of the embedded business will help improve the gross margin. At the same time, the client business is also expanding, but currently more focused on the consumer sector, with a gross margin lower than the average level of enterprise business, which will be a resistance factor that we need to address
Finally, our team has performed well in improving operational efficiency. As the company scales up, we will further benefit from economies of scale, continuously improving gross margins through enhanced operational efficiency.
Q: How would you describe the development of the data center supply chain after revenue exceeds $5 billion? Considering the production and delivery cycles of MI300 and MI325 GPUs, what measures have been taken on the supply side to ensure supply in 2025?
A: We are very pleased with the improvements in the supply chain over the past few quarters. Despite the ongoing tight supply environment, we have successfully ensured the entire supply chain's capacity, which is one of the reasons for the increase in third-quarter Instinct business revenue, reflecting the combined effect of customer demand and supply chain improvements.
Looking ahead to 2025, we expect the supply environment to remain tight, but we are well prepared for significant growth next year and are confident in the overall supply chain capabilities.
Q: Regarding the sequential revenue guidance of approximately $680 million growth, could you explain how much of it is being driven by the client business and how much is from the data center business?
A: In the sequential growth guidance, the data center business is the largest contributor, currently accounting for over 50% of our business, with further growth expected in Q4.
The client business is also performing well, with successful releases of desktop Zen 5 and laptop AIPC products this year, expecting growth in this area. Growth in other sectors such as gaming and embedded systems is relatively moderate.
Q: How is the performance of the embedded business specifically in the fourth quarter guidance? And what are the prospects for the embedded business in 2025?
A: In the embedded business, there was an improvement in Q3, and a slight growth is expected in Q4.
Performance varies across sub-sectors, with testing and simulation business performing well, and we are accelerating the deployment of the new Versal platform. The aerospace and defense sectors also remain relatively strong. However, the recovery in the communication sector is limited, and the industrial sector's performance is slightly weak.
Overall, a slight growth is expected in Q4, and there may be some growth in 2025, but the performance will vary across different sub-sectors.
Q: Has the increased demand for Instinct products affected the demand for EPYC, causing "crowding out" effects? Looking ahead to 2025, how do you view the demand dynamics between CPUs and GPUs from a customer spending perspective?
In the data center CPU business, we are very satisfied with recent progress. The market environment has significantly improved over the past few quarters, with large cloud customers increasing data center capacity and updating equipment. Additionally, some enterprises have begun modernization activities.
In such a market environment, our product portfolio has performed well. In the third quarter, there was strong demand for the Zen 4 series Genoa and Bergamo, with Zen 5 Turin gradually increasing. Furthermore, Zen 3 and Milan series continue to receive good market demand in terms of price performance ratio.
Overall, our product portfolio is very strong from top to bottom, and the server market also shows some growth momentum, providing support for the expansion of AI opportunities.
Q: Looking ahead to next year, AMD seems poised to further expand its share in the server market. Some opinions believe that GPUs are eating into the server market share. What specific trends have you observed in the Turin series? Especially in large-scale data centers, is this upgrade favored for its ability to integrate more AI devices? What do you think are the main driving factors for this upgrade?
A: Looking ahead to 2025, the Turin series has been well optimized for a wide range of server workloads and traditional CPU workloads, meeting the needs for scalable deployment and expansion, especially excelling in AI content. This also positions it as an important component in GPU configurations, overall providing us with a favorable growth catalyst.
Currently, we have made significant progress in workload expansion, deepening cooperation with major cloud customers, especially crucial in advancing in the enterprise sector. Enterprise sales cycles are longer, and we are ensuring that CIOs are familiar with our product portfolio through numerous POC projects. With the Zen 4 and Zen 5 series, we have a wide product range that can meet the needs of most traditional CPU workloads.
Q: Regarding the PC market, how do you observe the market in the fourth quarter? Are there any risks of seasonal declines beyond seasonal factors?
A: In the PC market, our business leans more towards consumer demand, so the second half of the year is usually stronger than the first half. This year, we launched the Zen 5, Ryzen 9000 series, and the new generation AIPC Ryzen AI 300 product portfolio, leading to a particularly strong performance in the second half. It is expected that there will be some seasonal fluctuations in the first half of next year, which is normal for the industry.
Our current PC product portfolio is the strongest in our history, covering desktop and laptop areas. Looking ahead to 2025, the overall market outlook is optimistic, with expected mid-single-digit growth. Additionally, the launch of AI PCs and the end of support for Windows 10 will also act as catalysts for market growth.
Overall, we are positive about the PC market in 2025, but expect seasonal adjustments in the first half of the year.
Q: In terms of data center GPUs, can you explain how much room for expansion there is among existing customers? Additionally, how do you view the catalytic effects of the MI325X based on the CDNA3 architecture and the new architecture MI355 in driving internal and external existing customer revenue and workload expansion?
A: In the data center GPU business, we have made significant progress. At the beginning of the year, we estimated revenue of around $2 billion by 2024 when we just launched the MI300, but now it has been raised to over $5 billion.
Looking ahead to 2025, we have strong confidence in the market and product portfolio. The continued capital expenditure demand in the market and the successful promotion of MI300 have greatly benefited us. Especially on the software side, we have significantly improved customer support, out-of-the-box performance, and open-source ecosystem, ensuring that customers can achieve scalable operations under good cost-effectiveness conditionsIn summary, we have multiple growth opportunities among existing customers, especially in terms of load expansion. Collaborations with clients such as Meta and Microsoft have further expanded our applications in large language models and training workloads. Additionally, we will continue to work closely with customers to optimize the compatibility of their software with our hardware.
Q: Is the actual revenue from data center GPUs in September around $1.5 billion? If so, will the revenue in December be close to $2 billion?
A: Our performance in the data center GPU business has exceeded initial expectations, with revenue in September actually surpassing $1.5 billion. We are now seeing the scale of the GPU business gradually approaching that of the CPU business.
Q: In the market environment with these large clients, how do you view the revenue growth trend for next year? Can it continue to grow on the basis of the current revenue in the fourth quarter? Because some clients may expect support for rack-level deployments. Are you concerned that there may be a brief pause before reaching rack-level deployments by 2026?
A: We are currently working closely with customers to meet their diverse needs in data center construction, including support for various configurations such as air cooling, liquid cooling environments, and rack-level deployments. Regarding growth opportunities in 2025, we remain optimistic but there may be fluctuations. Overall, these are acquisition projects for large clients, making it difficult to accurately predict the quarterly distribution of growth. However, we are satisfied with the overall development trajectory for 2025.
Q: Can you provide a detailed breakdown of the $500 million AI revenue in terms of training and inference? What are the future expectations?
A: Regarding the $500 million AI revenue, it currently mainly comes from inference applications as the MI300 product is highly optimized for inference tasks in terms of memory capacity and bandwidth. However, we have made progress in the training field and expect training demand to continue to grow over the next few quarters. Looking ahead a year from now, we anticipate a more balanced distribution of revenue between training and inference applications.
Q: Regarding the view that MI product technical specifications lag behind industry leaders by about a year, how does AMD plan to narrow this gap? When competitors launch new generations of products (such as Blackwell Ultra or Rubin), can AMD make breakthroughs in market share?
A: AMD does not agree with the view that MI products lag behind industry leaders. While the MI300 initially lagged behind the H100, this gap has been narrowed through an accelerated product roadmap. The MI325 will compete with the H200, and the MI350 series will also be well-positioned against Blackwell. Additionally, there are limitations in market demand for new generation products, as it takes time from initial samples to mass production. AMD has advantages in data center product transformation and strong infrastructure compatibility. Furthermore, there are broad opportunities for AI workloads, AMD's roadmap is continuously strengthening, and customer feedback is very positive.
Q: Regarding the growth in the PC market, how is the current channel situation? Considering the lackluster performance in shipment volume in the third quarter, is the growth mainly driven by an increase in average selling price (ASP) or by sales volume growth?Looking ahead to 2025, do you think this year's ASP advantage can continue?
A: Our client business has a high market share in the desktop field and performed strongly in the third quarter, reflected in revenue. In terms of laptops, it is more focused on the consumer market, usually performing better in the second half of the year. We have good growth momentum in the AI PC field, especially the activation rate of the latest Ryzen AI 300 processor is good. Although AI PC is still in the early stages, we see positive signs of growth.
Looking ahead, we expect shipments and average selling prices (ASP) to fluctuate based on the market share between consumers and enterprises, desktops and laptops. We believe that the client business is a potential growth point for us in the consumer and enterprise markets, with the market penetration rate in this area still not high enough, hence there is more market space in the future.
Q: Will the recovery trend of EPYC in multiple industries continue in this quarter, and can we expect quarter-on-quarter growth? Additionally, how do you view the enterprise and cloud demand trends in the Chinese market?
A: In the third quarter, the enterprise market showed positive growth momentum, with customers' acceptance of AMD's cloud instances increasing and a clear migration trend towards AMD platforms and new cloud instances. In the fourth quarter, we expect the server business in the enterprise and cloud domains to continue to achieve quarter-on-quarter growth.
Looking ahead, the enterprise business has great potential, and AMD still has room for improvement in this market. Our product portfolio has been strengthened, including the Zen 4 and upcoming Zen 5 platforms, which will all be growth opportunities in 2025. Our service representation in the Chinese market is insufficient, but it is also another opportunity for us to expand market share.
Q: The scale of the GPU business in the third quarter is close to 1.7 billion USD? If the full-year target is 5 billion USD, then in order to achieve growth in the fourth quarter, the full-year revenue needs to reach 52 to 53 billion USD?
A: In our data center business, apart from CPU and GPU, we also have other revenues such as FPGA. The GPU business revenue in the third quarter exceeded 15 billion USD. Although we have not provided precise guidance, we expect full-year revenue to exceed 50 billion USD.
Q: Are there any seasonal factors to consider in the data center GPU business in the first quarter of 2025? For example, the expected mass production of the MI300 series, will this bring seasonal fluctuations?
A: The data center GPU business does not have seasonal factors, but rather depends mainly on the specific needs of certain customers. The deployment needs of these large customers will bring business fluctuations. For example, the performance in the third quarter exceeded expectations mainly due to additional customer demand, so we may see this kind of volatility. As we enter 2025, we need to observe further developments in the business
Q: By 2025, major competitors may gain a market share of $50 billion to $60 billion in the AI and GPU markets, while the company expects revenue of only $5 billion to $10 billion. What do you think is the main obstacle to becoming a major player? Is it because the company started late, or are there other hardware incompatibilities or architectural differences?
A: In the evolution of our EPYC product line from Zen 1 to Zen 4, we have consistently performed well. Although we are in the same GPU product category as our competitors, there are indeed some differences in software environments, which require customers to gradually become familiar with and build trust. The current pace of product promotion is the fastest ever, and next year we will continue to expand our customer base and application scenarios, especially with the upcoming MI400 series. Overall, customers are very open to AMD and give us fair opportunities for cooperation.
Q: Can you provide new guidance or frameworks to help us understand how the gross margins of MI300 and MI325 can gradually reach the level of the company's overall business gross margin?
A: We are very satisfied with the overall revenue growth of our data center GPU business, and the team is committed to improving gross margins. Currently, the gross margin is below the company's average level, but in the future, our focus will be on meeting customer needs, providing total cost advantages, and significantly increasing market share and revenue.
With revenue growth, we expect to further improve gross margins. In the long term, the gross margin of the data center business is expected to exceed the company's average level. Over the past year, our data center segment revenue has doubled, and operating income has tripled.
Therefore, our goal is to gradually increase gross margins through achieving long-term growth and expanding market influence.
Q: Regarding the contribution of the MI300 platform in the fourth quarter, considering the release of the MI325 series, should we expect significant contributions from the new product in the fourth quarter, or will the main impact be seen in the first quarter due to the year-end release?
A: For the product mix in the fourth quarter, most of it will still be composed of the MI300 series. The MI325 will start mass production at the end of this quarter, with the main growth expected in the first quarter of next year.
Q: Could you introduce the development of Xilinx's end markets and embedded business? You previously mentioned the situation in the communication and some industrial markets, could you further elaborate on the specific performance from 24Q4 to 25Q1?
A: In terms of the embedded business, looking ahead to 2025, we observe trends similar to those in the third and fourth quarters. Some markets are expected to gradually recover, with strong performance currently in the testing and simulation fields, especially in the aerospace and defense industries. There are also signs of recovery in the automotive sector. For the communication and industrial markets, we are still waiting for more visibility, and expect a clearer picture by the end of this year.
Risk disclosure and statement of this article: Dolphin Research Disclaimer and General Disclosure
