博通公司 (AVGO.US) 2025财年第四季度业绩电话会
文章语言：
简
繁
EN
Share
Minutes
原文
会议摘要
Broadcom Inc. achieved record revenues, with a 65% year-over-year increase in AI semiconductor sales to $20 billion, driving total semiconductor revenue to $31 billion. Infrastructure software revenue also grew by 19%. The company anticipates continued AI revenue acceleration, stable non-AI semiconductor revenue, and low double-digit growth in infrastructure software. Broadcom expects fiscal Q1 revenue of $18.1 billion, with adjusted EBITDA at 67% of revenue, and has increased its dividend and share repurchase program. The company is focusing on innovation, supply chain resilience, and strategic partnerships, including a significant agreement with OpenAI, while managing challenges in non-AI semiconductor revenue segments.
会议速览
Broadcom's Q4 & FY2025 Results Highlight AI Revenue Surge and Strong Infrastructure Software GrowthBroadcom reported record revenues and EBITDA in Q4 and FY2025, driven by a 65% year-over-year increase in AI semiconductor revenue to $20 billion, and robust growth in infrastructure software, particularly VMware Cloud Foundation. The company secured significant orders, including a $11 billion AI accelerator deal, and expects AI revenue to double in Q1 of FY2026, projecting consolidated revenue of approximately $19 billion with adjusted EBITDA of around 45% of revenue.
Record Revenue and Robust Growth in Semiconductor Solutions Segment HighlightedConsolidated revenue reached a record 18 million, up 28% YoY, with semiconductor solutions segment leading growth at 61% of total revenue, driven by AI advancements. Operating income and adjusted EBITDA surpassed guidance, reflecting strong financial performance despite margin challenges from product mix.
Strong Financial Performance with Increased Margins and Cash Flow in Q4The dialogue highlights significant financial achievements, including a rise in operating margins, robust cash flow, and efficient inventory management, reflecting successful integration efforts and disciplined operational strategies.
Record Financial Performance with Increased Dividends and Share RepurchasesThe company achieved record revenue of 63.9 billion, with organic growth at 24% year on year. Semiconductor and infrastructure software revenues grew significantly. Adjusted EBITDA was 43 billion, representing 67% of revenue, with free cash flow increasing. The company returned substantial cash to shareholders through dividends and share repurchases, announcing an increase in quarterly dividends and expanding the share repurchase program.
Q1 Financial Guidance Highlights: Revenue Growth, Gross Margin Shifts, and Tax Rate IncreaseThe company provides Q1 financial guidance, forecasting a 1 billion LY revenue increase, a 100% YoY AI semiconductor revenue growth to 8.2 billion, and a 2% YoY rise in infrastructure software revenue. It anticipates a 100 basis point sequential drop in Q1 gross margins due to a higher AI revenue mix. Adjusted EBITDA is projected to be approximately LY, with a non-GAAP tax rate increase to 16.5% for Q1 and FY2026, influenced by global minimum tax and geographic income mix shifts.
AI Data Center Investment and Customer Tooling TrendsDiscussion on $73 billion backlog for AI data center components over 18 months, highlighting strong booking growth, particularly for Ta6 switches. Debate on customer preference for self-tooling versus reliance on service providers, with emphasis on multi-year trends and custom AI accelerators.
Custom vs Merchant Hardware in AI AccelerationDiscusses the advantages of custom-designed hardware over general-purpose GPUs, questioning the feasibility of customer-driven ASIC development due to silicon evolution. Highlights TPU's market expansion, differentiating its transactional shift from GPU to its strategic long-term investment for custom AI accelerators.
AI Backlog Growth, Supply Chain Security, and Advanced Packaging InitiativesDiscussed the anticipated increase in AI backlog, emphasizing the need for robust 3nm and 2nm wafer supply and addressing advanced packaging challenges with a new Singapore facility to enhance supply chain security and meet growing demand.
System Sales Approach for Hyperscaler AI SystemsThe dialogue discusses the transition to selling AI systems as complete solutions, rather than individual components, to hyperscalers. It highlights the integration of key components into a certified system, similar to selling a fully functional chip, and the strategic decision to market the fourth customer's deal as a comprehensive system sale.
AI Revenue Growth Impact on Margins and Operating LeverageDiscussion focuses on how increasing AI revenue, including system sales, will affect gross and operating margins. While gross margins may decrease due to higher component costs, the company anticipates significant operating leverage, maintaining high growth in operating margin dollars. Guidance on exact margin impacts will be provided closer to the end of the year.
AI Revenue Growth Projections and Backlog DynamicsDiscusses the acceleration of AI revenue growth, clarifies guidance for Q1, and addresses the dynamic nature of the backlog, emphasizing continued growth without specific full-year guidance.
OpenAI Contract Details and Growth ExpectationsDiscussion on OpenAI contract specifics, multi-year commitment, and customer growth, emphasizing no immediate contributions expected in 2026, with projections extending through 2029.
Exploring Custom Silicon and Customer Accelerators for Diverse WorkloadsDiscussion highlights the strategic approach to custom silicon design, emphasizing the creation of customer-specific accelerators for varied workloads. This includes tailoring chips for different aspects of training and inference, showcasing the versatility and innovation in hardware development to meet specific customer needs.
AI Growth and Silicon Photonics: Diverse Customer Expansion and Future Interconnect TechnologiesDiscusses robust AI revenue growth, emphasizing diverse customer contributions and strong demand for high-bandwidth switches. Addresses silicon photonics' potential as a future interconnect solution, highlighting current development and readiness for deployment in scalable systems.
Addressing Supply Chain Bottlenecks in AI Data CentersA discussion on enhancing supply chain resiliency and visibility, focusing on key materials and components for AI data centers. The conversation highlights the company's proactive approach in identifying and mitigating bottlenecks, particularly in leading-edge technologies like DSPs and lasers, to ensure smooth operations and meet growing demand.
OpenAI's Impact on Semiconductor Revenue and Future ProspectsDiscussion revolves around OpenAI's influence on non-AI semiconductor revenue, noting stability with a sharp recovery yet to be seen. Broadband shows promising recovery, while AI seems to divert enterprise spending. An agreement with OpenAI for developing Ed gigawatts over 2027-2029 is highlighted, distinct from the xPU program, indicating a committed phase in their collaboration.
Future of Rack Revenue and Demand for Compute CapacityThe dialogue explores the future of rack revenue, considering potential shifts in business models and the percentage of backlog for full systems. It emphasizes the demand for compute capacity beyond the next few months as a key factor influencing revenue trends.
Broadcom's Upcoming Presentations and Earnings Report AnnouncementsThe call concludes with announcements about Broadcom's upcoming presentation at the Street Research Virtual AI Big Ideas conference and its planned earnings report for the first quarter of fiscal year 2026, scheduled for release after market close. A public webcast of the earnings call will follow, marking the end of today's session.
要点回答
Q：What was the consolidated revenue growth for Broadcom in fiscal year 2025 and what were the main drivers?
A：In fiscal year 2025, Broadcom's consolidated revenue grew 20% year over year to $16 billion, driven by AI semiconductors and VMware.
Q：How is the demand for AI semiconductors and what impact are they having on revenue?
A：AI semiconductors have seen strong demand with robust revenue growth, with the AI semiconductor revenue growing 65% year over year to $20 billion, and the robust growth driving the semiconductor revenue for the entire company to a record $11 billion for the year.
Q：What were the revenue growth and achievement highlights in the fourth quarter for Broadcom?
A：In the fourth quarter, Broadcom's total revenue was a record $18 billion, up 20% year on year. Q4 consolidated adjusted EBITDA was a record $12.2 billion, up 34% year on year. Semiconductor revenue grew by more than double year over year, and AI revenue more than doubled, reaching $20 billion and driving semiconductor revenue to $11 billion for the quarter.
Q：What is the impact of strong AI demand on Broadcom's different business segments?
A：The strong AI demand has led to a robust increase in revenue for the semiconductor segment and a doubling of AI revenue in the quarter. It has also driven a significant increase in demand for AI networking, with a strong backlog of orders for AI switches exceeding $10 billion and a total order hand exceeding $73 billion.
Q：What is the forecasted revenue and delivery timeline for the AI backlog mentioned?
A：The forecasted AI backlog of $73 billion is expected to be delivered over the next several months, and in Q1 fiscal year 2026, AI revenue is expected to double year on year to $12 billion.
Q：What are the expected revenue changes and market conditions for non-AI semiconductors and infrastructure software in the upcoming fiscal year?
A：Non-AI semiconductor revenue is forecasted to be approximately $4.1 billion, up year on year due to favorable wireless seasonality and solid recovery in broadband, while wireless was flat. The infrastructure software revenue is expected to grow by a low double-digit percentage with renewals being seasonal in Q1.
Q：How is the financial performance and guidance for Q1 and the fiscal year structured for Broadcom?
A：For Q1 and fiscal year 2026, Broadcom expects AI revenue to continue accelerating, non-AI semiconductor revenue to be stable, and infrastructure software revenue to grow at a low double-digit percentage, largely driven by VMware. They anticipate a consolidated revenue of approximately $17.5 billion up year on year, with adjusted EBITDA to be approximately 70% of revenue for Q1. The forecast includes solid gross margins and sequential improvement in operating income and adjusted EBITDA.
Q：What were the key figures for free cash flow and days sales outstanding (DSO)?
A：Free cash flow in the quarter was $6 billion and represented 20% of revenue. DSO in the fourth quarter was 29 days, down from 34 days in the prior year's fourth quarter.
Q：How much cash was held at the end of the fourth quarter?
A：The company held $16.2 billion of cash at the end of the fourth quarter, up $5.5 billion from the prior year.
Q：What is the expected non-GAAP diluted share count for Q1?
A：The expected non-GAAP diluted share count for Q1 is approximately 4.97 billion shares, excluding the potential impact of any share repurchases.
Q：What were the financial performance figures for fiscal year 2025?
A：Fiscal year 2025 saw record revenue of $63.9 billion with organic growth accelerating to 24% year over year. Semiconductor revenue was $29 billion, up 20% year over year, and infrastructure software revenue was $20 billion, up 20% year over year.
Q：What was the dividend increase and share repurchase program expansion?
A：The quarterly common stock cash dividend was increased to $1.30 per share, an increase of 5% from the prior quarter. The Board also approved an expansion of the share repurchase program by an additional $6 billion, with a total of $13 billion remaining through the end of calendar year.
Q：What is the guidance for Q1 revenue and the forecast for AI and infrastructure software revenue?
A：The guidance for Q1 is for consolidated revenue of $74 billion, up 21% year on year. The forecast includes semiconductor revenue of approximately $77 billion, up 11% year on year, with AI semiconductor revenue of 8.2 billion, up approximately 100% year on year. Infrastructure software revenue is expected to be approximately $77 billion, up 2% year on year.
Q：What is the expected adjusted EBITDA and non-GAAP tax rate for Q1 and fiscal year 2026?
A：The expected Q1 adjusted EBITDA is approximately $10 billion. The expected non-GAAP tax rate for Q1 and fiscal year 2026 is to increase from 14% to approximately 16.5 percent due to the impact of the global minimum tax and a shift in geographic mix of income compared to fiscal year 2025.
Q：What is the difference between using GPUs and TPUs in AI applications?
A：Moving from GPUs to TPUs is considered a transactional move, while developing a custom AI accelerator is a long-term strategic move that involves a multiyear investment and is not typically made in a short-term or transactional context.
Q：How does the company interpret its AI backlog of $73 billion over the next 6/4 quarters?
A：The $73 billion is not just a snapshot of current orders but includes potential future orders given the lead times. The top-level number for AI shipments is expected to increase as time progresses and more orders are placed within the next 6/4 quarters.
Q：Does the company have sufficient 3 nm and 2 nm wafer supply commitments to meet the demand in the order book?
A：The company is focusing on advanced packaging with its Singapore facility and believes it has enough demand to source from in terms of costs and supply chain security. They are in the process of building up a substantial facility for advanced packaging in Singapore. However, the challenge lies in securing supply chain for silicon and packaging, which is ongoing as they keep progressing and increasing their backlog.
Q：What is the focus of the Singapore facility in terms of advanced packaging?
A：The Singapore facility is focused on partially sourcing advanced packaging for multi-chip solutions required in creating customer accelerators, as well as building up capacity in script nanometers for TSMC.
Q：How does the new $10 million deal with a fifth customer involve networking, and will it be a replication of Google's setup?
A：The new deal is for a system sale, which includes multiple components beyond the AI accelerators used by hyperscalers. The company believes it makes sense to sell the system with key components included, and it plans to be fully responsible for the entire system. This approach is similar to selling a chip that is certified and finalized for the wholesale selling process. However, details on whether it will replicate Google's networking setup or include the company's own networking elements were not explicitly stated in the transcript.
Q：What should investors expect in terms of gross margin and operating margin as AI revenue and system sales ramp up?
A：As AI revenue and system sales increase, investors should expect a dilution in gross margins. However, the company expects the rate of growth in AI revenue to generate operating leverage, which means that despite a decrease in gross margins, operating margins will still show high-level growth. The company anticipates that the operating leverage will benefit the operating margin, keeping it at a high level even as gross margins start to deteriorate. Additionally, when more systems are shipped in the second half of the year, there will be a lower gross margin on those systems, but overall, operating margin dollars are expected to grow.
Q：What is the projected revenue from XPS over the next few months and what is the significance of silicon photonics?
A：The projected revenue from XPS over the next few months is potentially around 20 billion, which includes all other revenue types. Silicon photonics is seen as a means to create more efficient and lower power interconnects, with potential to be the only viable option in the future. It is currently developed for 400 Gb, 800 Gb, and now 1.6 terabit width to grid silicon photonics switches and interconnects.
Q：What challenges are faced with scaling up using copper and what technology is being developed to potentially replace it?
A：The challenges with scaling up using copper include the difficulty to achieve the necessary scale while maintaining efficiency. The company has technology developed in silicon photonics, which is being advanced from initial uses in 400 Gb to now 1.6 terabit widths, as an alternative to copper for high-scale data centers.
Q：What visibility does the company have over its supply chain and how does it address potential bottlenecks?
A：The company has visibility across its supply chain and is able to manage potential bottlenecks by being part of the system racks for AI data centers. They have visibility into potential bottlenecks because they are part of the solution. The company feels confident about managing any bottlenecks that arise through the end of the year.
Q：What is the nature of the agreement with OpenAI, and how does it relate to non-AI semiconductor revenue?
A：The agreement with OpenAI is not a general binding agreement but is more like the ones with Nvidia and AMD. OpenAI is significantly impacting spending in the enterprise and hyperscaler sectors. Regarding non-AI semiconductors, there is a recovery in broadband, stability in other segments, and no sharp sustainable recovery yet. There is no inventory overhang in non-AI semiconductors, and growth in this business is not expected to recover quickly, except for broadband.
Q：What is the outlook for future revenue from racks and what percentage of the backlog is for full systems?
A：The outlook for future revenue from racks is uncertain and dependent on customer demand for compute capacity. If customers need more capacity, revenue could continue to grow; if not, it may not. The current demand being seen is what will dictate future revenue. As for the backlog, the exact percentage of full systems within the script month backlog is not provided, but the focus is on the compute capacity needs of customers over the next period.
Broadcom Inc.
Follow
