博通公司(AVGO.US)2025财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
Broadcom reported record revenues of $15 billion, with a 20% YoY increase, driven by robust AI semiconductor and VMware software sales. AI semiconductor revenue surged 30% YoY, with custom AI accelerators and networking growing significantly. The company announced the next-generation Tomahawk 6 switch and anticipates accelerated custom AI accelerator demand. Non-AI semiconductor revenue showed signs of recovery, and infrastructure software revenue rose 16% YoY. Guidance for Q3 projects $15.8 billion in revenue, with AI semiconductor revenue expected to jump 60% YoY. The transition of VMware customers to a subscription model is progressing, with the process expected to continue for another year to a year and a half.
会议速览
Broadcom Inc's Fiscal Year 2025 Second Quarter Financial Results and Third Quarter Guidance Conference CallBroadcom Inc held a conference call to discuss its financial performance for the second quarter of fiscal year 2025, providing guidance for the third quarter and commentary on the business environment. The call was led by the head of investor relations, with contributions from the president and CEO, the chief financial officer, and the president of the Semiconductor solutions group. Participants were directed to the company's website for the press release and financial tables, and were informed about accessing the webcast and replay. The discussion primarily referred to non-GAAP financial results, with a note on the availability of a GAAP to non-GAAP reconciliation in the press release.
Record Fiscal Qi Xi Performance Driven by AI Semiconductors and Infrastructure Software GrowthThe company reports a record revenue of $15 billion, up 20% year-on-year, driven by organic growth in AI semiconductors and strong performance in VMware. AI semiconductor revenue reached $4.4 billion, growing significantly year-on-year, with custom AI accelerators and AI networking showing robust growth. Non-AI semiconductor revenue is near its bottom, with some segments showing sequential growth. Infrastructure software revenue increased above expectations, reflecting successful conversion of enterprise customers to the Vcf software stack, with subscription customers turning to Vcf for modernized private cloud solutions. Q3 consolidated revenue is guided to be approximately $15.8 billion, up 21% year-on-year.
Review of QR Financial Performance: Record Revenue and Strong Semiconductor Segment Growth in Q2 2025Consolidated revenue reached a record script billion in the quarter, with gross margins at 79.4% and operating income up Ed from the previous year. Semiconductor solutions segment revenue grew to 8.4 billion, representing 56% of total revenue, driven by AI, with a gross margin of approximately 60%.
Comprehensive Financial Update and Q3 Guidance for a Global Tech Company - June 06, 2025The company reports a 12% year-on-year increase in operating expenses due to investments in AI semiconductors, with an improved semiconductor operating margin. Infrastructure software revenue grew significantly, representing a substantial portion of total revenue, and the operating margin improved due to disciplined integration of VMware. Free cash flow stood at $6.4 billion, impacted by interest expenses and taxes. Inventory increased in anticipation of future revenue growth. The company repurchased $6 billion in debt and paid $2.8 billion in cash dividends. Guidance for Q3 includes a 21% year-on-year revenue increase, with a focus on AI semiconductor and infrastructure software segments.
Increased Deployment of XPU and Networking for Enhanced AI Capabilities in the Upcoming Fiscal YearThere's a notable increase in the deployment of XPU expected for the next year, coupled with growing networking advancements. This visibility suggests a significant boost in AI capabilities, matching the growth rate of the current year.
AI Business Growth Trajectory and Fiscal Year Expectations as of June 06, 2025The dialogue discusses the AI business's positive growth inflection quarter-on-quarter and year-over-year, aligning with the scripted year-over-year growth rate. Visibility into hyperscale partners' deployments and improved updates provide a trajectory for sustained growth into the next fiscal year, consistent with the previously established growth CAGR.
AI Networking Strength and Tomahawk's Impact on Future AccelerationThe strength in AI networking is closely tied to the deployment of AI accelerator clusters, experiencing higher density in scale-up scenarios than initially anticipated. This trend has led to sustained performance in the AI networking sector, contrary to previous expectations of a drop. Looking ahead, the role of Tomahawk is anticipated to drive further acceleration in the coming year.
Strong Market Demand for Innovative Quest Therapy Per Second Tomahawk Switches as of June 06, 2025Despite only shipping basic proof of concept orders, there's a significant interest in the new therapy switches. The current date is June 06, 2025, highlighting the high demand for this technology.
Inquiry on Shipping Timeline for Ethernet Scale Up Network to Major Customers by June 06, 2025A representative from Jefferies asks about the visibility and timing for shipping an Ethernet scale-up network to primary customers, focusing on the scale-out opportunity.
Rapid Transition to Ethernet for Hyperscale Customers in Scaling Up Processes as of June 06, 2025The dialogue highlights a swift shift towards ethernet usage among a specific group of hyperscale customers in the context of scaling up operations.
AI Revenue Growth Trajectory and Visibility for Fiscal Year 2026The company discusses its AI revenue growth, aiming for a significant increase by the end of the fiscal year, with a focus on maintaining the current trajectory based on visibility into the upcoming quarters. They clarify that while they are not updating specific numbers at present, they are monitoring milestones with new customers to confidently incorporate their addressable opportunities into future revenue projections.
Exploring the Evolution and Benefits of Custom Accelerators in Optimizing Large Language ModelsThe discussion highlights the increasing preference for custom accelerators in optimizing large language models, emphasizing the value in integrating hardware and software for enhanced performance over time. It underscores the learning curve associated with developing and optimizing algorithms on custom silicon, leading to superior performance compared to using third-party merchant silicon. The journey towards mastering this integration is portrayed as a continuous process of improvement and optimization.
Transition from Copper to Optical Interconnects in Scaling Up Networking: An In-depth ExplorationThe discussion highlights the significance of transitioning from copper to optical interconnects as scale-up networking clusters grow larger, emphasizing the role of optical interconnects in addressing distance limitations. It explores the potential of co-packaged optics and traditional low-cost optics in facilitating this shift, anticipating a major transformation within a year or so.
Analysis of Gross Margins and Margin Profiles in Business Diversification as of June 06, 2025The discussion revolves around the implications of a significant incremental revenue increase and its effect on gross profit margins, highlighting the dilutive impact of certain business segments. Further, it explores the long-term margin profile expectations as the business scales and diversifies, noting that the simplistic analysis underestimates the complexity of margin dynamics.
Discussion on AI Networking Ecosystems and the Role of Ethernet in Scale Up vs Scale Out StrategiesThe dialogue explores the competitive landscape of AI networking ecosystems, highlighting the debate around scale up versus scale out strategies. It discusses the significance of Ethernet as an open standard for networking, emphasizing its prevalence and efficiency over proprietary protocols. The conversation also touches upon the impact of new ecosystem openings and the potential for growth in AI networking.
Acceleration in AI Demand: Monetizing Inference for Hyperscalers and ROI JustificationThe surge in AI demand is attributed to hyperscalers' need to justify high spending on training through monetizing inference, driving a shift towards xPU from GPU technology.
Forecasting AI Revenue Growth Amidst Hyperscale Customer Demands and Inference ExpansionThe discussion centers around the expected acceleration of AI revenue growth into fiscal 26, driven by increased demands from hyperscale customers and the addition of inference demands on top of training. Visibility suggests sustained growth from the twenties into 26 as the current best forecast.
Broadcom's Capital Allocation Strategy Amidst VMware Integration and Debt Reduction in 2025The discussion highlights Broadcom's approach to capital allocation, emphasizing the prioritization of debt reduction and dividend returns following the VMware integration. The company indicates a cautious stance on M&A, suggesting significant acquisitions would require additional debt, and outlines the use of free cash flow for debt reduction and opportunistic share buybacks.
Discussion on Networking's Role in AI Growth and Contribution Expectations from Current CustomersThe conversation focuses on clarifying expectations for contributions from current customers rather than prospects, and the anticipated balance between networking and XPU in the AI sector's growth.
Uncertainty Surrounding Export Controls on AI and Their Potential Impact on the Industry as of June 06, 2025The speaker acknowledges the dynamic environment of bilateral trade agreements and export controls in the AI industry, expressing uncertainty about future impacts due to rapidly changing rules.
Progress and Timeline for VMware's Subscription Model Conversion as of June 06, 2025The company is more than halfway through the process of converting VMware customers to a subscription model, with contracts typically lasting four years. Expectations are that the conversion will continue for at least another year to a year and a half.
Broadcom's Fiscal Year 2025 Third Quarter Earnings Call Announcement and Closing RemarksBroadcom announces plans to report its third-quarter earnings after market close on September 29, followed by a public webcast of the earnings conference call at 5 PM Pacific.
要点回答
Q：What was the main driver of revenue growth in the second quarter of fiscal year 2025?
A：The main driver of revenue growth in the second quarter of fiscal year 2025 was the continued strength in AI semiconductors and the momentum achieved in VMware, reflecting operating leverage.
Q：What is the outlook for non-AI semiconductors in the second quarter of fiscal year 2025?
A：The outlook for non-AI semiconductors in the second quarter of fiscal year 2025 is a revenue of $4 billion, which is down year on year. The segment is expected to remain around $4 billion, with enterprise networking and broadband growing sequentially, but server storage, wireless, and industrial expected to be largely flat.
Q：What recent progress has been made in the development of custom AI accelerators?
A：Recent progress in the development of custom AI accelerators includes the successful deployment of script customers and prospects, with plans for each to deploy custom AI accelerated clusters in 2027. The company anticipates an acceleration of demand for custom XPS (eXtensible Processing Elements) in the back half of the year to meet urgent demand for inference.
Q：What is the projected growth rate for AI semiconductor revenue in fiscal 2026?
A：The projected growth rate for AI semiconductor revenue in fiscal 2026 is expected to sustain the growth pattern established in fiscal 2025, indicating continued strong growth into the next fiscal year.
Q：What revenue growth is anticipated for the infrastructure software segment in the third quarter?
A：The anticipated revenue growth for the infrastructure software segment in the third quarter is $6.7 billion, which is up 16% year on year. This growth is attributed to the successful conversion of enterprise customers from perpetual vSphere to the full VCF software stack.
Q：How did gross margins perform in the semiconductor and infrastructure software segments?
A：Gross margins for the semiconductor solutions segment were approximately 60%, up 1 ly basis points year on year, driven by product mix. For the infrastructure software segment, gross margin was Ed in the quarter compared to Ed a year ago.
Q：What was the year-over-year change in operating expenses and operating margin for the semiconductor segment?
A：The year-over-year change in operating expenses for the semiconductor segment increased by 12% to $971 million, with a semiconductor operating margin of Ed, up Ed basis points year on year.
Q：What was the impact of capital expenditures and inventory levels on cash flow and balance sheet?
A：The company spent $144 million on capital expenditures. Ending inventory was up at $9.5 billion, anticipated to support future revenue growth. The company ended the second quarter with $69.4 billion in gross debt and $9.5 billion of cash.
Q：What were the details regarding the subsequent debt reduction and its impact on the weighted average interest rate?
A：Subsequent to the quarter, the company reduced $6 billion of debt, resulting in a weighted average rate of 3.8% for the remaining 69.8 billion in 6.5% debt, and a weighted interest rate of 5.3% for the remaining 8 billion in floating rate debt.
Q：What was the dividend payment in Q2, and how many shares were repurchased?
A：In Q2, the company paid stockholders $2.8 billion in cash dividends based on a quarterly common stock cash dividend of 15 cents per share. Approximately 25 million shares of common stock were repurchased in Q1, and the non-share to be 4.7 shares were mentioned, with the potential impact of any shares.
Q：What is the guidance for Q3 in terms of consolidated revenue and semiconductor revenue?
A：The guidance for Q3 is for consolidated revenue of $15.8 billion, a year-over-year increase of 21%. The forecast includes semiconductor revenue of approximately $15 billion, a year-over-year increase of 60%, and infrastructure software revenue of approximately $5 billion, a year-over-year increase.
Q：How is the Q3 adjusted EBITDA expected to be, and what is the forecasted non GAAP tax rate?
A：Q3 adjusted EBITDA is expected to be at least 66%, and the forecasted non GAAP tax rate for Q3 and the fiscal year is to remain at the current level.
Q：Can you provide more color on the inference commentary regarding growth rates for next fiscal year?
A：The company is seeing increased deployment of XPS next year, which, combined with more networking, is expected to contribute to growth. Confidence in the growth rate is based on the combination of these factors.
Q：What caused the strength in the networking AI segment and how much of it is attributed to the acceleration into next year?
A：The strength in the networking AI segment is closely tied to the deployment of AI accelerator clusters. The increase in scale up within data centers has led to higher consumption and density of switches, contributing to the surprise and continued robustness in this segment.
Q：When does the Tomahawk switch start contributing to the acceleration in networking?
A：There is strong interest in the new 6 terabit per second Tomahawk switches, with demand being extremely high. However, the company is not shipping big orders, except for basic proof of concepts, indicating that the acceleration will begin gradually.
Q：What is the visibility and timing regarding the shipping of a switch Ethernet scale-up network?
A：The visibility and timing regarding the shipping of a switch Ethernet scale-up network were not clearly communicated in the transcript. The speaker mentioned that the company is rapidly converting scale up to Ethernet for a specific group of hyperscale customers, but there was no mention of when a switch Ethernet scale-up network would be shipped to customers.
Q：What is the expected growth in AI revenues and is it related to an increase in scale?
A：The expected growth in AI revenues is tied to an increase in scale, with a projected growth in the Edwards script, which would put AI revenues at more than $1 billion for the year. The speaker mentioned that if the growth in the Edwards script continues, it aligns with the trajectory of ramping up AI revenue. They also indicated that based on current visibility, there is no reason to believe this growth trajectory will change.
Q：What factors should be considered when evaluating the trajectory of the XBU business and its transition from networking to AI?
A：When evaluating the trajectory of the XBU (Cross-Bar Unit) business, it's important to note that while networking is a challenging field, XBU is progressing along the expected trajectory without any signs of lumpy demand or softening. The business is expected to continue on this path into the next quarter and possibly beyond, providing clear visibility on the short-term trajectory. The company is not updating any numbers at this point, but future updates could be provided if clearer visibility becomes available, likely after 2026.
Q：How does the use of custom accelerators impact workload and the potential for inference versus training in the XBU business?
A：The use of custom accelerators impacts workload by allowing for a more optimized approach to developing algorithms, particularly for large language models. This leads to a learning curve on how to optimize algorithms and enables the combination of hardware and software for better performance. The ability to optimize software for custom hardware results in higher performance and is a significant value proposition for creating algorithms. The speaker indicates that the use of custom accelerators is not just about cost but also about performance and that the differentiation between training and inference is less relevant when using merchant accelerators compared to customer accelerators.
Q：What is the importance of demand for co-packaged optics in achieving the higher content opportunity for scale-up networks?
A：Demand for co-packaged optics is seen as important for achieving higher content opportunities in scale-up networks, especially as the size of scale-up clusters increases. While much of the scaling up currently uses copper interconnects for clusters that are not yet large enough to necessitate optical interconnects, as the clusters grow larger and require more connections, the demand for co-packaged optics and other advanced interconnect solutions will likely become more relevant. However, the extent to which co-packaged optics are needed for driving the scale-up opportunity was not definitively stated in the transcript.
Q：What are the reasons for using optical interconnects and when will this transition occur?
A：Optical interconnects will start replacing copper due to distance and the ability to deliver high-bandwidth connections. This transition is expected to happen within a year, with potential methods including package optics or lower-cost optical cables.
Q：What potential benefits will the switch to optical interconnects bring to the company?
A：The switch to optical interconnects will allow for a large increase in the number of connections, potentially reaching 512 connections per switch, which is significant for interconnecting GPUs and may be implemented within a year.
Q：What is the company's position on margins and how should investors view the margin profile of the custom business as it scales?
A：Historically, the company's margins for the XP business have been lower than the rest of the business. The quarter-over-quarter decline is mainly attributed to additional excus. The company acknowledges the complexity of the margin analysis and believes a simple analysis could be misleading.
Q：What is the company's stance on the emerging ecosystems such as HOP and Nvidia's NvLink?
A：The company believes in the simplicity of using existing protocols and standards like Ethernet for scaling up, rather than creating new systems. They contend that new protocols are mostly proprietary, while Ethernet, being open and standard, has historically prevailed in networking.
Q：What is driving the recent acceleration in AI demand and how might this affect the market share shift towards XPU from GPU?
A：The recent acceleration in AI demand is attributed to the need for hyperscalers to justify their spending on model training, which consumes a lot of compute power. Monetizing inference is now a priority, and availability of inference is driving the upside. This is expected to lead to a faster market share shift towards XPU from GPU.
Q：Is the company expecting a growth acceleration in AI revenue in fiscal 2026?
A：The company expects the growth rate in AI revenue to sustain into fiscal 2026 based on improved visibility and the continued demand for inference in addition to training, as cluster sizes increase.
Q：What are the expectations for the market shift from Nvlink to scale up in terms of Ethernet?
A：The company does not participate in Nvlink and therefore does not offer specific insights into the potential shift from Nvlink to scale up in terms of Ethernet. They suggest that the market may follow the trends in top-of-rack and scale up.
Q：What are the company's plans for M&A and debt levels?
A：The company intends to maintain its focus on reducing debt to a level that does not exceed two times the ratio of debt to EBITDA. This approach is also intended to preserve borrowing capacity for future M&A activities, which are expected to be significant and substantial.
Q：How does the company view its networking business in relation to other segments like AI?
A：The company expects the networking segment to be a smaller portion of the mix compared to other segments like AI, with the proportion of networking revenue to AI expected to be closer to 30% rather than 40%.
Q：Is the company affected by new export controls on AI technology?
A：The company cannot provide certainty regarding the impact of new export controls on AI technology due to the dynamic and rapidly changing nature of bilateral trade agreements and rules.
Q：What progress has been made in converting VMware customers to the subscription model?
A：The conversion of VMware customers to the subscription model is about two-thirds complete, with less than one-third of the renewals still to be renewed. Therefore, there is still a significant portion of the renewal process to be completed, suggesting the conversion is not yet close to completion.
Broadcom Inc.
Follow
