美国超微公司 (AMD.US) 2026年第二季度业绩电话会
文章语言:
简
繁
EN
Share
Minutes
原文
会议摘要
AMD reported a 50% YoY revenue increase to $11.5B, driven by EPYC CPUs, Instinct accelerators, and Ryzen processors. Data center segment grew over 100% YoY, with expectations for continued growth in AI and server CPUs. Client segment grew, but gaming declined. Q3 guidance projects $13B revenue, with double-digit growth in data center and embedded segments, and improving margins.
会议速览
AMD's earnings call discusses Q2 financials, excluding 800M inventory charges, with forward-looking statements about future performance and upcoming investor events.
The dialogue highlights AMD's exceptional financial performance, with a 50% year-over-year revenue increase to $11.5 billion, primarily fueled by Epic Instinct, Ryzen, and embedded processors. The data center segment, now accounting for a significant portion of total revenue, experienced more than a doubling in sales, reflecting the expanding scale of server and data center AI businesses. The company anticipates sustained growth due to the multi-year AI adoption cycle, with increasing demand for compute resources across various markets and workloads. AMD's leadership in the x86 server market is attributed to the success of fifth-gen Epyc Turin and fourth-gen Epyc Genoa processors, gaining market share and expanding in cloud hyperscalers and enterprise sectors.
A gente AI introduces the Venice family, featuring the 6th Gen Epic Vents CPUs, designed for diverse workloads with Zen 6 core and 2nm technology. This new generation offers over twice the performance per watt compared to leading x86 CPUs and up to 3.3 times that of ARM-based CPUs. With more than 30 processors, Venice supports a wide range of memory and IO configurations, enhancing performance efficiency and TCO for cloud, enterprise, and HPC applications. Major OEMs and cloud providers are set to launch platforms and deployments, driven by strong customer demand, aiming to expand market share in the coming quarters.
AMD's Instinct accelerators, particularly the Mi 355x and Mi 450 series GPUs, have seen significant adoption in AI and cloud sectors. The launch of the Helios AI platform, featuring enhanced throughput and efficiency, is receiving strong customer interest. Strategic partnerships with Anthropic and Microsoft are expanding AMD's presence in next-generation AI infrastructure, with plans for annual releases of advanced rack-scale platforms, including the Mi 500 series GPUs, aiming for substantial performance and efficiency gains.
AMD highlights significant advancements in its AI software stack, particularly with the introduction of Racom Do AI, which enhances development and performance capabilities for AMD GPUs. The company forecasts robust growth in the data center AI accelerator market, expecting it to expand over 45% annually, reaching $1.4 trillion by 2030. Client and gaming segment revenues also saw notable increases, with client revenue up 23% year-over-year, driven by record mobile processor sales and market share gains.
The company reported record revenue and profitability, driven by strong commercial adoption, embedded segment growth, and strategic product launches. Despite softer PC market expectations, the client business is poised to outperform, with gaming and embedded segments leading the charge. AI is catalyzing demand for high-performance computing, positioning the company for significant market growth.
The company reports significant revenue growth, exceeding prior targets, with data center sales leading the charge. AI adoption is highlighted as a major driver, with AMD poised to capitalize on long-term growth opportunities, showcasing record sales and expanding margins.
The company reported a 6% YoY and 7% sequential increase in Client Gaming segment revenue, primarily fueled by record mobile revenue. Data Center and Embedded segments showed strong double-digit growth, while Gaming segment revenue declined 31% YoY but increased 8% sequentially. Guidance for Q3 2026 projects revenue of $13 billion, with double-digit growth in Data Center and Embedded segments offset by a decline in Gaming. Non-GAAP gross margin is expected to be around 56%, with operating expenses of $3.65 billion.
The dialogue discusses the growth trajectories of data center AI and server CPU businesses, highlighting their significant contributions to 2027 growth, with a focus on the timing of the crossover point and the relative strength of each segment.
The dialogue explores the company's gross margin strategy, emphasizing the impact of expanding survey and embedded businesses, alongside the ramp-up of data center AI, which offers substantial revenue opportunities despite slightly lower gross margins. The speaker expresses confidence in managing the business transition to sustain gross margin improvements, particularly highlighting the tailwind effects from pricing and mix shifts towards CPUs.
The dialogue discusses robust growth expectations for server CPUs and data center AI segments, with a focus on Q3 and Q4 projections. Strong demand across hyperscale and enterprise customers, coupled with increased supply, is anticipated to drive double-digit sequential growth in Q3, with even higher growth forecasted for Q4 as healios ramps up significantly.
The dialogue discusses the company's ambitious plan to surpass the 40% TAM growth in the compute market through 2030, driven by AI tailwinds across server CPUs, data center AI, embedded, and PC businesses, aiming for overall revenue growth above market expectations.
The dialogue discusses AMD's expectations for significant growth in data center AI, projecting over 100% growth for their segment in 2027, driven by strategic partnerships with leading companies. It highlights AMD's supply capacity to meet and exceed guidance, with a focus on continued collaboration with partners to satisfy high demand for compute. Revenue per gigawatt is anticipated to remain in the double-digit billions, emphasizing the importance of aligning with customer build-outs.
Discusses AMD's close collaboration with memory partners to manage high HPM benefits and memory cost inflation, ensuring profitability and allocation requirements are met, while adapting memory footprint based on workload needs.
AMD discusses robust server CPU demand, emphasizing growth in both units and ASPs. They highlight strong market demand and capacity expansion efforts, setting expectations for continued growth in 2027.
Discussion focused on Healios exceeding volume forecasts, with expectations of improving yields and margins. The server market was segmented into general purpose and AI front-end nodes, highlighting the potential growth of the sandbox AI market.
The dialogue discusses the potential for significant growth in the AI and server markets by 2030, with a focus on the company's strategy to lead with a family of CPUs optimized for various workloads. The new Venice CPUs are highlighted for their performance improvements, enabling the expansion of workloads and securing a strong market position, particularly among hyperscalers, driving confidence in substantial growth ahead of the market.
The dialogue discusses optimism regarding the client business growth despite PC market pressures, emphasizing the importance of AI PCs and a strong product portfolio. It also clarifies the acceleration in the data center business, indicating higher growth in the second half compared to the first.
Discussion focuses on data center growth, emphasizing server CPU targets exceeding 70%, with AI business forecasted to surpass 100%, driven by strategic customer strength and Helios ramp-up.
The company anticipates a diverse customer base for its Mi 400 series, including large frontier model companies and regular scale customers, with a focus on aligning with data center build-out schedules. This strategy ensures readiness and flexibility, especially during Q4 and Q1, to meet varying customer demands.
The dialogue highlights the company's strong confidence in meeting data center AI business targets, emphasizing the readiness of data centers, supply chain robustness, and collaborative efforts with operators to accelerate AI compute capacity.
The dialogue focuses on revenue expectations for the data center AI segment in 2027, with discussions on first-half versus second-half revenue cadence, and gross margin projections for the 455x in heels product starting in Q1 and exiting in Q4. It suggests that revenue growth should be considered well over 100%, with quarterly growth expected to continue.
The dialogue highlights the company's optimistic outlook on managing gross margin, driven by the ramp of the Mi 450 and growth in server CPU business. It emphasizes the importance of multiple levers, including embedded and client businesses, in improving margins, with further details to be provided for 2027.
The dialogue discusses the current tightness in the server CPU supply chain, attributed to unforecasted demand, with expectations of improvement in 2027 due to better demand forecasting. It highlights AMD's approach leveraging chiplet technology to efficiently ramp up new nodes, focusing on securing necessary supply for strong customer demand in servers and data center AI, while addressing overall supply chain aspects including backend and packaging capacity.
The partnership with Cerebrus, leveraging advanced AI and disaggregated computing, is expected to significantly grow sales in the fast inference market, with a customized solution anticipated to be available by Q4 and extending into 2027. This strategic move underscores the importance of optimizing technologies for various workloads, reinforcing the commitment to an open ecosystem.
The company plans to manage Opex increase at a rate lower than the expected 40% top line revenue growth, aiming to leverage operational efficiency for higher earnings per share, exceeding $20 as previously outlined. This strategy underscores the commitment to optimizing costs while capitalizing on large market opportunities.
要点回答
Q:What were the financial results for AMD's second quarter?
A:For the second quarter, AMD reported record revenue and profitability with a 50% year-over-year increase in revenue to $11.5 billion, driven by strong sales of EPYC, Radeon Instinct, and embedded processors. Data center revenue grew more than twofold, representing 58% of total revenue.
Q:What are the expected future prospects for AMD's AI business?
A:AMD anticipates continued strong demand for AI, driven by new growth vectors for server CPUs and new product families like EPYC, which is built for a range of AI workloads and offers significant performance gains. The company also expects to grow market share across cloud and enterprise sectors. The data center AI business is expected to be driven by strong demand for Instinct accelerators and strategic partnerships with leading companies.
Q:How much year-over-year growth did AMD's data center revenue experience?
A:AMD's data center revenue grew 107% year over year to a record $6.7 billion.
Q:What performance and efficiency improvements does the new 6 Gen EPYC processor family offer?
A:The new 6 Gen EPYC processor family, built on Zen 6 core and 2 nm technology, is purpose-built for a range of workloads including high-frequency AI, high-density compute, and general-purpose cloud and enterprise workloads. It offers one of the largest generational performance gains in EPYC history, with more than double the performance per watt of leading x86 CPUs and up to 3.3 times the performance per watt of leading ARM-based CPUs.
Q:What are the details of the new AI platform, Helios, and its strategic partnerships?
A:Helios is a new rackscale AI platform combining EPYC CPUs, Radeon Instinct GPUs, Pensando networking, and ROM software, tailored for a broad range of inferencing workloads. It is expected to deliver up to 15% more throughput and up to 30% more tokens per dollar than competitors at the same rack power. Strategic partnerships include a deal with Anthropic for 200 GW of Radeon Instinct GPUs and a multi-year joint engineering collaboration using HoloLens to optimize workloads. Microsoft will deploy Helios at scale on Azure for front-end model inferencing. Initial shipments of Helios are on track to begin later this quarter.
Q:What performance capabilities and developer experience advancements has the AI software stack achieved?
A:The AI software stack, specifically Radeon, has reached an important inflection point with enhanced performance capabilities and developer experience, including the ability to deploy AI in production at scale. The breadth of the ecosystem has expanded with more than 3 million models now running out of the box on AMD, leading open models with day 0 support for Radeon and significant open source contributions. Additionally, Radeon Do AI, a new AI-assisted development platform, allows developers to use modern coding agents to create and optimize code for Radeon GPUs, significantly speeding up model and workload deployment to AMD.
Q:What are the new forecasts for the data center AI accelerator market and the server CPU market?
A:The new forecast for the data center AI accelerator market is an annual growth of more than 45%, reaching approximately $1.4 trillion by 2030. The server CPU market is expected to grow more than Ed annually, reaching approximately $Ly billion by Ry for AMD. These markets are being driven by the increasing demand for AI across various applications and workloads.
Q:How is commercial adoption progressing and what new products have been launched?
A:Commercial adoption continued to expand with Ryanne pro sales growing more than Ed year over year. New wins were closed with major companies across healthcare, technology, automotive, and financial services sectors. Dell, HP, Lenovo, and Asus have launched a portfolio of new commercial PCs powered by the latest Ryzen AI Pro 400 series processors. Additionally, demand was strong for Ryzen AI halo developer systems, which feature the new Gorgon halo processor with an industry-leading 64 GB of unified memory.
Q:What is the expected performance of the client business and the gaming revenue in the second half of the year?
A:The client business is expected to perform better than the market in the second half of the year, with a focus on a softer PC market due to higher memory and component costs affecting demand. Despite a decline in gaming revenue of 31% year over year to $779 million, which is attributed to lower semi-custom sales and higher industry-wide component costs, the client business is anticipated to grow strongly driven by the Ryzen portfolio and commercial adoption.
Q:What growth has been seen in the embedded segment and what new products were introduced?
A:The embedded segment experienced year-over-year revenue growth of 19% to $977 million, marking the strongest growth in more than three years. Growth was broad-based across various industries, with a significant increase in the embedded x 86 business as hyperscalers and networking customers adopted AMD CPUs for data center functions. New products include the Ryzen AI embedded X 100 processors for edge AI workloads and the Creo AI robotics platform for physical AI applications.
Q:How does the company view the future growth potential in high-performance computing?
A:The company views a substantial growth potential in high-performance computing, with the market expected to grow approximately script annually over the next several years, approaching $2 trillion by 2030. As a result, the company expects to grow significantly above the market due to its focused investment, broad and strong product portfolio, and strategic relationships.
Q:What is the updated financial outlook and how is the company positioned for future growth?
A:The company is tracking ahead of its long-term financial model and now expects revenue growth to substantially exceed the prior target of greater than 35%, with the potential to significantly exceed the $20 annual EPS target within the strategic timeframe. The company is well positioned to capitalize on the ongoing multi-year AI adoption cycle, which is expected to provide significant growth opportunities in the coming years.
Q:What is the nature of growth seen in the company's business?
A:The nature of the company's business growth is unit and ASP (Average Sales Price) growth, with strong demand from the overall market resulting in over 70% growth in both cloud and enterprise segments, and even higher unit growth than the ASP growth.
Q:What actions has the company taken in response to the significant inflection point in server demand?
A:The company has been working diligently across its supply chain, including wafers, back-end capacity, and substrates, to raise overall server capacity in response to the significant inflection point in server demand.
Q:What is the expected impact of increased server capacity on future results?
A:The increased server capacity is expected to play out through the year, contributing to the company's ability to raise second-half guidance and supporting the growth previously discussed.
Q:How should one expect yields to improve as Helios ramps up?
A:It is expected that overall yields will improve as Helios ramps up over the next few quarters, starting from Q3. Initial yields are anticipated to continue improving, especially on a highly complex product like Helios.
Q:What is the projected size of the server market's general purpose and AI front-end node segments by 2030?
A:By 2030, the company projects the AI sandboxes to be the largest piece of the total addressable market (TAM), growing rapidly and representing the smallest piece today. The general purpose and AI front-end node segments are seen as significant contributors to the server market's growth.
Q:How does the company plan to grow in each segment of the server market?
A:The company believes it can grow in every segment of the server market by offering different optimization points for the types of CPUs needed for various workloads, indicating a diverse family of CPUs for each segment.
Q:What is the company's position in the server CPU market, particularly in terms of general purpose and AI applications?
A:The company holds a strong position in the server CPU market, with Venice being at the forefront in all categories including general purpose and AI applications. This is evidenced by customer traction and momentum, with workloads expanding on the next generation of products.
Q:How is the client business performing and what are the expectations for growth?
A:The client CPU business is seen as very important, with local AI becoming more prevalent and AI PCs gaining importance. Despite a market decline expected in the second half, the company's view is that there will be year-over-year growth in client business, and a strong product portfolio is expected to address both traditional and AI-centric PC experiences.
Q:What is the projected growth for the data center business in terms of year over year?
A:The data center business is expected to show a year over year growth with an acceleration in the second half versus the first half. Specifically, the server CPU is projected to grow more than 70% and the overall data center business is expected to be well over 100% due to strength in strategic customers and the ramp of Helios.
Q:How should the expectation of more than a 70% increase in server CPUs be interpreted?
A:The expectation of more than a 70% increase in server CPUs indicates a very strong trajectory for the data center business. This reflects the anticipated growth due to the strength of strategic customers, the expansion of the data center AI business which is expected to be well over 100%, and the contributions from strategic initiatives like Helios.
Q:What is the expected customer diversity for the data center business in the early stages of the Mi 400 series ramp?
A:The expected customer diversity for the data center business includes a mix of large model companies such as OpenAI, Anthropic, and Meta, who will consume resources through various Cloud Service Providers (CSPs), both hyperscalers and others. In addition, there are other customers who are interested in Helios and have a need for a more regular scale rather than a gigawatt scale. The diversity is expected to be good as the business ramps up through the next couple of quarters.
Q:Is the company facing any constraints in the data center supply chain that would hinder meeting their targets?
A:The company does not see any constraints in the data center supply chain that would hinder meeting their targets. They feel very good about the supply chain, including silicon, CPU, GPU, networking components, and Helios components. They are working closely with data center operators to ensure visibility and are optimistic about meeting the data center AI business targets. The company is also focused on accelerating the construction of data centers with ecosystem partners.
Q:How does the projected revenue growth for the data center AI business compare to the previous expectations?
A:The projected revenue growth for the data center AI business is significantly higher than initial expectations. The data center AI number is expected to be much more than just over 100%, with a progression over the next few quarters, growing each quarter as the business continues to ramp. The growth reflects an acceleration over time.
Q:What is the expected gross margin for the data center AI business in 2027?
A:The expected gross margin for the data center AI business in 2027 is optimistic, and it will be influenced by the pace of server CPU business ramp and data center AI ramp. The company is very positive about managing the ramp of the Mi 450 while also seeing significant improvement and growth in the server business in 2027. There are multiple levers available to improve gross margins, and the company will provide more guidance as they approach those milestones.
Q:Is the server CPU business currently supply constrained, and how is the company addressing the supply chain challenges?
A:The server CPU business is currently supply constrained due to unforecasted demand for the first half of the year. The company is feeling optimistic as demand is better forecasted for 2027, which is expected to improve the server supply situation compared to 2026. The company is confident in meeting the projected over 70% year-over-year growth and, depending on future events, there may be opportunities for higher growth. The company is actively managing the supply chain, including wafers, backend capacity, packaging capacity, and substrates, to satisfy both the strong ramp in servers and the data center AI business.
Q:What is the impact of 3nm node availability on the company's ability to ramp new capacity?
A:Ramping new capacity on the 3nm node is challenging, but the company's chiplet technology allows for a smaller number of wafers in the new node, which helps in securing the necessary supply to meet strong customer demand. The company is actively working on ensuring supply availability and feels optimistic about satisfying both the server and data center AI business demands.
Q:What are the expectations for sales growth in the fast inference market with the partnership with Cerebras?
A:The inference market, particularly fast inference, is experiencing substantial growth this year and into next year. The partnership with Cerebras is expected to provide a strong solution for customers with the integration of their technology with Helios and the wafer-scale engine. This solution is anticipated to become available in Cerebras' cloud offerings in Q4 and extend into 2027. The company views fast inference as an important part of the market and is focused on customizing and optimizing technologies for various workloads.
Q:How is the company planning to manage Opex growth in relation to the projected market growth?
A:The company expects Opex growth to continue at a faster pace than the top-line revenue growth. As such, they plan to manage Opex increases less than the revenue growth to drive operating leverage and deliver earnings per share. The business model is designed to achieve this, and the company is confident in their ability to manage expenses effectively.






