Okay, great. We'd like to get started, if you can take your seats. Good afternoon. My name is Toshiya Hari. I cover the semiconductor and semi-cap equipment space at Goldman Sachs. Thank you all for coming. It's a great pleasure, a great honor to have Kirsten Spears from Broadcom. She's the Chief Financial Officer and Chief Accounting Officer of the company. Kirsten, thank you so much for coming.
Thank you, Toshiya. I just wanted to say I'm happy to be here today. I've been with Broadcom since 2014. I was acquired as part of the LSI transaction, and I've really enjoyed the journey with Broadcom, as you can imagine.
Great. Certainly want to spend a bunch of time on some of the longer-term, more strategic aspects of the business, but wanted to kick off with a near-term question. You know, you reported results last week, very strong results, you know, very strong outlook. Maybe, you know, spend a couple of minutes speaking to what stood out in the quarter, any product areas, customer types, geographies that sort of, you know, stood out in the quarter. That would be great. Thank you.
I'd be happy to do that. The overall macroeconomic environment is super dynamic, and from where I sit today, I believe we're executing very well in this environment. We just reported last week, and to summarize a few highlights, we reported Q3 revenues of $8.9 billion, up 5% year-over-year, with our semi and infrastructure software businesses also up 5% year-over-year. Hyperscale continued to grow double digits year-over-year, but enterprise and telco did moderate. By end market, our businesses performed exactly as we expected and in line with expectations.
We did provide some revenue detail on generative AI, stating it represented just over $1 billion in our Q3, and we also stated that excluding the benefit of generative AI, our semi business was approximately flat year-over-year and has stabilized at around $6 billion for the past three quarters. Free cash flows in the quarter were 52% of revenue, and we've stated we expect cash flows to be strong in the Q4.
I think what's really important that I point out, just as we kick things off, and it seems to be a common theme with investors in the questions that they ask, it's just important for me to reiterate that we really manage the ecosystem in terms of inventory. We really believe, you've heard us say, we scrub our backlog. We believe we're shipping to real end demand.
We believe our revenue line represents real end demand. We do not want to overship into the ecosystem. We work really hard with all of our customers, distributors, OEMs, and customers directly, to make sure that we're not doing that. We do our best there. That's really important to Broadcom.
Great. Thank you for that. We'll definitely come back to the scrubbing the backlog bit. But before we do, wanted to ask you a couple questions on AI. Obviously continues to be a very, very hot topic throughout the conference, throughout the sector. You talked about your AI business on the call, and I think you said it was roughly 15%, or will be 15% of your semiconductor business this year, and possibly more than 25% in the out year. Before we dive into the individual products and your strategy, maybe level set the audience by, you know, describing what role Broadcom plays and what you enable in AI.
Certainly. In order to support, the large or elephant workloads that generative AI demands, the network really becomes a computer. More specifically, generative AI requires large-scale networks to transfer packets of data in a lossless and low-latency manner, and in a manner that is open in terms of the architecture and scalable. This is where Broadcom plays. We're in the network, and we are very well-positioned with some of the best products out there in order to enable generative AI engines. Now, the two largest areas where we participate are, number one, custom silicon for compute offload, and with our Ethernet switching and routing platforms. We also provide, next-generation PCI Express and optical components for generative AI data center infrastructure.
Given what you just described, you must, you know, engage with a broad set of customers, whether it be in the enterprise or your hyperscale customers. Can you give us a feel as to how, you know, customer engagements have evolved since the beginning of the year?
Since the beginning of the year. I just want to step back and just say that, now, AI's been around for 20 years, right? But generative AI, we see infrastructure spending coming mostly from hyperscalers. It's not quite at enterprises just yet, so just the hyperscalers. And engagements with our hyperscaler customers are going very well. And we're currently benefiting from the race to develop and scale these large language models.
Got it. You talked a little bit about your custom compute business, which I think today is $2 billion in revenue a year. I guess, A, what are your growth expectations for the custom compute market? And I think today you primarily service one customer. What would you need to see to pursue customers outside of that one customer in custom compute?
Well, my point of view, you know, Broadcom is an engineering-first company. We're always open to customers who value our leading engineering capabilities and execution and want to have deep strategic and multi-year sustainable relationships with us. As you mentioned, we have one major customer for offload compute silicon today. On our call last week, we highlighted that we are excited that generative AI is pushing our world-class engineers to develop cutting-edge technology in silicon that has not been developed before. I mean, Hock mentioned that, and it is exciting to see inside of Broadcom, really. For example, we're investing in industry-leading high-speed SerDes, not only in optics but even copper cables... we are investing in high bandwidth memory, high speed and ultra-low power chip-to-chip connectivity, and complex 2.5D and 3D packaging.
Got it. So I think, you know, custom versus merchant is something that, you know, we get questions on, and it's a debate among investors. I think Hock has held the view that merchant silicon will ultimately win. I'm curious, has his thoughts or the company's thoughts evolved over the past 6-9 months since the emergence of GenAI?
You know, it's our view that merchant silicon will continue to be in higher demand than custom and will win out, as Hock says. It's just that when Broadcom sells merchant silicons, we're selling our leading-edge technologies to multiple verticals and end markets. But when you're working on an ASIC engagement, you're ultimately just selling that custom solution to one customer, right? So it's you know, you don't have the scale that you would get with merchant, so we do believe merchant would win in the end. But having said that, we do have very deep and strategic engagements with our silicon customers in the custom space, and so while I think merchant will ultimately win in the end, there is upside on the ASIC side. Obviously, you're seeing that now in GenAI. So you decide.
Got it. And then I guess, you know, transitioning a little bit to the networking side. On the earnings call, you guys talked about, you know, receiving substantial orders for some of your next-generation products. You know, you talked a little bit about, you know, the Tomahawk 5 switches, the Jericho 3 routers. Can you remind us where your hyperscale customers are, specifically hyperscale customers are in that 400-800 gig transition? And how should we think about 1.6 T going forward?
Ah! Well, our hyperscale customers are deploying Tomahawk 5 for 800G deployments, and we expect this to be mainstream for hyperscaler deployments in 2024 and in 2025. It is still early on 1.6T. That's currently in development now, but early adoption could start towards the end of 2025 and beyond.
Got it. Ethernet versus InfiniBand is another sort of topic or debate in the financial markets today. You know, NVIDIA has a lot of mind share with AI emerging. What is your view on this topic? Where do you see this headed? And from a Broadcom standpoint, you know, is InfiniBand and the growth there a threat to your business, or are you agnostic to that dynamic?
I think we've always said that InfiniBand will coexist with Ethernet. But having said that, I think when you start talking about generative AI networks, we generally believe that Ethernet is the best networking protocol to scale out large AI clusters, and we have the best technology. So we definitely think Ethernet's the answer to generative AI, and we believe we've got the right products there. Our Ethernet is lossless and has low latency. Most importantly, it's able to scale to tens of thousands of compute nodes and beyond, and it's based on an open standard architecture and a broad ecosystem. So that's kind of the view on that.
Got it.
Yeah.
Shifting gears a little bit on the non-AI side of your semiconductor business, you talked about weakness across, you know, enterprise customers, telco customers, last week on your call. But at the same time, you seemed to imply that the outlook for non-AI semis is, you know, $6 billion, give or take, over the next couple of quarters. First off, like, is it fair to say, and I think you talked about this earlier, but is it fair to say that you've achieved sort of a, quote-unquote, "soft landing?
Yes. I think a year ago, we signaled we expected a soft landing in 2023 and believe our non-AI semi revenues have remained stable at around $6 billion for the last three quarters. In terms of inventory, as I mentioned in my opening remarks, we really are careful not to overship into the ecosystem, so I don't believe we have excess inventory in the ecosystem. In terms of our own inventory on hand, we just announced last week that we exited the Q3 with 80 days of inventory on hand, which was down from 86 days in the Q2, and it's consistent with the range that we have maintained for over the last couple of years.
Okay. Got it. Your, your wireless business, you know, continues to be an important part of your portfolio. You know, revenue, I think you guided revenue to be down in the low single digits on a year-over-year basis in the current quarter. Maybe give us an idea as to how you're thinking about units versus content growth.
There's really only, you know, one customer here. We sell to this customer really on our superior technology. Our relationship with this customer is strategic, multi-year, and long-term in nature, which is customary with silicon engagements. We supply custom products to this customer across their entire product line. These include Wi-Fi, Bluetooth, GPS, high-end RF filters, inductive charging, among the few, and touchscreen controllers. These are all very unique to this particular customer. In terms of content increases, while I cannot comment on specifics year on year, I expect content to increase steadily, low single digits, but to steadily increase.
The low single-digit growth is going forward as well?
Yeah.
Gotcha.
Yes.
Okay, got it. Again, kind of sticking to the wireless business, clearly, you know, historically and going forward, it's really content growth that drives the business. But curious how you're thinking about, you know, the smartphone replacement cycle. There's, you know, hope that in year 2, year 3, things like AI could potentially come into the picture and drive a replacement cycle. Is that something that you guys think about?
... You know, I have to be really careful with this particular customer. I can't really talk about anything coming up that's new. I can tell you that in terms of wireless connectivity in general, I'm excited about Wi-Fi 7. Wi-Fi 7 will get devices to even higher speeds with higher bandwidth, and it can also reduce latency. And this, in turn, will drive better gaming and immersive internet experiences for customers. But you know, think about it with all of our products, with all of the end market technologies that we have, we work on improvements in power, which adds value by driving down power requirements and improving overall experiences for the end markets.
Got it. The geopolitical backdrop, you know, has been a challenging one, and it continues to be, again, a part of the debate. Remind us what percentage of your business is currently in China or consumed in China. I'm curious if this evolution in US-China relations has driven any change to your strategy, if at all?
No. I mean, it really hasn't impacted our business. I think that. Remember that we're not in GPUs or CPUs, and we think that end consumption represents roughly 10% of semiconductor revenue in China.
Okay.
It's not huge.
Okay. Got it. Shifting gears a little bit, on the software side, so you kicked off your journey in software with the acquisition of CA. That was about five years ago. You followed it up with your purchase of Symantec's enterprise security business about a year after. You've grown the business at a very steady pace. How would you grade your team's performance, looking back? And maybe talk about your long-term growth aspirations in software business as well.
Well, in terms of acquisitions, you know, we buy the right technology asset. That can either be in semiconductor or in the software space. It's about getting the right asset that has the key characteristics we are looking for. It has to be an incumbent leader in the markets that they participate in. It has to be a sustainable franchise with line of sight into to maintain leadership in the long term. We look to drive 10% cash-on-cash returns, and of course, the assets need to be actionable. In terms of your question on the software businesses, we have five divisions, and I think they're doing really well. As you pointed out, our software business drives strong cash flows and has steadily grown year-over-year.
Our strategy in our core software business, and you've heard Hock talk about this many times, is to target the largest 600-700 customers, and we work with them to expand their use and adoption of our software. It's really a renewals business. This strategy enables us to focus our operating spend dollars on just a targeted group of customers, enabling us to generate high operating margins, upwards of 70%. So I think they're doing pretty well.
Got it. And assuming, I know this is hypothetical, but assuming that the current pending, you know, acquisition is ultimately approved, would you consider additional acquisition in the software space? Is that something you'd talk about internally?
We would, semiconductor or software.
Okay.
Yep.
Got it. Perfect. So moving on to the financials, you know, gross margins, puts and takes, again, is something that I often get questions on. Roughly 70% in your semiconductor business, which is top-notch, within the industry. That said, I think margins have been coming down a little bit. Is that primarily mix that's causing that, or are there other dynamics at play?
That's generally product mix, truly. Yep, definitely. I mean, there's different product margins in each of our end markets, and within each end market, every single product will have different margins, so there's a lot of variability there. So it just definitely comes down to product mix.
Okay. As you look forward, I think, you know, both yourself and Hock, as you know, have talked about, you know, 50-150 basis points of annual improvements in gross margins. I know some years you do better, some years it's a little bit slower, but is that still the right target as you think about gross margins in your semiconductor business going forward?
Absolutely. It's definitely the correct range to look at when you consider our gross margins going forward. I think... Remember that, I think, what we're going through now is, in 2021 and 2022, we benefited from really early adoption of next-generation technology. Part of that had to do with COVID and people working from home, but, so, so we really benefited from that during those years. And so the long-term view, you definitely need to consider the 50-150 basis point increase. That, that'll happen, but ultimately, it's product mix in any given quarter, any given year.
Outside of product mix, any other potential drivers as it pertains to gross margin, or is it, again, primarily just mix that drives-
At-
50-150 basis points?
Yes, at the end of the day, mix, even-
Yep
... you know, with the macroeconomic environment that we have, it still ultimately comes down to product mix. Yep.
Got it. The other question that I often get, kind of moving down the P&L, you know, is: Is Broadcom underinvesting from an R&D perspective? Given your track record across all your businesses, you know, you're very competitive. You've outgrown the market, so you've done really well. Because you've been so efficient with OpEx, that is a question we get. How would you respond to that?
Well, let me step back and explain my view. We're a technology company. We are innovators. Of the 20,000 employees we have, about 80% of them are engineers, so it's a very engineering-first company. Our business model is really about focus and execution. We have 22 different franchises, 17 are in semiconductor and 5 in software. The way to look at this is each of the general managers of these businesses have autonomy to invest in R&D as they need to maintain their technology and product leadership. As you have heard Hock and now Charlie say, there are guardrails, though, when you consider R&D spending. The lens that I'd like you to look through this at is we invest to maintain our incumbency in the markets we participate in.
To us, incumbency is everything, and because we are well financially managed, we guide the spending to be focused on our core competencies instead of really chasing random moonshots. In fiscal 2022, we spent $5 billion on GAAP R&D, and so we believe we are investing properly in research and development. It's just very focused.
Yep, got it. Before, when we talked about the geopolitics and that having impact on your strategy, obviously you said no, but curious how you're thinking about your foundry strategy. You have a strong relationship, very deep relationship with TSMC. It's been a win-win for many years. Given sort of the ongoing tensions in the region, in Asia, that is, is there a plan B in place when you kind of debate your foundry strategy internally? Or is TSMC's plans to build capacity in the United States good enough from your vantage point?
We do have a strong relationship with TSMC, and we disclose this in our filings. 75% of our wafers are purchased from TSMC, and so it's a great strategic, you know, deep engagement that we have with TSMC. Now, having said that, I don't think we'd be doing, you know, ourselves, you know, a good risk management if we didn't look at other foundry partners as other opportunities, and we do. We do absolutely look at them.
But what I would point out there is that, you know, it takes a very long time to transfer process technology from one fab to another, be it Samsung, Intel, or even TSMC in Arizona. It just takes a long time. It's not a lift and quickly shift. It's definitely, you know, engineers on both sides have to get together. You've got to get that process working perfectly with all the know-how, the cell library transfer and everything else. So it's not easy to do that, but yes, we do consider other partners.
Okay, great. Kirsten, I know we've got to close a little earlier than scheduled, but maybe one last question on capital allocation. You know, you have a policy of returning 50% of trailing twelve months free cash flow to shareholders. You've been extremely consistent with that policy. Is there a debate internally as to whether you can go higher, just given how much cash flow you've been generating?
You know, our capital allocation strategy is going to remain the same. We're going to be returning, you know, to the target of returning approximately 50% of our prior year free cash flows, you know, in the form of dividends. We increased our annual dividend in 2023 for the twelfth consecutive year since we initiated dividends in 2011. If you look at the CAGR of dividend growth from FY16 to FY23, it's about 38%.
Really, it's in generating more free cash flows for our shareholders, right? That then, in turn, would bring more dividends, essentially. And so to do that, you know, that's going to be the-- in the form of continually to carefully financially manage the business and improve profitability and through future M&A. And of course, we are committed to maintaining our strong balance sheet and our investment-grade rating.
Okay. I guess in the last couple of minutes, you know, you interact with a bunch of investors and analysts. Anything that we collectively overlook or underappreciate about, you know, the Broadcom story specifically, or anything about the markets you play in?
I think we're a technology company first and foremost. You know, we have a strong IP portfolio, and we have the deepest and broadest portfolio of silicon IP out there, and other than that, I really appreciated being here today.
Awesome. Thank you so much.
Thanks.
Appreciate the time.