Broadcom Inc (AVGO) — Q2 2023 Earnings Call Transcript

Thank you, operator, and good afternoon, everyone. Joining me on today’s call are Hock Tan, President and CEO; Kirsten Spears, Chief Financial Officer; and Charlie Kawwas, President, Semiconductor Solutions Group. Broadcom distributed a press release and financial tables after the market closed, describing our financial performance for the second quarter fiscal year 2023. If you did not receive a copy, you may obtain the information from the Investors section of Broadcom’s website at broadcom.com. This conference call is being webcast live and an audio replay of the call can be accessed for one year through the Investors section of Broadcom’s website. During the prepared comments, Hock and Kirsten will be providing details of our second quarter fiscal year 2023 results, guidance for our third quarter, as well as commentary regarding the business environment. We will take questions after the end of our prepared comments. Please refer to our press release today and our recent filings with the SEC for information on the specific risk factors that could cause our actual results to differ materially from the forward-looking statements made on this call. In addition to U.S. GAAP reporting, Broadcom reports certain financial measures on a non-GAAP basis. A reconciliation between GAAP and non-GAAP measures is included in the tables attached to today’s press release. Comments made during today’s call will primarily refer to our non-GAAP financial results. I’ll now turn the call over to Hock.

Thank you, Ji, and thank you, everyone, for joining us today. So, in our fiscal Q2 2023, consolidated net revenue was $8.7 billion, up 8% year-on-year. Semiconductor Solutions revenue increased 9% year-on-year to $6.8 billion. And Infrastructure Software grew 3% year-on-year to $1.9 billion as the stable growth in core software more than offset softness in the Brocade business. Now, as I start this call, I know you all want to hear about how we are benefiting from this strong deployment of generative AI by our customers. Put this in perspective, our revenue today from this opportunity represents about 15% of our semiconductor business. Having said this, it was only 10% in fiscal '22. And we believe it could be over 25% of semiconductor revenue in fiscal '24. In fact, over the course of fiscal '23 that we're in, we are seeing a trajectory where our quarterly revenue entering the year doubles by the time we exceed '23. And in fiscal third quarter '23, we expect this revenue to exceed $1 billion in the quarter. But, as you well know, we are also a broadly diversified semiconductor and infrastructure software company. And in our fiscal Q2, demand for IT infrastructure was driven by hyperscale, while service providers and enterprise continued to hold up. Following the 30% year-on-year increases we have experienced over the past five quarters, overall IT infrastructure demand in Q2 moderated to mid-teens percentage growth year-on-year. As we have always told you, we continue to ship only to end-user demand. We remain very disciplined on how we manage inventory across our ecosystem. We exited the quarter with less than 86 days on hand, a level of inventory consistent with what we have maintained over the past eight quarters. Now, let me give you more color on our end markets. Let me begin with wireless. As you saw in our recent 8-K filing, we entered into a multiyear collaboration with a North American wireless OEM on cutting-edge wireless connectivity and 5G components. Our engagement in technology and supply remains deep, strategic, and long term. Q2 wireless revenue of $1.6 billion represented 23% of semiconductor revenue. Wireless revenue declined seasonally, down 24% quarter-on-quarter and down 9% year-on-year. In Q3, as we just begin the seasonal ramp of the next-generation phone platform, we expect wireless revenue to be up low single digits sequentially. We expect however, it will remain around flattish year-on-year. Moving on to networking. Networking revenue was $2.6 billion and was up 20% year-on-year, in line with guidance, representing 39% of our semiconductor revenue. There are two growth drivers here. One, continued strength in deployment of our merchant Tomahawk switching for traditional enterprise workloads as well as Jericho routing platforms for telcos; and two, strong growth in AI infrastructure at hyperscalers from compute offload and networking. And speaking of AI networks, Broadcom’s next-generation Ethernet switching portfolio consisting of Tomahawk 5 and Jericho3-AI offers the industry's highest performance fabric for large-scale AI clusters by optimizing the demanding and costly AI resources. These switches based on an open distributed disaggregated architecture will support 32,000 GPU clusters running at 800 gigabit per second bandwidth. Ethernet fabric, as we know it, already supports multi-tenancy capability and end-to-end congestion management. This lossless connectivity with high QoS performance has been well proven over the last 10 years of network deployment in the public cloud and telcos. In other words, the technology is not new. And we are, as Broadcom, very well positioned to simply extend our best-in-class networking technology into generative AI infrastructure while supporting standard connectivity, which enables vendor interoperability. In Q3, we expect networking revenue to maintain its growth year-on-year of around 20%. Next, our server storage connectivity revenue was $1.1 billion or 17% of semiconductor revenue and up 20% year-on-year. And as we noted last quarter, with the transition to next-generation MegaRAID largely completed and enterprise demand moderating, we expect server storage connectivity revenue in Q3 to be up low single digits year-on-year. Moving on to broadband. Revenue grew 10% year-on-year to $1.2 billion and represented 18% of semiconductor revenue. Growth in broadband was driven by continued deployments by telcos of next-generation 10G PON and cable operators of DOCSIS 3.1 with high attach rates of Wi-Fi 6 and 6E. And in Q3, we expect our broadband growth to moderate to low-single-digit percent year-on-year. And finally, Q2 industrial resales of $260 million increased 2% year-on-year as the softness in China was offset by strength globally in renewable energy and robotics. And in Q3, we forecast industrial resales to be flattish year-on-year on continuing softness in Asia, offset by strength in Europe. So summary, Q2 Semiconductor Solutions revenue was up 9% year-on-year. And in Q3, we expect semiconductor revenue growth of mid-single-digit year-on-year growth. Turning to software. In Q2, Infrastructure Software revenue of $1.9 billion grew 3% year-on-year and represented 22% of total revenue. As expected, continued softness in Brocade was offset by the continuing stable growth in core software. Relating to core software, consolidated renewal rates averaged 114% over expiring contracts. And in our strategic accounts, we averaged 120%. Within the strategic accounts, annualized bookings of $564 million included $133 million or 23% of cross-selling of other portfolio products to these same core customers. Over 90% of the renewal value represented recurring subscription and maintenance. And over the last 12 months, consolidated renewal rates averaged 117% over expiring contracts. And among our strategic accounts, we averaged 128%. Because of this, our ARR, the indicator of forward revenue at the end of Q2 was $5.3 billion, up 2% from a year ago. And in Q3, we expect our Infrastructure Software segment revenue to be up low single digits percentage year-on-year as the core software growth continues to be offset by weakness in Brocade. On a consolidated basis, we're guiding Q3 revenue of $8.85 billion, up 5% year-on-year. Before Kirsten tells you more about our financial performance for the quarter, let me provide a brief update on our pending acquisition of VMware. We're making good progress with our various regulatory filings around the world, having received legal merger clearance in Australia, Brazil, Canada, South Africa, and Taiwan and foreign investment control clearance in all necessary jurisdictions. We still expect the transaction will close in Broadcom's fiscal 2023. The combination of Broadcom and VMware is about enabling enterprises to accelerate innovation and expand choice by addressing their most complex technology challenges in this multi-cloud era. And we are confident that regulators will see this when they conclude their review. With that, let me turn the call over to Kirsten.

Thank you, Hock. Let me now provide additional detail on our financial performance. Consolidated revenue was $8.7 billion for the quarter, up 8% from a year ago. Gross margins were 75.6% of revenue in the quarter, about 30 basis points higher than we expected on product mix. Operating expenses were $1.2 billion, down 4% year-on-year. R&D of $958 million was also down 4% year-on-year on lower variable spending. Operating income for the quarter was $5.4 billion and was up 10% from a year ago. Operating margin was 62% of revenue, up approximately 100 basis points year-on-year. Adjusted EBITDA was $5.7 billion or 65% of revenue. This figure excludes $129 million of depreciation. Now, a review of the P&L for our two segments. Revenue for our Semiconductor Solutions segment was $6.8 billion and represented 78% of total revenue in the quarter. This was up 9% year-on-year. Gross margins for our Semiconductor Solutions segment were approximately 71%, down approximately 120 basis points year-on-year, driven primarily by product mix within our semiconductor end markets. Operating expenses were $833 million in Q2, down 5% year-on-year. R&D was $739 million in the quarter, down 4% year-on-year. Q2 semiconductor operating margins were 59%. So, while semiconductor revenue was up 9%, operating profit grew 10% year-on-year. Moving to the P&L for our Infrastructure Software segment. Revenue for Infrastructure Software was $1.9 billion, up 3% year-on-year and represented 22% of revenue. Gross margins for Infrastructure Software were 92% in the quarter and operating expenses were $361 million in the quarter, down 3% year-over-year. Infrastructure Software operating margin was 73% in Q2, and operating profit grew 8% year-on-year. Moving to cash flow. Free cash flow in the quarter was $4.4 billion and represented 50% of revenues in Q2. We spent $122 million on capital expenditures. Days sales outstanding were 32 days in the second quarter compared to 33 days in the first quarter. We ended the second quarter with inventory of $1.9 billion, down 1% from the end of the prior quarter. We ended the second quarter with $11.6 billion of cash and $39.3 billion of gross debt, of which $1.1 billion is short term. The weighted average coupon rate and years to maturity of our fixed rate debt is 3.61% and 9.9 years, respectively. Turning to capital allocation. In the quarter, we paid stockholders $1.9 billion of cash dividends. Consistent with our commitment to return excess cash to shareholders, we repurchased $2.8 billion of our common stock and eliminated $614 million of common stock for taxes due on the vesting of employee equity, resulting in the repurchase and elimination of approximately 5.6 million AVGO shares. The non-GAAP diluted share count in Q2 was $435 million. As of the end of Q2, $9 billion was remaining under the share repurchase authorization. Excluding the potential impact of any share repurchases, in Q3, we expect the non-GAAP diluted share count to be $438 million. Based on current business trends and conditions, our guidance for the third quarter of fiscal 2023 is for consolidated revenues of $8.85 billion, and adjusted EBITDA of approximately 65% of projected revenue. In Q3, we expect gross margins to be down approximately 60 basis points sequentially on product mix. That concludes my prepared remarks. Operator, please open up the call for questions.

Hock, I might just as well start off with the topic that you started, AI these days is everywhere. Thanks for the color that you gave and the percentage of the sales that it was potentially going to represent into the future. I wanted to just get a little bit more color on two aspects of that. How you've seen the demand evolve during the course of your quarter? Has it accelerated, in what areas, et cetera? And is there any competitive implications for it? We've heard from some of the compute folks that they want to do more on the networking side. And then obviously, you want to do more into the compute side. So I just wondered how the competitive intensity is changing, given the AI workload increases these days.

Sure. Regarding your first question, yes, during the last earnings call, we indicated a strong sense of demand, and that has continued without interruption. However, we all understand that manufacturing lead times for most of these advanced products are quite long. Producing these products typically takes at least six months. While there is significant demand and urgency, our capacity to ramp up production will be cautious, focusing on the most immediate needs. As for your second question, we have always faced competition. Even in traditional environments like enterprise and hyperscale data centers, our segments in networking, switching, and routing experience competition. This is nothing new; competition persists, and we each strive to excel in our respective strengths.

Hock, I just wanted to first clarify. I think you might have mentioned it, but I think last quarter, you gave very specific numerical targets of $3 billion in ASICs and $800 million in switches for fiscal '23. I just wanted to make sure if there is any specific update to those numbers. Is it more than $4 billion in total now, et cetera? And then my question is, longer term, what do you think the share is going to be between kind of general-purpose GPU-type solutions versus ASICs? Do you think that share shifts towards ASICs? Do you think it shifts towards general purpose solutions? Because if I look outside of the compute offload opportunity, you have generally favored, right, more the general-purpose market. So I'm curious, how do you see this share between general purpose versus ASICs play out in this AI processing opportunity longer term?

In response to the first part of your question, we have indicated that for fiscal year '23, we are expecting revenue in this area to be $3.8 billion. We don't plan to change that forecast mid-year. We maintain that forecast for fiscal '23 and are providing some insights into what we see for '24, which reflects a general trajectory. This is not a detailed forecast but rather a sense of an accelerated trend from '22 through '24. Regarding the specific numbers you've mentioned, we stand by our fiscal '23 forecast of 3.8%, as it's still early for any revision. As for your broader question about ASICs versus merchant solutions, I generally prefer merchant options, whether in computing or networking, as I believe they have a better chance of succeeding in the long run. However, today's discussion pertains to shorter-term issues. While there is a presence of compute offloading, the number of players in compute offload ASICs remains very limited, which is a trend we continue to observe.

It's great to see the strong and growing ramp of your AI compute offload and networking products. Regarding your next-generation AI and compute offload programs that are currently in the design phase, you have next-gen switching and routing platforms that are being qualified. Are your customers continuing to urge the team to speed up the design process and adjust the program ramp timing? Additionally, I believe you may have addressed this, but I wanted to clarify that all of these solutions utilize the same advanced packaging technologies, such as stack die and HBM memory packaging. Not surprisingly, this is the same architecture employed by your AI GPU peers, which are experiencing similar strong trends. So, is the Broadcom team currently facing or expected to face constraints in advanced packaging and substrate supply? How will the operations team manage through these challenges?

You are correct that this type of AI product, including both next-generation and current generative AI products, relies on cutting-edge technologies in silicon wafers and packaging, such as memory stacking. However, there are still products and capacity available, as I mentioned. This is not something that can be shipped or deployed immediately; it requires time. We anticipate a gradual increase, which began in fiscal '23 and will maintain its pace through '24.

And on the design win funnel, are you seeing customers still trying to pull in all of their designs?

Our main opportunity continues to be in the networking of AI networks. We have products available and are collaborating with numerous customers to establish this distributed disaggregated architecture using Ethernet fabric for AI. There is significant interest and many designs already in place.

Hock, I was wondering if you can sort of help shed some light on the general perception that all this AI spending is sort of boxing out traditional compute. Can you talk about that? Or is it that just CapEx budgets are going to have to grow to support all this extra AI CapEx? I mean, the trick is probably somewhere in between, but I'm wondering if you can help shed some light on just the general perception that all of this is coming at the expense of the traditional compute and the traditional infrastructure. Thanks.

Your guess is as good as mine. I can share this: you're correct about the increasing allocation of funds towards AI networks by hyperscale companies. However, this does not necessarily mean that enterprise spending is diminishing for traditional workloads and data centers. There will definitely be coexistence. Much of the current investment in AI we observe is concentrated in the hyperscale sector. Enterprises are still directing a significant portion of their budgets towards traditional data centers and x86 workloads. It might be too early for us to determine if this translates into cannibalization.

I have a less sexy topic to talk about, but obviously very important in how you manage the business. Can you talk about lead times and especially in the light of demand moderating, manufacturing cycle times coming down, not to mention the six months that you highlighted for the cutting edge? Are you still staying with the 52-week kind of lead quoting to customers, or has that changed? Thank you.

By the way, it's 50. Yes, my standard lead time for our products is 50 weeks, and we are still staying with it because it's not about as much lead time to manufacture the products as our interest and, frankly, mutual interest between our customers and ourselves to take a hard look at providing visibility for us in ensuring we can supply and supply in the right amount at the right time the requirements. So yes, we're still sticking to 50 weeks.

Yes. Hock, I was hoping you could clarify something for us. Earlier in the call, when you discussed AI, you mentioned that generative AI revenues are currently around 15% and are expected to reach 25% by the end of 2024. This seems to account for most of your growth, about $3 billion to $4 billion. While I recognize that your core business is performing well, I'm concerned that I might be misinterpreting this. I'm hoping there isn't significant cannibalization occurring in your business, and I would appreciate your clarification on this.

In response to an earlier question from a peer, we do not observe any cannibalization at this stage, although it is too early to make definitive conclusions, and budget changes do not happen quickly. If there were any cannibalization, it would depend on how spending priorities shift, but we do not have sufficient clarity to assert that there is cannibalization. Currently, as we look at the numbers, we see growth coming from various areas. As we discuss things in 2023, we continue to show some growth across our business and product lines. While our growth in AI revenue contributes to overall growth, it does not account for all of it. I would estimate that at least half of our growth still comes from our traditional business, while the other half is likely from generative AI.

Hock, you rightly pointed to the custom silicon opportunity that supports your cloud AI initiatives. However, your AI revenue that's not tied to custom silicon appears to be doubling in fiscal '23. And the outlook for fiscal '24 implies that it will double again. Obviously, Broadcom has multiple areas of exposure to AI really across PCI switches, Tomahawk, Jericho, and Ramon ASICs and electro-optics. I guess what sort of opportunity do you see your electric optics portfolio playing a role in high-performance networking environments for inferencing and training AI applications?

What you mentioned is very insightful. A significant part of our growth in AI comes from the networking components we provide for creating Ethernet fabrics for AI clusters. You've touched on a crucial point. The growth rate in this area is likely outpacing the growth of our offload computing. This is where we are concentrating our efforts, as our networking products are standard merchant offerings that are supporting the rapid expansion of generative AI clusters in computing. For us, the growth in networking is indeed the fastest segment of our overall growth.

I wanted to ask about the renewal of the wireless contract. Can you give us a sense for how much sort of concrete visibility you have into content over the duration of that? As you mentioned, it's both, RF and wireless connectivity. Just any the additional color you can give us would be great.

I don't want to be overly specific, but I would say this is an extension of our existing long-term agreement. It's best described as a collaboration and strategic arrangement rather than a renewal. The characteristics are quite similar, as we supply technology and various specific products related to 5G components and wireless connectivity, which is our area of strength and where we continue to lead in the marketplace. This arrangement is set for multiple years. Beyond that, I would direct you to our 8-K for more information, as there are sensitivities involved that prevent me from sharing additional details.

Hock, I'm curious how you're thinking about your semiconductor business long term. You've discussed AI pretty extensively throughout this call. Could this be something that drives higher growth for your semiconductor business on a sustained basis? I think historically, you've given relatively subdued or muted growth rates for your business vis-à-vis many of your competitors. Is this something that can drive sustained growth acceleration for your business? And if so, how should we think about the rate of R&D growth going forward as well? Because I think your peers are growing R&D faster than what you guys are doing today. Thank you.

That’s a very good question, Toshiya. We are still a broadly diversified semiconductor company, as I emphasized, with several end markets beyond just AI. Most of our AI revenue is in my networking segment, as you all are aware. We still have plenty of other areas. For fiscal '24, we believe AI could represent over 25% of our semiconductor revenue, but we have many solid foundations for the rest of our semiconductor business. For instance, our wireless business has a very strong outlook for multiple years, making it a significant part of our operations. The AI sector is just beginning to match its size. Additionally, our broadband server storage enterprise segment continues to be quite sustainable. Overall, we haven't updated our long-term forecast recently to provide any new insights. I really don’t have more to add than what we've previously communicated. As for whether this will impact our long-term growth rate, we haven't considered that yet. I’ll leave it to you to speculate before I put anything in writing.

Hock, I'm wondering if you can talk about your foundry relationships. You've got a very strong relationship with TSM. And of course, Intel has been very vocal about winning new customers potentially. I wonder if you can talk about your flexibility and openness and considering new partners. And then maybe also talk about pricing from foundry and whether that's influencing any changes quarter-to-quarter. There have been certainly a lot of price increases that we've heard about recently, and I'd love to hear your comments. Thank you.

Thank you. We are very loyal to our suppliers, just as we are with our customers. This loyalty works both ways, and we have a strong, enduring partnership with all our key suppliers. However, we must also be realistic about the current geopolitical environment. Therefore, we are open to exploring specific technologies to expand our supply base. We are consistently evaluating this while still aiming to be loyal and fair to our existing partners. Because of this partnership and loyalty, price increases are viewed as a long-term element of the relationship. Simply put, we do not make decisions solely based on price; we remain committed due to the support, service, and a strong mutual sense of commitment.

Hock, I have a housekeeping question. It seems your comments in the press release regarding the wireless deal did not mention Mixed Signal, which is part of your previous agreement. Additionally, everything you've mentioned today does not suggest that Mixed Signal will be a focus in wireless and RF, despite your ongoing work in that area. Can you clarify this? Also, why shouldn’t we anticipate that the increased interest in AI will enhance the prospects, if not immediate orders, for the electro-optic products being developed? Given the clusters and the sizes of the arrays people are trying to assemble, I would expect much greater demand, as they should provide significant benefits, particularly in terms of power. Could you provide some insights on this?

All right. You have two questions here, don't you? You're right. Our long-term collaboration agreement that we recently announced, it includes, as it indicated, wireless connectivity and 5G components. It does not include the high-performance analog components, mixed signal components that we also sell to the North American OEM customer. right? That doesn't make it any less, I would add, strategic, not deeply engaged with each other. I would definitely hasten to add. And on the second part, Ed, if you could indulge me, could you repeat that question?

You're discussing general AI and the rising demand from hyperscale customers. We can already see the potential size of these clients. While this isn't necessarily stressing our networking assets, the release of next year's electro-optic products, particularly Tomahawk 5, which integrates Tomahawks directly on the chip, might become more appealing due to its lower power requirements. Although it hasn't been deployed yet, I believe interest in this technology should grow. Is that accurate?

You're correct. As I mentioned earlier, we're looking forward to our next generation, Tomahawk 5, which will feature silicon photonics and co-packaging as crucial components. It will support a bandwidth of up to 51 terabits per second, making it ideal for high-demand AI networks, particularly those utilizing over 32,000 GPU clusters operating at 800 gigabits per second. Such networks require substantial switching capabilities to maintain low latency and virtually lossless transmission. Achieving Ethernet losslessness involves significant scientific and technological advancements, as traditional Ethernet has inherent limitations. However, our technology enables that losslessness. This aligns with our new generation of products, including Jericho3-AI, which offers unique technology for extremely low tail latency, efficiently managing packet transmission and reordering to minimize loss and latency. We are excited about applying this network routing expertise from telcos to AI networks with our innovative products. So, we are fully embracing this opportunity in the networking space, and we see the most compelling prospects within the AI sector.

I'll stick to a single-part question. Can you maybe double-click on your compute offload business? What can you maybe tell us about how growth could split between revenues from existing customers or potential diversification of that business going forward? Thank you.

Thank you for the question. I'll restate my earlier points for clarity. We have essentially one rail customer, and my previous comments regarding compute offload have largely centered on this single customer. The business is not very diversified; it remains highly focused. That's the situation with our compute offload business.

This is Kurt Swartz on behalf of C.J. I wanted to discuss software gross margins, which are continuing to improve despite the challenges in Brocade. I'm interested in your insights on the stabilization of Brocade and how we should approach software gross margins as the mix returns to normal. Thank you.

Our software segment consists of two parts: our core software product revenues, which are sold directly to enterprises, and typical infrastructure software products typically under multiyear contracts. We have significant backlog, around $17 billion, averaging nearly 2.5 to 3 years. Each quarter, a portion of that backlog renews, and we provide data on it. This segment is very stable, and based on our historical renewal patterns with an expanding customer base, we can drive this growth in a stable manner. The growth rate is very predictable, and we're satisfied with it. Additionally, we have Brocade's fiber channel SAN business, which, while also part of our software offerings, is primarily enterprise-driven and tends to be cyclical. Last year was a robust up cycle for this segment, but this year, the cycles are not as strong, especially when compared to last year’s strong performance. The combination of these factors is what we are currently experiencing. However, I believe that next year could see a turnaround, potentially leading Brocade to improve. Instead of a 3% year-on-year growth in the entire segment, we could achieve high single-digit growth, as I have previously noted that you should plan for long-term mid-single-digit growth in core software revenues, which remain a very predictable aspect of our financials.

Yes. Hi Hock, I have a two-part question to wrap up. I'm curious about the content uplift for Broadcom on an AI server compared to a general compute server. Additionally, regarding generative AI, what percentage of servers today are being equipped for generative AI? Given your strong market position, how do you anticipate that uptake ratio will evolve in fiscal '24 and '25?

I'm sorry to disappoint you on your two parts, but it's too early for me to provide a solid answer. The majority of servers in use today are traditional servers utilizing x86 CPUs, while networking is predominantly still based on traditional Ethernet in data centers. Most enterprises are still operating their own traditional servers on x86. Generative AI is relatively new, and the possibilities with it are vast. Currently, we mainly see hyperscale companies deploying generative AI infrastructure at scale. Enterprises continue to use standard x86 servers and Ethernet networking in their traditional data centers. What we're observing may just be the beginning of a larger cycle, which is why I cannot provide a definitive view on the attach rate, the ratio, or any potential stability in the near term. We see both traditional servers and generative AI coexisting for now.

Thank you, operator. In closing, we would like to highlight that Broadcom will be attending the BofA Global Technology Conference on Tuesday, June 6. Broadcom currently plans to report its earnings for the third quarter of fiscal '23 after close of market on Thursday, August 31, 2023. A public webcast of Broadcom's earnings conference call will follow at 2:00 p.m. Pacific. That will conclude our earnings call today. Thank you all for joining. Operator, you may end the call.