Broadcom Inc. (AVGO.US) Q4 FY2024 Conference Call: AI revenue grew by 220% this year, and semiconductor business is expected to recover.

December 12, Broadcom Inc. (AVGO.US) held its FY2024Q4 financial report conference call. CEO Hock E. Tan stated that the combined net income for the 2024 fiscal year was $51.6 billion, a 44% year-on-year increase. Excluding the impact of VMware's consolidation, the 2024 fiscal year was still able to achieve 9% organic growth. The strong performance this year was mainly due to growth in two areas. First, the acquisition of VMware was completed in the first few weeks of the 24 fiscal year, and VMware's focus was shifted to its leading position in data center virtualization technology. The integration of VMware is almost complete. Revenue is on a growth trajectory, and by 2024, the operating profit margin will reach 70%. Steady growth in adjusted EBITDA is being achieved and will significantly exceed the initial announced acquisition price of $8.5 billion. The goal is to achieve this target earlier than the original three-year target. The second driver of 2024 was artificial intelligence. Artificial intelligence revenue from custom AI accelerators or XPUs and networking strength grew by 220%, from $3.8 billion in fiscal year 2023 to $12.2 billion in fiscal year 2024, accounting for 41% of semiconductor revenue. This brought semiconductor revenue in 2024 to a record $30.1 billion. Before summarizing and providing guidance for the first quarter of the 2025 fiscal year, let's take a long-term perspective on the development of the semiconductor business over the next three years. With a wide-ranging semiconductor business portfolio, non-AI business reached a cyclical bottom in the 2024 fiscal year at $17.8 billion, and non-AI-related semiconductor business is expected to recover from this level at a mid-single-digit historical growth rate. Overall, it is expected that first-quarter consolidated revenue will be approximately $14.6 billion, a 22% year-on-year increase, which is expected to drive first-quarter adjusted EBITDA to approximately 66% of revenue. CFO Kirsten Spears stated that consolidated revenue for this quarter was $14.1 billion, a 51% year-on-year increase. Excluding VMware's contribution, Q4 revenue increased by 11% year-on-year. This quarter's gross profit margin was 76.9%. Research and development expenses were $1.4 billion, and consolidated operating expenses were $2 billion, a main reason for the year-on-year increase being the consolidation of VMware. Operating profit for the third quarter was $8.8 billion, a 53% year-on-year increase, and the operating profit margin was 63% of revenue. Excluding transition costs, operating profit was $9.1 billion, with an operating profit margin of 65% of revenue. Adjusted EBITDA was $8.2 billion, 63% of revenue. This figure does not include $156 million in depreciation. Regarding the Q1 guidance, consolidated revenue is expected to reach $14.6 billion, with semiconductor revenue at $8.1 billion and software revenue at $6.5 billion, and adjusted EBITDA is expected to be around 66%. For modeling purposes, it is expected that the consolidated gross margin will increase by approximately 100 basis points month-on-month, due to the higher revenue share of the high-margin combination of software business and intra-semiconductor products. [Analyst Q&A session] Q: Can you provide more detailed information on the growth trends of ASIC business and Ethernet business in next year's Q1 AI revenue? A: Both are growing, but at different rates. In the second half of this year compared to the first half, the Ethernet trend is growing faster. We believe this trend will continue in the first half of next year, but as 3nm XPU starts shipping in the second half of next year, this trend will change. Q: In the guidance for AI-related scale reaching $60-90 billion by 2027, can you break down the share between ASIC and network? A: First, it needs to be clarified that this is not our revenue volume, but our forecast market size. The network side accounts for approximately 15-20% share. Q: How do you evaluate your competitors for NVIDIA Corporation's Rack-level products? Can you also provide rack-level solutions? A: When you scale from a cluster of thousands of cards to 100,000, 500,000, or even a million, engineering-wise, it's a completely different problem. So, a rack-level solution will be the method necessary for the next step in scaling up. There's also scale out, which is still a continuously evolving engineering problem. But I think all of our ultra-scale customers are well aware of how to achieve this at this point. Therefore, the roadmap will progress from 100,000 XPU clusters to over a million. Over the next three to four years, we expect to see more and more similar architectures being built. Q: How do you view the match between cloud provider capital expenditures and the growth rate of your AI revenue? We expect the Capex spending of the four major cloud providers in the 25 fiscal year to increase by around 35-40%. Considering Ethernet taking share from InfiniBand and ASIC growing faster than GPU, can we assume that your growth rate will be faster than this? A: I don't think these two can be matched. I think large enterprises often provide you with a total capital expenditure number. I'm not sure whether they really differentiate between AI and non-AI, but clearly, AI spending exceeds non-AI spending, even in terms of capital expenditure. Q: In the next quarter, due to the delay of software customer orders, the gross margin rate will increase, but will this impact gradually diminish after Q2? Will the gross margin be under pressure after Q2? A: This part of the order volume is very small, just a temporary delay, and it will not have a profound impact on the gross margin after Q2. Q: Regarding the AI related scale reaching $60-90 billion in 2027, is this an accumulated value or an annual market size? How much of a market share do you expect to have in this market? If you also consider the other two potential customers, how much additional market size will it bring? A: It is just the market size for the fiscal year 2027. If potential customers become actual customers, our market share would be better than anticipated and additional market size would be significant.Progress has indeed been made, but I do not want to discuss this issue in depth at the moment.Q: How is the free cash being used? A: It is mainly used to reduce debt leverage. Considering the amount of debt we are currently carrying or have carried since the acquisition of VMware, we do intend to use a portion of the 50% of free cash flow not used for dividends towards our own deleveraging. Q: How do you view the AI market size for 2024, in order to understand your current market share in the AI market? How do you anticipate your future market share changing? Also, as your AI revenue in semiconductor solutions increases, how will your gross margin evolve? A: This is a very insightful question. In regards to the first question, we talked about a baseline of $600-900 billion within three years and specified our three clients. I estimate the AI reachable market size for 2024 to be below $200 billion, around $150-200 billion. As for profit margins, do not worry too much about gross margins, you are correct that the gross margin in semiconductors may dilute, but please note that our operating leverage will increase significantly, leading to an improvement in operating profit margins compared to our current levels. Q: What is the ratio between ASIC and networking spending? How much networking spending corresponds to $1 in ASIC spending? Are there sovereign customers interested in ASIC? A: Sovereign nations are more like enterprise customers, as they do not have the capability to complete full-stack hardware and software development. The ratio relationship between ASIC and networking is constantly changing. With the expansion of clusters, the demand for Scale-up and Scale-out will increase. So we believe that the current network ratio may be around 5-10%, but as clusters expand to 500,000 cards or even 1 million cards, this ratio will rise to 15-20%. Q: Considering the guideline of a $600-900 billion market size, could you talk about your expectations for the entire TAM of AI? For the parts you are not involved in, do you think it will be absorbed internally by cloud vendors, or will there be a second supplier, or is it low-margin business that you do not want to pursue? A: We are not clear about the overall TAM of AI. Our calculation of our TAM is this: I find customers, I try to figure out how many roadmaps the customers have, not just products, technologies, but also what they are building and how they consume, that's how we create our SAM. In a way, it is bottom-up, then top-down. Therefore, apart from our closely serviced and collaborated clients, I do not know how much TAM exists in other areas. What we need to do is get our fair share. Today, we are in a very advantageous position, with the best technology, and so far, we have one of the best combination technologies for making XPU and connecting these XPUs, which is why we are in a favorable position with these three clients. Q: With the explosion of AI revenue, will interest in M&A diminish in the future? A: No, it will not. We always maintain an open attitude, and whenever we come across very high-quality technical patented products, we actively add them to our business structure, whether it is software or hardware, as long as they meet our quite stringent standards, we are always willing to acquire these assets and add them to our portfolio. Q: Based on your recent figures, with a TAM of $175 billion this year, you hold approximately 70% of the share. So assuming 70% share will continue to 27 years, your AI revenue in fiscal year 27 should be around $500 billion. So, do we have good expectations for fiscal year 26, showing a fairly good growth momentum? A: I do not have the appropriate information to share with you in this regard. Considering that we will only provide guidance for AI and non-AI businesses in the future, for the non-AI business part, they are already stable and mature, with a high probability of mid-single-digit growth in the future. However, for non-AI revenue, we cannot give you a clear linear guidance, as the deployment pace of each client will vary, and we expect differences between quarters. So, my best answer to you is that besides what I have provided so far, I cannot give you any clear information. Q: For the other two potential clients, when do you think their revenue growth will begin? A: First, we must obtain orders to start production before we can discuss the start of revenue growth. Their software stack is not yet complete, they still need validation, so right now I am not sure, but it will definitely happen within the next three years.