Marvell Technology Group Ltd. (NASDAQ:MRVL) Deutsche Bank 2020 Virtual Technology Conference September 14, 2020 12:15 PM ET
Matt Murphy – CEO
Ashish Saran – VP, IR
Conference Call Participants
Ross Seymore – Deutsche Bank
Hi, everybody. Thanks for dialing in for the next presentation. I am Ross Seymore, the semiconductor analyst here at Deutsche Bank. We’re very pleased to have the management of Marvell here today CEO Matt Murphy, as well as the VP of Investor Relations, Ashish Saran.
So as of the other presentations, if you have any questions, feel free to email them to me either within the webcast system that you’re listening to this with, or you can just email me directly at [email protected] And I can ask them anonymously.
So with that being said, we’re going to just jump straight into the fireside chat with Matt. So, Matt, thank you very much for joining us.
Q – Ross Seymore
I think the high level question I want to start off with is, is largely about the transformation of Marvell itself. I know you have a rebranded logo, et cetera. Focus on the infrastructure side, data infrastructure, talk a little bit beyond just the logo obviously, about the focus of Marvell going forward and how big a change you’ve put into place over the last couple of years?
Yes great Ross, and thanks for the opportunity today. Yes, it has been a, pretty significant transformation process we’ve driven in the company. I joined in July of 2016 and early in my tenure, we sat down as a management team and really spent time on our mission statement and where we were going to kick the company. And at that time, what I wrote, which is still the same statement we have today is that we want to be the leading semiconductor company in the world that makes chips to move data, store data, process data and secure the world’s data.
We want to do that faster, more reliably and better than anybody else. And we did that, with the franchise at the time that was had a consumer component to it. And even at that point, we saw a little bit of mobile type of business in there.
And so, we drove this transformation Ross both organically as well as through M&A. Obviously, we added the Cavium, Avera, Aquantia and then we divested a number of product lines, the most notable of which was Wi-Fi.
And so, I had a lot on my plate when I first started and the one thing I did decide upfront was we weren’t going to do any external rebranding of the company until there was actually subsubstance. And so that was an intentional decision to wait and so the rebrand and the logo changed and all the look and feel was really emblematic of all the work we had done over the prior two or three years to get the company to where it was.
And now if you look at our consumer exposure is extremely small, and we’re now attached to some of the most exciting high growth opportunities in the world including 5G cloud and automotive.
And so, as I look forward with the platform that we put together with really the key components that you need to be an infrastructure powerhouse which is best-in-class processing capability which we have with our partnership with ARM, along with the expertise of the Cavium team for the last 20 years, leadership in networking technology both from the Marvell side in terms of the Ethernet switches and PHYs and also we bolstered that with Aquantia.
And then finally with the addition of Avera, we now have a best-in-class custom ASIC team, which can address very unique opportunities that we couldn’t service before in markets like cloud automotive, and also in 5G, there was a segment that we couldn’t quite address and standalone Marvell.
So, pretty exciting in terms of the journey that we’ve been on and certainly the brand was one part of it just to signal externally what our intentions were.
So you mentioned part of this transformation was M&A and you’ve done three deals of various sizes, but definitely is transformed the focus of the company to the infrastructure side of things? How is Marvell viewing M&A right now you’ve just done three deals, one divestiture and two acquisitions? Are you in more digestion mode right now or are there pieces of the puzzle that you still have to fill?
Yes I think, I would say we’re more in execution mode meaning we – I think very efficiently integrated both Aquantia and Avera on the IT front. I mean we were done, in the case of Aquantia our IT teams, the entire lift and shift of the ERP systems in one day. The Avera stuff was within a week and then the R&D side, we’ve now fully integrated those two teams. So that part’s done.
What I’ve seen this year Ross as we’ve narrowed our focus and by the way, I think the Wi-Fi divestiture which went very well. Again, it’s helped hone our focus into our core markets. Our design win funnel this year, Ross has never been larger. Our design activity, especially with the development and now announcement we’ve made of our 5-nanometer platform, has been extremely well received by customers.
And we’re very busy, right trying to execute our product roadmap and pipeline and continue to close, design-wins at a higher rate than we have previously and that all bodes well for the future. So, we like the assets we have, and we’re really in execution mode now which we’ve had to really focus on given the COVID-19 pandemic and shifting all of our engineers to home office environments, but that’s gone well.
And what I’d finally added at the highest level I think, it does seem like tech M&A has come back. Certainly, we had two big deals over the weekend. And I do think that bodes well for chip and broader tech M&A in general, but that – there seems to be an appetite to go do that. But for Marvell specifically, we’re really just executing this great platform that we’ve put together right now. And we’re busy doing that.
So we’ll get into the different revenue segments and product segments here in a bit, but since you mentioned about the M&A side that got announced today with a couple big deals, a lot of your business, the OCTEON side of things, the ThunderX side of things, a decent amount of what is going on in Cavium and even some portions I believe of classic Marvell are based on the ARM architecture. What are your thoughts on the proposed deal of NVIDIA and ARM? Does it have any implications for your roadmap or create any concerns or opportunities for Marvell?
Sure yes, so obviously, it seem the rumors and then saw the announcement today. And I think everything that’s being presented, certainly makes sense. As you point out, ARM is an integral partner for our company. And it spans pretty much all of our products, we use ARM-based microcontrollers and CPUs as embedded inside of many of our products that don’t even see the outside world and then all the way to the Thunder-based platforms, which is a server class CPU.
So long history there, I’d also say that we have a very good working relationship in history with NVIDIA. If you look at what some of the announcements we’ve done with them over the last few years with them, with the two companies’ together working on their GPUs plus our high end CPUs for high performance computing applications.
So we’ll have to see where all this goes in terms of what there – now their detailed plans are going to be but at a high level, I think having somebody like NVIDIA involved has the potential to accelerate the development of their high end part of their portfolio and their high end cores, which we certainly can utilize for our business and potentially there’s opportunities to work together.
But I’d say, we also have to – we’ll have to see what their exact plans are now that this is announced, but I think net-net it certainly could be a positive, but we’ll have to see.
And with the historical licensing agreements with ARM, how much IP exchange is there and how much concern could that create if you’re exchanging IP with ARM historically that’s one thing, but if it’s a potential competitor, is there something that would need to be more limited then it was in the past?
No I don’t think so, there’s not a lot of exchange per se from our side to theirs. I think we’ve had – we certainly licensed, we have an architecture license. We have teams that design their own CPUs based on that. We also licensed directly from them IP and cores, and we instantiate those in our products. And I think for this deal to be – to get through candidly and be successful in the market, I think NVIDIA will need to follow through on its commitment to keeping this as a separate entity and as a licensing ARM where the customer base of ARM is not disadvantaged in anyway.
And so that’s really what I mean by watching how this plays out. But no, I don’t sense any – I don’t have any concerns at this point about competitive issues. And I think if those were to arise, they would impact us and many other partners in the ecosystem, and I think that’s where challenges could occur from a regulatory approval standpoint.
So I would believe that their intent would be to do everything they can to obviously maximize the value of their investment they’re making, and also keep their partners and customers and the ecosystem happy so that ARM remains a vibrant architecture in the industry for people to use. And especially given how the landscape has changed in terms of customers wanting choice and ARM now, not only just at the low end and IoT applications but also you see it knock the ARM and even our high end processors, but really the market opening up on the infrastructure side. I think there’s just a big opportunity for companies like Marvell, who historically in partnership with ARM have been very successful and I think have a very bright future in that market. So we’ll have to see that Ross, only day one.
Get off that topic and go to one other aspect that you mentioned as far as kind of the new Marvell and the transformation of the company. And that is specifically on the manufacturing technology side. You have a big relationship now in partnership with the 5-nanometer node and probably beyond with TSMC. Talk about the importance of that relationship and where Marvell follows now and kind of the hierarchy of chip companies with its ability to address the bleeding edge of Moore’s Law?
Sure yes, I think it’s unique and important on a number of levels. And certainly, you could look at it and just say well everybody has access to nanometers. If you want to go get PHY, you just go get PHY and you develop a product and that’s it. And the reality is that – it’s very difficult to assemble the team, the intellectual property set, and be able to execute in a cost efficient manner the development of any of these advanced process nodes.
And historically, Marvell is a standalone company, always was a fast follower, typically would be one node behind, but differentiate and innovate on architecture, software, circuit design techniques, et cetera to gain the advantages back and it was very similar with Cavium, right. They could never sort of invest in the bleeding, bleeding edge, but they had this very amazing team of architects and developers to develop very, very competitive products and, compete head-to-head with much larger scale companies and even Avera, has been really on IBM/ GlobalFoundries process platform which was, has not been in a leadership position for some time.
So the unique part Ross is that we’ve now got the scale and the leverage to invest and bring all three companies into the bleeding edge. And that’s important because we’ve achieved a fair amount of success to-date with this prior strategy. So we’re going to keep the architectural innovation going, whether it’s our chiplet-based approach or whether it’s uniqueness in the microprocessor architectures we have or the features that we add to our products, let’s say in our networking area, but they’re going to be now in the best-in-class technology.
And so, when I look to the future and we see this and the design win momentum that we’re seeing this, having this platform in 5-nanometers is very unique. And I think it could give us an over-time a step function change in our relevance in the markets that we play in.
So it’s a pretty big deal and it’s not something that everybody can do. And it’s not something that we decided overnight either. We’ve been at this for almost two years in terms of developing the IP set and the platform and now we expect this year to get our first silicon back from a sample’s perspective from the [shuttle]. And then actually sample real products at the end of next year in 5-nanometer, which I think is a big deal in our industry.
And have you guys said or are you willing to say timing wise, you just mentioned kind of in the next year sampling to customers. What will be the general product area you’re going to target first with this leading edge?
Well, the nice thing about it, Ross is that all of our various product lines are now mapping into this 5-nanometer roadmap. So you would – the obvious one that you would think of would be our processor platform based on OCTEON. That’s certainly underway, both for baseband processors as well as the embedded processors. We’re working at our 10th generation OCTEON processor now, but also our entire ASIC roadmap, we’ve been marketing to customers since the fall of last year.
And targeting designs and that’s all in 5-nanometer. You should assume our networking products, right are making that jump. Storage will probably be slightly behind just because dyes are typically smaller and don’t have the needs, but even they have their own roadmap. So Ross, it’s going to be very broad in nature. And the way to think of it is we’ve in some cases, jumped all the way over seven and gone directly into five and the whole company roadmaps are converging.
So we plan on getting a lot of efficiency out of this development, and not having to support a disparate wide array of various process technologies, but be very focused. And then we get the IP reuse and the leverage by having all these various product lines, dovetail into PHY. So it’s not just a one shot deal. There are multiple products, the first of which we’ll be sampling at the end of next year by 2022/2023 it’s going to be very robust in terms of our product roadmaps and also product ramps.
Got it. Thanks for the color on that. So while we pivot over to the revenue segments that you talk about on your call, within networking obviously, 5G is the end market that people are most excited I would say for your company. Talk a little bit about where we are today, I know you have an analyst meeting coming up early next month. So I don’t want you to obviously front run and steal your own Thunder too much on that. But talk about the breadth of customer base, the geographic dispersion of that business, just any sort of metrics that we should use externally to monitor the progress that Marvell is making in that segment of your company?
Sure yes, we do have the Analyst Day. I hope all of you can tune in and we’ll certainly do a much deeper dive than what I’m going to cover today. But yes, I think the 5G opportunity for us is quite unique and has come a long way. If you go back to our last Investor Day, which was in October of 2018, we had closed Cavium and we had come out and shown what our combined company strategy was.
And part of that was, so I think was eye opening at the time was that we had a significant. We highlighted a significant gain we were seeing and content in 5G base stations, at least with our lead customer, which was Samsung at the time. And since that time, we’ve now had public announcements with companies like Nokia to be a much more important part of their portfolio on multiple products, not just the baseband.
We did it in acquisition, which was Avera, which came with two additional 5G customers that weren’t Huawei. So basically, we have now content and share it for the top five base station OEMs. We also highlighted on our last call that we’ve been winning additional designs at companies that are outside the top five, which in aggregate will actually make up a fairly reasonable volume over time.
And so, our platform is clearly shown a lot of success. And so, I think the first point is that it’s broad from a customer standpoint. The second is that it’s broad from a product standpoint. We have chips all the way from baseband processors to the transport processing sockets. We have content now in the Radiohead, for massive MIMO and beamforming. And then we have, , custom ASIC capability where we’ve got a number of wins in the digital front end area.
So, I think that’s the second part is multitude of customers multitude of products. And then all those customers we have are selling into pretty much every geography around the world. We’ve seen China ramping up this year. There has been a lot of positive announcements about U.S. 5G for next year. Korea was an early adopter in 2019. So, as these geographies roll out over time and our customers gain share in terms of the spent on 5G versus 4G equipment.
We see a multiyear opportunity for the company as our new wins layer in, our new customers layer in and then geographies ramp up with significantly higher content or exposure than we ever had as a company as Marvell in 4G.
So is there a way to summarize that content increase from the 4G to the 5G side of things? I know you did at your last Analyst Meeting, if I recall right, it was kind of a 4x increase. But given the breadth of the customer base you have now is there any simple summary to that number that you could update us on?
Well, I think the way to think about it Ross is we’ve said that the SAM for our products, is around $6 billion. And the rough math to get you there is call it 1.5 million base stations, with $4,000 of opportunity for Marvell in each of those, that’s the rough math. Now, you could argue that actually there might be a lot more 5G base station shift and especially what happens if there’s a lot more massive MIMO radioheads, and there’s a lot of other scenarios you can get into.
But just to keep it simple, let’s call it 6 billion I mean, could be bigger, who knows. And historically, about 2 billion of that we really haven’t been able to go after because that was serviced by Huawei and high silicon themselves. That is one X factor right now right, given all the government restrictions and some of the issues going on geopolitically with them, there’s certainly an opportunity for some of that SAM to actually open back up for us through our existing customer channel.
So that’s a call option or an opportunity that we haven’t quite sized. But basically yes, if you look at each of these base stations, there’s about $4,000 worth of content. And we’ve shown that at our lead customer, if we can kind of win the full suite we can approach that number. But you should assume with all of our key customers, we’re in deep discussions about what we can offer. And certainly, there’s a range of offerings we have today, but as we penetrated into places like the radioheads, we’ve got more and more content there.
Showcasing of where is ASIC capability, they were really limited primarily to one customer before and now we’re able to showcase that. So, we certainly are driving a higher target content Ross per base station. But then some of that this also comes down to what are the configurations that we win, what share does our customers get in fact. But we still think the $4,000 bogey is kind of a dream, the dream number depending for each individual, it’s kind of a content goal.
And then that’s up pretty significantly. And if you look at the content we had, in prior designs at least at our lead customer, it was around call it $700/$800. So, content opportunity is up share shift is a very distinct possibility. And certainly the design when traction and momentum that we’re seeing indicates that there’s more opportunity for us.
And the fungibility of that share that the Huawei high silicon side of things goes obviously I’m not sure you’re going to have any greater color on what the government restrictions may or may not look like. But how fungible do you think that is to go to other vendors? Are you seeing your traction of activity and design wins ramp in places that you otherwise would have historically assigned to being Huawei strongholds or is that more still in kind of a discussion phase and theoretical than it is practical?
Yes, I call it more theoretical at this point. And I think the best companies to really give you the beat on that ultimately would be their direct competitors. And I think in general, everybody’s been very reticent to be able to size what that impact would be or even call it out. Let’s remember, it’s a very resilient company. I’ve worked with them for a very long time, and they’ve always sort of found a way to make it happen.
But it does appear to be fairly serious at this point in terms of some of the actions that are taken. So hard to know Ross, none of our short-term outlook is really driven by that. But certainly, it’s clear to us now that there’s got to be some impact from all of this. And is that 2%, share shift to 5% share shift, to 10% share shift? I don’t think we know, but given the SAM numbers and content numbers I gave you, you can start to imagine if there was a meaningful move there over time, what’s called the next couple years as share moved.
Then it would certainly be a benefit to us, but that’s currently not in any of our internal plans or our internal models. That’s really what I would call it an upside case when we look at it and plan our company.
Got it. Let’s pivot a little bit to another metric that you’ve given in the last couple earnings calls and that was your – I think you’d refer to it as your cloud data center business exceeding 10% of your revenue. I know that doesn’t overlap perfectly with your networking versus storage side and kind of straddles those two definitions. But generally speaking, talk about how you’re attacking that data center business, why you decided to break it out? And what are your growth aspirations in that if there’s any sort of quantification you can provide?
Sure yes no, I think we thought it was important to break out for a few reasons. One is, as I had mentioned earlier, when we developed our company strategy and mission statement, clearly the pivot to the cloud was a key part of that. Because that’s where, that’s where the growth we saw a lot of growth opportunity across a number of our product line. And so, kind of quietly behind the scenes, we’ve been driving this pivot to the cloud and that’s across the Marvell storage franchise.
If I look back four years ago, bulk of our storage business was in HDD. And it was in notebooks, and SSD was very small. Now if you look at our storage portfolio, we’ve reduced that consumer type exposure on both the HDD and SSD side. We’ve grown SSD meaningfully over the last four years and the nearline focus portion of the HDD business. We’ve actually won multiple accounts now with both SoCs and preamp.
So all of a sudden, and then we see SSD as well going in various forms into the data center as well. So Marvell storage has made that pivot. We acquired Cavium, during that timeframe. They also had exposure to the cloud both with their liquid security, as well as SmartNIC offerings, which is a product called LiquidIO. And then we acquired Avera and they also came with revenue. And then design wins as well in hyperscale.
And so you start adding it all up. And it was a fairly small number over the last few years and two quarters ago, it exceeded 10% of revenue. And as we looked out, we believe that that would continue and only continue to grow as a percentage of our total. And so that was the case in Q1, it was also the case in Q2. And to your point Ross, we typically only talk about networking and storage.
So it’s sort of markets on our quarterly results. We’ll give some more view at the Investor Day in terms of the end markets that we sell into. And a little bit more clarity, but yes the cloud one is exciting because it’s not a one trick pony, not like we just got one win on one product and it shipping into one hyperscaler and it’s going to come and go. We’re in a little bit like the 5G commentary its multiple hyperscalers, multiple products and technologies.
And we’re now able to service that market in a number of ways. And we do believe it’s going to continue to grow and be a much more important part of our company revenue and composition going forward.
Another aspect that you mentioned within that cloud side of things was the ARM server CPU ThunderX. Without going into the NVIDIA/ARM deal, you guys before that also changed the business model a little bit made it more custom rather than standard silicon? How are we to take that away as I get that the customers want it to be customized but I also believe that the standard silicon market was always the goal there? And so, I’ve heard some rather cynical folks say this is just one step further down the road of exiting that business. So talk a little bit about that evolution and what does it mean, going forward to your Thunder business?
Sure, yes I’m happy to do that. So, let’s take it back to even when we did the diligence on Cavium and we acquired the company. There was a question back then, should we keep it or should we not from investors. My dealings with the key customers at the time was very much in support of continuing that roadmap and the strategy. When we bought it was that – the business was being run like a standard processor business which is you have a roadmap based on process node and architecture improvement.
It was sort of like just map of A and B, you map Intel, map of ARM and figure out where people are going to be drive the ecosystem, get boards built in Taiwan, drive the whole thing as a normal processor business with at that time, no funding from anyone else just funding the dream on our own. And I think pretty quickly, within the first year or so we sort of said look, this is a pretty significant investment and to do it right, we’re going to need to get some additional funding.
And so, we did do that. We secured that from some of the customer base as well as from ARM themselves. And have continued that. What has become apparent though, in the last year or so is that the bulk of that market is really the highest growth opportunity and the biggest SAM for us is really in the hyperscale area. Enterprise is going to take a long time, high performance computing, I think is a bit lumpy and probably – and just not the same scale opportunity as hyperscale.
And as we’ve engaged deeply with those customers I’ve kind of felt like the same story over and over again what’s happening, which is hey, we love your ThunderX2, we love your ThunderX3 and you’re roadmap. But we actually want something a little bit different. Can you add these other features, can you add more functionality that we need specifically for our case, and each of those is a new chip, you got to go spin it.
And at these advanced nodes, it’s very expensive to do that, you need customers to really commit, you can’t just make a unique core standard product for every large hyperscaler. And ultimately, the discussion is really turned to well if that’s really what you want, and this looks like more like a customer effort.
At the same time, we acquired Avera which actually gave us a very high functioning and working business model where we can size at our ease. We can get customers to put skin in the game. And I think our view fundamentally was, given how strategic and important this is, we need to go about – above this market in a different way. And I didn’t feel like it made sense to keep going as a merchant offering because quite frankly, there is very little market for a one size fits all park. And so it’s led us to, I think a pretty logical conclusion that we – if we’re going to do these high end complex chips with companies with trillion dollar market caps that we should both be in this together.
And we shouldn’t be funding 100% of it ourselves. And certainly we’re not in a position to be spinning out, new tape outs and mass sets for every individual customer on our own dime. And so that’s really, that’s really where the pivot came from. And in doing that, it also has saved us some OpEx because we’re able to now do that in a much more efficient manner given that we’re not trying to market this product and sell it, and create an ecosystem to support 10s or dozens of customers. It’s really for a handful of targeted accounts where we expect to drive the opportunity from.
Thanks for that color there, we have I think one minute left.
Yes, yes, it’s a market driven decision. At some point this is where the markets going Ross and so we can’t just anyway. So go ahead sorry you had one more question.
Yes sure, just to kind of wrap it up bringing it back to a high level. It seems to me that the data center side and the 5G side are great businesses, they’re growing fast. The storage side kind of has it fits and starts. If I had thought about the company two, three years down the road versus how you have networking, it kind of 55%/60% in the storage side being the remainder of the company – a little bit of other. Two, three years down the road, how do you see the revenue mix of this company looking and what does that mean to the operating margin and profitability?
Sure well, I think there’s two ways to cut it. I think just to make it simple, I think on the market side which, as you mentioned, we very much expect cloud and the 5G business to be a much more significant portion of the company total as we go forward. Both of those have got the design wins to back it up, they’re already growing it at a much faster rate than the overall company. And by the way on top of that, we have this kicker, which is now starting to manifest itself and will continue, which is an automotive.
So we see those three end markets driving the bulk of the company growth. Enterprise we believe still will grow. That’s sort of what you would call the remainder excluding this other piece, which has got printers and some legacy stuff in it. We’ve always said that that will decline over time. But the enterprise business while it’s a little bit soft right now, we do see from the traction there.
So but I would think about the first three I mentioned is the highest growth and if you flip that back around then I said what does that mean to our traditional carve out – breakout of networking and storage. Yes, you’re going to continue to see networking grow over time as a percentage of the total, because most of the content for the cloud and the 5G and the automotive business are from the networking type of products.
Although on the cloud side there is some contribution from airlines. So I think over time, that’s what you’ll see storage being more stable. Obviously, it’s a higher beta business. But if you just normalize it over time we believe that that business from here on out can grow a little bit. And then networking is really where you’ll see the higher growth and we’ll update the financial model coming up here at the Investor Day and what does that mean for growth and gross margins? So maybe I’ll save the suspense for the – till the Investor Day on the company financial model as a result of all this.
Got it. Well, thank you so much for all those details, Matt, and Ashish thanks for joining as well. We’re at the end of our time. So everyone this ends the Marvell fireside chat.