Broadcom Inc. (AVGO) Earnings Call Transcript & Summary

Good afternoon, everyone. I'm Vivek Arya. I cover semiconductors and semicap equipment at BofA Securities. And I'm really delighted that and honored to have our afternoon keynote speaker, Hock Tan, CEO of Broadcom. And Hock doesn't need a lot of introduction, but suffice to say, under Hock's leadership, Broadcom has sort of what I consider been a role model in the industry in terms of driving really disciplined, profitable, diversified growth, and we are really happy that he's able to come and share his perspectives with us.

Thank you, Vivek. Glad to be here.

Thank you, Hock.

Maybe not so glad, but yes, feel good to be here.

We will see. Let's go through the questions. It will make you happier.

Yes, that's what I mean.

So Hock, maybe let's just start at the top. Give us a state of the union in that there seems to be sort of industry dynamic where the semiconductor industry is sounding very strong. But on the macroeconomic side, that there appear to be a lot of cross cut and in terms of weaker consumer and so forth. So how do you see the demand picture today?

Okay. Well, I don't pretend to be an economic's [indiscernible]. And I'm sure the macro economy at some point could impact even the sector of the economy within Broadcom sits in. But I won't give you a sense of what we are seeing at Broadcom. And cut to the chase, and where we're sitting, obviously, for those of you sitting out here, at the risk of repeating what you already know, we are the leading -- one of the leading providers of infrastructure technology to the digital economy, simplest way to say that. And what we're seeing today is demand has been for the last 2 years almost now, 2.5 years, call it, definitely in '21, late '20, so far in 2022, demand continues to exceed supply. And as far as we can see, that's going to continue through to at least late '23, if not early '24. And I'm saying that simply because like many of our peers, we provide lead times for delivery of our components, our technology. And the lead times we have put in place, 15 weeks, 1 year, and that has been in place for 18 months. We haven't changed it because we intentionally do that to not disrupt or change behavior of our customers in the way they foresee and do their planning. And even up to end of last quarter, we track this very religiously. Our rate of bookings which accumulates into a level of backlog at the end of each quarter, each month continues to grow. The date when those backlog starts to decline or flatten out would probably be the time it tells us that demand out 50 weeks starting to roll over. And what we've seen, it hasn't.

Understood. And Hock, we will go through the different segments, but I'm really very interested in kind of taking one step back from when the company was, Avago, right, to the time that you went through the process of making very strategic acquisitions in the semiconductor industry and then in the software side. So take us through -- what was the strategy from that point in terms of first going through the semiconductor industry M&A and now the software industry M&A? How do you select these targets? What's the grand strategy at play here?

I would not necessarily dignify it as a strategy. But I guess you could say there's an overarching investment thesis behind our strategy, as you call it, is that technology and in the areas where our routes come from, semiconductors which is really technology embedded in little pieces of silicon, is a very deep profit pool. And so is infrastructure software, very deep profit pool. And the game for us simply is to find those opportunities, which -- where these pools exist, thinking about what I mean by a deep profit pool. It's products that cost us maybe $5 to manufacture, design of course and manufacture. And we sell it into the ecosystem. And at the end of the day, the last person who buys it to use it is, whether it's particularly enterprise pays $500. Somewhere between $5 and $500, it's a lot of value. And when I say it's deep profit pool, is we go in and mine that $495, and that's basically our overarching strategy. And the game is to find those assets out there that are very, very mission-critical to consumers that can add value in the way and come back to my earlier statement, why the area we're in, infrastructure, technology, whether it's in the form of semiconductors or software, why it appears at this point to fairly so far, recession proof is simply because there's some strong tailwind behind us. I mean enterprises globally are finding themselves in a situation where to be relevant, to be productive, competitive, even they have to expand if they haven't already create their digital footprint. And what -- who enables them? Technology, technology like the kind of technology Broadcom provides. And we have up to now 22 different product franchises, each of which have those characteristics have been very, very sustainable, very mission-critical and where we are the technology and market leader.

Got it. So I can understand the synergy from a financial perspective, right? That you are bringing diversification into the business and right, obviously, by optimizing the costs of the target, you are able to really improve its financial profile. What about technical synergy, right, that since the time you started an infrastructure software business, have you actually seen the semiconductor side benefit from that or vice versa?

Well, some of it, yes. But broadly at this point, mostly none. And a big part of it is simply because a lot of the enterprise, a lot of software assets we have acquired so far are still very enterprise-driven. And when we sell semiconductors, we sell through system integrators or original equipment manufacturers, people like HP, Dell, Cisco, who then puts it together in a box with operating software in many cases and resell with support to real end users, be it Bank of America, JPMorgan, Wells Fargo. So we are one step removed. In infrastructure software, however, we have to have and which we do direct access to the Bank of America, to the JPMorgan of the world where we sell -- and where we sell and support those infrastructure software. So that's still not that level -- and that does not what I call, co-mingle those 2 products at this point. And to the customers, we sell direct, particularly hyperscale, they literally are their own system integrators. They don't buy much software, which are more enterprise driven, they will tend to be the one to write a software, but they do buy our hardware directly. So that level of synergy that you're addressing does not quite exist as much as we would like it to at this point. At some point, one could believe that we could go that enterprise sales models we have in software today could extend to semiconductors, but not at this stage.

Got it. So what is the core value, Hock, that you think Broadcom bears or brings to the table in a software transaction? Is it your relationships with the top CIOs? Is the -- or is it just focused, right? And having that cost discipline that maybe the target did not have?

Well, in one consistent thing that does happen directly and direct, and it applies particularly in hyperscale these days, right? Is -- semiconductors is starting -- has started to become a matured industry, particularly the last 10 years. I've always said that, nothing new to you guys. What is also interesting is workloads continue to increase going to the cloud, close to 20% per year. The last data I've seen annually increase of workloads. And what they don't measure is the complexity of those workloads. And it's getting to a point where software workloads, applications is starting to overwhelm the ability of silicon technology to, as it has in generations before, being able to absorb and handle, it's called the end of Moore's Law, as you all hear. And does that mean technology stops evolving far from it. Well, people are very, very creative, innovative animals. So what we have started to see in a major way recently, is that in the past, driving down process technology in silicon enables you to drive up speed, performance of your transistor so that you can create CPUs, compute engines that handles all the software you can draw at it by moving down deeper and deeper submicron process technology. Not anymore. So what do a lot of hyperscale companies do to offset this function, this lack of ability to scale vertically downwards? They go sideways. They start creating silicon accelerators to handle specific function. We've seen that in AI machine learning workloads, the creation of GPUs, the creation, even one step further of TPUs from hyperscaler. We've seen it on orchestration. We've seen -- we're starting to see it even in congestion control of networking, where you're creating specific silicon accelerators for a very unique function. Effectively, for the hyperscale, which might extend eventually to enterprise, large enterprises, they're moving sideways, they're expanding the use of silicon to offset the fact that the compute engine, general purpose compute engine cannot handle this kind of complex and large workloads. And what does that imply? Well, it implies -- there is a kickoff -- there's kicking up a notch of demand for silicon. And it's early stage, but we can see that actually happening, and it's led by the hyperscale guys. And the growth rate we are seeing in our business, especially in the space we call networking. In another space, we use -- we call server storage connectivity, has accelerated as we reported very recently. And it's all due to this additional demand from horizontal creation of demand of hyperscale company.

Got it. Makes sense. And just one last thing, Hock, on the overall industry structure. So it seems the semiconductor industry -- Historically, units would grow 8%, 9%, right? But then we would have price declines of 3, 4 points. And the industry will grow 5%, 6%, right, at a time. Now it seems like that price reduction, right, has gone. And what I'm interested to know is that is that just in near-term cyclical dynamic? Or do you think the semiconductor industry is now in this new place where pricing can continue to be a very positive aspect of the industry?

I doubt it. With the amount of capacity we are seeing being planned in to by wafer foundries. We're going to see an excess of supply of wafers in back end and substrates in the next few years. Now it's not coming in probably until -- which is part of the reason why I'm projecting they won't be late '23, early '24 before we start seeing perhaps some balance or rebalancing of supply and demand, but it will come. And when it comes, we are back to the typical issue, semiconductor industry has -- that gives its bad rep sometimes, cyclicality, which is supply will start to exceed demand as inevitable. Because the demand, the supply that invests today comes in large chunks. You will exceed it, maybe not '23, probably '24. And then guess what, market forces takes over. And I hate to be a foundry at that time. It's going to be a foundry now. Be bad foundry at that time. But that's breaks in the industry, which means if they drop their prices to us, we'll drop our prices to our customers. But right now, to answer your question, you're probably going to come up, we have price increases in our supply chain. Given our products, we're able to pass it on to our customers as much as we can. And that's...

And you've seen that happening potentially in '23, but then in '24, there is the potential for a reverse to happen. But how about if I ask the foundries or even the equipment companies, that question, they will say, well, don't forget that there is a move to 3 nanometers or 2-nanometer, like you have all these advanced geometries where, yes, there might be a lot of CapEx, but the cost is going up so much that it will take a long time for the foundries to start cutting to have a glut of capacity?

Probably so. I mean this is starting to look beyond all the physics, beyond looking at physics and economics and costs altogether. But I tell you one nice thing about semiconductors that -- a little dirty secret and why you say our margins keep expanding, is the dirty little trick is -- in semiconductors, you keep coming out with new generation of products. That one makes it nice. That's what keeps out the leaders from the followers, new generation of products and technology. And that leads to newer process technology in silicon. And the dirty little secret is, every time we come and brief a new generation with more performance, more bandwidth, lower power, all that nice stuff, we'll raise our price to capture the value we create for our customers, which are who obviously want it or they won't buy from us. But fact is it's the same cost to produce, the same -- the next generation of product. It is. Non-R&D., R&D is still there, but the manufacturing cost is the same. That's the thing about semiconductors. So that's why your gross margin over time in this industry expands if you're mindful of what you can extract out of what you're offering in value.

Got it. Makes sense. And then specifically on your business. So you mentioned on the recent call that lead times are extended, but stable. But the semiconductor backlog grew to $29 billion from $25 billion before. That means you have more than a year's worth of backlog. What's your level of confidence, right, into this. Is this real? Is there any sense of double ordering or any abnormal situation in this backlog?

Oh, I'm pretty sure there's double ordering within that backlog. Of course, there is. But let me be -- let me pass through that statement. We deal directly in semiconductors focusing on that the customers -- we have the segments we deal with broadly for us. As I say, we're largely infrastructure. So there is hyperscale. Most exciting all you guys, and we deal directly with them. They buy our technology, our silicon hardware directly, whether it's in the form of merchant silicon, switching, routing, network interface cards or custom products that I talked about, which is fast expanding in those hyperscale. And by the way, I should add that the horizontal expansion of silicon by hyperscale as I said, for specific functions, thinking of it bigger picture, what that actually does, right? It is hardware silicon replacing finally, software. When you point so much software into a CPU that the CPU can't run, basically create a silicon accelerator. And you don't write that much software into your CPU and you say you make it happen in hardware. And that's really the bigger transformational thing. The hyperscalers are reducing software workloads to make it into hardware, just a point of interest. But coming back to this, direct. So we know back to demand. We know what they want, real demand largely. Telco, service provider, another big segment, again, we deal directly with the technology So we tend to have a good idea. And they are, they tend to be very multiyear programs. The area that is hardest to judge is enterprise. And enterprise represents almost 50% of our revenues -- of our semiconductor revenue. And as you know, enterprise are served by OEMs, large services -- system integrators, OEMs, multiple of them. So we support all those OEMs who are all chasing of the same customers in enterprise. So we will see double ordering from OEMs as we expect. So the level of backlog at any snapshot in time per se, it's not necessarily an indication what my revenue -- and what will translate into revenue over time eventually. Why I highlight backlog as a way to think about is if I keep by any time the same and the customers are all trained OEM, particularly to see that and the level of orders still keep coming in higher and higher is an indication. A very loud, clear indication they believe a year from now, 50 weeks lead time, demand will be there. If that stops increasing and it starts rolling over is a good indication, sentiment, purchasing sentiment has gone down. That's always good for at this point. But to say is that 52 weeks? Yes, it is because our backlog is noncancelable. But is that a true reflection of actual demand? No, it's not. Is it a reflection of 52 weeks of revenue somehow? Sure it is, because it's noncancelable backlog, but that doesn't mean it's true demand. All it does is perhaps sit in inventory eventually.

Got it. So what about the non-enterprise part of that backlog? Do you think that is a little more secure, that there's a lot more confidence and visibility into the...

From hyperscale and telcos, it's very, very strict -- very, very [indiscernible] to the extent because we not just believe them as giving us the right answers, we judge it. We -- or that's what we're doing these days. Most of my -- a lot of my operations and sales teams spend their time judging, second-guessing orders placed by our customers.

Got it. So within the enterprise, that has actually been one of the more interesting dynamics that the growth that the semiconductor industry and Broadcom are seeing that it is not just hyperscale-driven, that enterprise is being accretive to that. Do you think enterprise demand signals today are well above trend line? Or do you think that there really are new product cycles, whether it's multi-gig or whether it's WiFi or other things, that means that are driving higher than expected demand in enterprise. Is this sustainable, I guess, is the question?

Part of it is, I believe, part of it is not. I'll put another way. I think -- I mean, I'm open to the thoughts -- I've always told some of you guys whom I've seen before and who have listened to my earnings call historically, that semiconductor is a matured industry. Last 10 years, a 5% compounded growth rate through cycles, 5%. Next 10 years, why not 5%? I'm not so sure now, but if it's not 5%, maybe 8%, 9% but it's still not the 29% I'm seeing today. So yes, that's a big part of it, not sustainable, but it will be an increasing portion that is sustainable.

Is it visual thinking, Hock, to think that there is a higher probability the semiconductor industry can grow 8% or 9% over the next decade?

No, I don't think it's wishful thinking. I think it's very real. That's right. And simply because what I mentioned earlier, the phenomenon, especially among hyperscale who are getting, as I mentioned, growth in workloads, double digits and having to translate those workloads -- the software they would need to run those workloads more and more into hardware. I actually believe that the transformation is happening. And with that, it steps up incrementally the demand for semiconductors. So maybe 8% is possible.

Got it. So that's the interesting dynamic that you alluded to Hock, which is that, it's essentially these 3 segments, that one is that just my general purpose processor, right? The other is that I'm going to design internal silicon, but I would still rely and work with a Broadcom, right? Like a number of top hyperscalers do. And then there is a third segment, which is that I don't need traditional semiconductor companies. I can do all the design myself and just take it to a foundry and be done with it. How do you think that thing plays out? Do you see more companies being able to do more things internally or more companies wanting to rely on a partner with Broadcom?

That's -- those companies want to do it internally in my view, is wishful thinking. They cannot do it internally. It's -- the level of technology, the level of various different capabilities required for someone to design the complexity of some of these integrated circuits, as we call them, or system-on-a-chip, SoCs, is far beyond, I would say, what most, if not all companies have out there who are doing social media, you're doing devices, it's hot. It's very hot what the scale is doing. Now you may be able to do it for one generation, that's it. I mean after one, your CFO will look at the team and say, what are you trying to do here? Why do you have a car when you can get the milk cheaper. It's basically what it is. Because the amount of investment required, it's hard to keep sustaining. And that's why the world -- I always said all along, will move eventually, another phenomenon we see in semiconductors from ASICs which comes from the lines, historically, of vertically integrated system manufacturers. Whether it's HP, the roots of Broadcom; whether it's Siemens in Europe, whether it's General Electric, a long time ago, AT&T then, where eventually, those ASICs developed vertically by system manufacturers because they think that gives them an advantage will eventually disappear and give -- and it will be -- it has been -- will be has been replaced by merchant silicon, standard silicon. Because a standard silicon one covers whole market. And two, is -- it's hard to compete with a dedicated silicon supplier who is seeing across the entire market when trying to do it yourself, you're doing just in the very tunnel visioned manner. So merchant silicon is beating this ASIC. And that's very, very interesting, which is why when you don't think about the hyperscalers going into ASICs and creating this horizontal demand, you might say, isn't there a conflict here? I'd probably say, yes, that ASIC -- horizontal ASICs by hyperscalers might eventually give way to merchant silicon to in those specific areas.

Got it. Now there is an important consumer part of your business, Hock, right? With smartphones and...

One customer.

One customer, yes. What are the opportunities for further content growth in that market? Because when we look at -- it's easy to look at these phones and say, look, there are X number of 5G bands, they're all kind of supported, so where are really the incremental opportunities to expand content for Broadcom?

Well, you always have with this particular customer. And why -- this is our only consumer -- customer, the best way to describe it. And why it has remained that way is simple. This is a customer because of its business model, which is coming out with really leading-edge cool devices to the consuming public, that is the consuming public who is willing to pay for the high price of those devices, high-end devices. And because only one need to do is hooked on those devices and get on iOS. Nothing wrong in that. It's an ecosystem play. But what they do is because of that model, they value technology, they value engineering features in their -- any devices, whether it's a phone, iPad, iWatch, whatever. And that's right up our alley. We sell technology. We sell technology not easily found anywhere else. And so it's a perfect match. This is the kind of customer we want. They just happen to be huge, so what. And we sell them multiple products, not just one, beyond RF analog, which is what people tend to tie us with. FBAR filters and RF front end, we sell a lot more. We sell in -- and it's -- we sell at least 5 different product technologies into the whole portfolio of product devices Cupertino has. And we see them as very sustainable product technology franchise.

Got it. Is that -- so historically, you've had 3-, 4-year type of contracts, and there is one that I believe ends next year. How do you see the long-term future of that relationship that you continue to see areas of growth and cooperation there?

It is very strong. And it's again, all our engagements with them extends beyond 1 year. It's always been multiyear because -- which is the interesting part about why I say they're engineering driven. If you are engineering-driven where you're trying to create leading-edge technology into your device, you cannot think in terms of 1 year, right? It takes longer than between hardware, software, any iteration in between to go -- you have to go beyond 1 year. 2, 3, even 4 years is where our engagement tends to be, depending on the particular product. Given, as you say, there's a long-term contract, it's more papering over what is the true engagement model.

Got it. Now from a financial perspective, Hock. So Broadcom, 60% plus EBITDA margins, right? Which is not just among the most profitable in semiconductors, I think maybe in all of the market. What are additional levers for you to expand margins from here? And I think the other important part of that is, is there a trade-off with growth? So let's say, if I were to say that your EBITDA margins would stay at these levels, but you can grow 1.5x faster, would you take that?

It's not a trade-off in a fair area, but let me check that quickly. In semiconductors, it's not a trade-off. We're the market leader in each of the franchises we're in. There's no growth to be had other than a natural business growth in semiconductors, where each of our 17 product franchises, we are the market leader, not just a technology leader. In software, yes, there is. And that's the difference between hardware and software the way we manage it. And back to semiconductors first, and then I'll touch on software. In semiconductors, really, it isn't. It's about naturally -- I mean, to define our franchise, it's about that you're well positioned, you're the market leader, sometimes by a large amount. And you keep reinvesting for next generation. And where the growth comes from is other than the basic economic growth, more like low single digits to mid-single digits is the fact they always indicated you bring in a new generation of product, you get to raise prices because to extract more value. And that's what we do on a natural basis. In -- so it's not about trying to trade off growth, you get -- you capture what is a natural motion. Software, however, it's a different motion. I always want to -- I've done it for now 4 years, so I'm by no means an expert. But 4 years is enough to perhaps taught me. There is a trade-off between growth and profitability in selling infrastructure software. Most software products out there, stacks of it, pretty good, pretty useful. And people like to use it. The trick is monetization. And if you're out there selling software to as many customers as you can land it without regard to whether it's a profitable sale, you grow, you land and you find those customers. The problem is it's hard to make money. It's not just selling, the expense of selling. I tell you sell someone license and then -- license someone by subscription or something else, the software. You need to teach them how to use it. Then you need to do feature requests, then you need to support them. By the time you add up all these costs, any money you get out of any new logo is gone. And now if you want to make it a profitable business, you focus as we do on a core group of software customers, which has large footprint, and you basically then focus on giving them the best support, best service, and they have usually -- they are the biggest enterprises around and budget because it's mission critical to them to pay you for it. And then by just focusing on a small group of customers, you can make, as we do, as you see, a lot of money, but you don't show much growth, 5% a year. All we can try to grow at 30% a year and probably cut that 70% operating margin easily by hand.

Got it. And then, Hock, in the few minutes that we have. So with the software deal that you recently announced, right? The mix of the business will kind of be even semiconductors and software, is that by design? Is it just coincidence? And if you were to kind of fast forward 5, 6 years from now, do you think it -- is this the right mix for Broadcom?

So we're not literally designing for a mix of hardware and software. It is just as I mentioned many times, it's about the asset -- the quality of the asset and VMware, as we indicated, by all measures is a very high-quality asset with all the attributes we seek in an asset that we acquired. It's the leader in growing global markets on its platform, blue chip set of customers. And lastly, great bunch of engineers with a very innovation-centric culture. So very much like us, actually, in many ways, very similar. And so to us, it's a great asset and it was actionable. So we bought it. And now we're going to make something out of it. And that's when we talk about focus and execution.

Got it. And the debt on the balance sheet, at what point -- so I understand, right, versus EBITDA generation, right, it's extremely manageable, right? It's maybe in the lower kind of half of the industry. But just from an absolute dollar perspective, does it concern you ever to have so much debt on the balance sheet?

Not really because we measure it against our ability on a sustained basis, obviously, to generate earnings and cash flow. Given right now, before VMware, Broadcom has just below $39 billion of gross debt. This year, we're generating north of $20 billion of EBITDA. And I think we are solidly investment grade to handle that kind of debt. And now we buy VMware, we are prudent enough to put in a mix of stock and cash. So that cash requires additional debt, but we expect the incremental earnings that we will generate out of VMware will again easily carry that into incremental debt.

Got it. And before I close, somebody had asked me to ask this question of you, which is, is semiconductor industry consolidation over for Broadcom? Or do you think that there can still be opportunities for you to look at semiconductor in...

You've got 16 seconds. You asked me a very tricky question, I like that, isn't it? You know, to start with, we're not a consolidator. And that's the fact that I know we get stereotyped that. We do not do rollouts. I'm sure many of you realize that. I know many of the naysayers out there would say, oh, this is just a rollout, they cut cost. We don't. We do cut costs where they're not needed. We double down where they need it. We are not a consolidator. We're not a roll-up of the industry. We look at assets, particularly as I say, we're mining a deep pool of profit. How do you do that? You pick great assets and leverage on them. And best way to indicate is, since you asked that question, let the numbers talk for themselves. Over the -- since our -- we went public 2009, August to date, 2021, end of '21, through acquisition and consolidation -- not consolidate, acquisition and organic growth, we increased our revenue from 2009, 16x. If you look at our R&D spend, we increased it 25x. That's doubling down. We invest even more. But here's the zinger on the model, a compelling value to shareholders. We increased over that same period operating profit 84x. We're not a consolidator.

So zero SG&A model.

That's the hope eventually.

Perfect. Thank you so much. I really appreciate your time. Thank you.

Thank you.