|
|
|
|
||
11/29https://www.marketscreener.com/amp/quote/stock/RAMBUS-INC-10681/news/Transcript-Rambus-Inc-Presents-at-Cred it-Suisse-26th-Annual-Technology-Conference-Nov-29-2022-02-42435967/ Transcript : Rambus Inc. Presents at Credit Suisse 26th Annual Technology Conference, Nov-29-2022 02:20 PM 11/29/2022 | 04:20pm Presenter Speech Christopher Caso (Analysts) All right, looks like we're just in time. So welcome back, everyone. I'm Chris Caso, semiconductor analyst here at Credit Suisse. With us for our next presentation, pleased to present Rambus, with us from Rambus is Luc Seraphin, President and CEO and Des Lynch from Investor Relations. Gentlemen, thanks both of you for attending. Thanks everyone for joining. Question Christopher Caso (Analysts) Luc, maybe I'll open it to you to start and Rambus is a company that's changed a lot over the years and strategy has certainly changed. Maybe I'll leave it open to you to speak about that the change in strategy that's happened over the past couple of years and why you've pursued that change? Answer Luc Seraphin (Executives) Yes, good afternoon, everyone. So I'd encourage everyone to read our Safe-Harbor and read our filings where you can find good information. Yes, you're right. We started about 30 years ago by pioneering the foundational technology used in memory interfaces. And the way we started the business was through a pure patent licensing and that's how we were known about 20 years ago. Overtime, we've transitioned the business from pure patent licensing to developing technologies and developing products for the data center. And today, we go-to-market in 3 different ways. We still have our patent licensing business and for this business we deal about $200 million to $220 million a year, that's a very stable business, very long-term visibility over the business. For those who followed us, we've signed a 10-year renewal agreement with Samsung, because it gives us visibility and cash flows for the long-run. So that's the first part we go-to-market. The second way we go-to-market is through what we call silicon IP. We actually develop pieces of specialized IP for high-speed interfaces, controllers and security that we sell to semiconductor companies. This business is about $120 million to $130 million. It grew quite nicely last year 25%, this year 25% over last year. We expect to continue to grow but not at the same rate given the current economic environment and the slowdown the funding in small companies. And the third leg of our business is semiconductor products and here we actually sell chips to the 3 memory vendors building DRAM modules, that business has grown quite nicely over the last few years. It was $78 million in 2018, this year if we take the midpoint of our guidance for Q4 is going to be $225 million. So we grew that business close to 60% over last year, when the market grew only 20%, 25%. So that strategy has paid-off for us. What we've done is, we've refocused the company on to data centers. We've eliminated the activities that were not central to that goal and the money we actually collected from our patent licensing business was re-injected in developing products that matter for the same customers. Question Christopher Caso (Analysts) Right. Answer Luc Seraphin (Executives) So we've also changed the relationship we have with the ecosystem and we become a force in the datacenter. Question Christopher Caso (Analysts) Right. So maybe I'll start with the licensing business and you spoke that now you have long-term agreements in that space. Is that really just a source of cash now at this point and degree of -- is there any variability to the licensing revenue stream because of macro conditions because of what's going on with memory right now, how stable is that revenue stream? Answer Luc Seraphin (Executives) That revenue stream of $200 million to $220 million a year is very stable. Typically, the license agreements are between 5 to 10 years. So once we sign a renewal, that revenue stream is stable for this long period of time. So we don't expect to see any fluctuations from that part of the business. We've continued to invest in our patent portfolio, so that our patent portfolio remains relevant when you come to renewals and that explains why we've been able to maintain that stable cash inflow into the company, right? Question Christopher Caso (Analysts) And when are the renewals up, are there any renewals up in the next several years? Answer Luc Seraphin (Executives) So we are in constant renewal mode, if you wish, we're not too public about it except in the case of Samsung recently for obvious reasons. But typically, we go through 2 to 3 renewals per year and this has been the case over the last few years, we just don't make them public, just [ pickers ] our licensees, don't like it to be public, sometimes you don't like it to be public. Question Christopher Caso (Analysts) Right. So accepting that, that licensing business is stable, you're taking that cash and you're putting that into R&D for the product business then. Maybe talk a little bit about the product business and some of the cycles that you're undergoing there, DDR5 is very important to you, CXL is very important. So there's certain things we'll wish to discuss. Answer Luc Seraphin (Executives) Yes. So we actually develop products that sits between processors in datacenters and memory. And there has been several generations of memory, which we call DDR3, DDR4, DDR5 and these generations last for many years. The DDR4 generation, that lasted for the last 7 years, for example. And within each one of these generations of memory, you have sub generations, their speed grades, they're going faster and faster. So we entered that market by building what we call these buffer shape which interfaces between the processor from Intel and AMD to the memory modules from Micron, SK hynix and Samsung. And through the DDR4 generations, we've grown our share from 0% about 5 years ago to mid-20% this year. And we've done that by qualifying our products early, making sure that we have products of good quality, we have good service and security of supply to our customers. Now the market is moving to DDR5. This is really, really important because when you talk to our customers, customers, the cloud service providers, they are hungry for more memory, they need more capacity and more bandwidth. And DDR5 offers that as a new generation of memory. So we have developed buffer chips, the same thing chips that sit between the processor and the memory. For DDR5, we started those developments in 2018, so we're really first-to-market and by being first-to-market our qualification footprint is been very, very large. So when the market transition to DDR5 in 2023, we see this translating into a bump into our share on that -- on that part of the business. You mentioned CXL, CXL is really exciting. In the datacenter, you have processors and memory, but people always want more memory. So if you saturate the memory bus and you want to add memory you have no other ways in adding another processor, which is not economically viable. So what the industry has thought about is, well, can I actually add more memory when my memory bus is saturated without adding a new processor and the answer is yes, you can if you attach more memory to the CXL bus. And that creates the opportunity for CXL memory modules for which we're going to produce CXL products. So that's an additional TAM for us and they're also use cases for CXL, cloud service providers, our concern with the fact that a lot of the memory they have in their datacenter is trended, meaning that is not being used. CXL allows you to pool those memory from different memory banks and share those memory with different processors that use memory CXL pulling devices for that and there are other use cases that CXL is going to bring to market. So the buffer chip market for us today, the market we are in is about $850 million in size. Today we expect this market to grow to $1.3 billion to $1.4 billion by 2025. The CXL market is going to add another $800 million to that number by 2025. So we do see just in the datacenter to manage the interface between processors and memory, a pretty large TAM expansion that we want to capitalize on. Question Christopher Caso (Analysts) Right. And DDR5 transition has been delayed because of the delay in Intel and AMD processor. So maybe first you could speak to what impact that's had to DDR4 market because also had an impact to DDR4 and then on your business it obviously delayed the transition to the DDR5 product that you have. Answer Luc Seraphin (Executives) Yes. So this has been a very interesting dynamic this year. Question Christopher Caso (Analysts) Yes. Answer Luc Seraphin (Executives) We had seen as you said the platform delays which delayed the introduction of DDR5 combined ways, the supply constraints that we all had. So what we saw happening is that people were building DDR5 modules a bit too early because of the delay and have built some inventory, but not too much, but they have built some inventory expecting DDR5 to ramp earlier. But at the same time, they didn't plan for enough DDR4 when DDR4 is lasting longer and they didn't plan for enough DDR4 in the market there were supply constraint. So we had very different dynamics between DDR4 and DDR5, in DDR5 there was a slight increase of inventory because of the delays. On DDR4, we were supply constrained and we're still supply constrained and I think it's going to change, but Q3 for us we were supply constrained, we shipped 58.6 million, but we could have shipped more if we had the supply. Q4 is going to be about 66 million at the midpoint, it could be higher if we had the supply is mostly DDR4. But the supply situation is improving now. I think the market is going to use their backlog on DDR4, there's going to be a pause in DDR5 and that's why we're saying that is going to be lumpy in the first 2 quarters of the year. But fundamentally, you know the market is going to grow for us and our revenue is going to grow next year compared to this year. We just need to manage that transition. Question Christopher Caso (Analysts) Yes. Because at this point now, you are closer to the DDR5 transition imagine... Answer Luc Seraphin (Executives) Right. Question Christopher Caso (Analysts) Customers don't want to build too much inventory of DDR4 now particularly as the market soften up a little bit and you still have to wait for DDR5 to kick-in? Answer Luc Seraphin (Executives) Right. And we expect DDR5 to ramp in volume in the second half of 2023. We see the crossover point in units between DDR4 and DDR5 to happen the end of the first half of '24 in volumes. So, DDR4 is going to continue to last. It's always been the case in every generation, is going to be a long-tail while the new generation ramps up and we just have to manage those inventories to do and backlog doing the transition in the first half of '23. Question Christopher Caso (Analysts) What's your expectation with regard to pricing as that transition happens, because in the past, you've seen some pricing erosion on the older generation as the new generation comes up? Answer Luc Seraphin (Executives) So we've seen a pretty long pricing erosion on the DDR4 generations because DDR4 lasted for 7 years now, so it's pretty why they're. We typically see a price reset when there's a new generation of memory, so we'll see this with DDR5. So we're going to see a price reset which is going to help improve the time for us. The other thing that's going to happen in DDR5 is that it's been defined slightly differently than DDR4 and some of the chips that used to be used on the motherboard in DDR4 environment have moved to these memory module in the DDR5. So there is a potential for more dollar content on the DDR5 module because in addition to the buffer chip we have what we call companion chips and we're developing those companion chips as well as part of our product offering so board to the... Question Christopher Caso (Analysts) What do you estimate the incremental content to be to Rambus? Answer Luc Seraphin (Executives) Typically it's going to be around couple of dollars per module for these companion chip, they're small chips, but that counts percentage wise what just 10% or so something like that. Yes, it's difficult to say, especially at the beginning of that transition when the prices are bit distorted. Question Christopher Caso (Analysts) Sure. Answer Luc Seraphin (Executives) And also we tend to not talk too much pricing when there's 3 customers and 3 competitors in the market, but it clears offset by as DDR4 kind of ramps down to that turn getting off. There's going to be a better pricing environment for DDR5 content environment. Question Christopher Caso (Analysts) Sure. What about with respect to CXL and my understanding is the first generation of CXL, you're not participating as broadly in that one and as I guess CXL 2.0 is, if memory serves correctly is where you're planning on participating more broadly? Answer Luc Seraphin (Executives) So we are approaching the CXL market in 2 waves. Actually, the first wave of CXL market is not through semiconductor products. But remember, we have patent licensing, silicon IP and semiconductor products. In the silicon IP space, we have CXL IP that we are selling today to any semiconductor companies that want to develop CXL based products. So we actually generate revenue today from CXL, but through silicon IP sales in first generation of CXL products. Our own semiconductor products the third leg of our business, for our own semiconductor products, we are developing CXL products that are going to be CXL 2.0 compatible and that will hit the market in 2024. Question Christopher Caso (Analysts) Right. Answer Luc Seraphin (Executives) So that first wave this is silicon IP sales this year, next year 2024, we're going to move into product sales as well. Question Christopher Caso (Analysts) Right. And what's the content and perhaps the CXL space adding so much to the TAM? What's the competitive environment? Is it a new set of competitors that you'll be dealing with when CXL heads? Answer Luc Seraphin (Executives) Yes. CXL is a very exciting market, buffer chip, [ Zedek ] defines what the chip should look like. So we all develop the same chip and we compete. CXL is defined as an interface, not as a chip. So anyone who wants to communicate -- every component that needs to communicate to another component in a data center needs to have a CXL interface. So the number of products that you can develop based on CXL is much larger than what you have in the buffer chip for example because CXL can connect 2 processors, can connect one processor to a GPU, we can connect processors to memory and all types of memories. We see 4 main categories of CXL products. The first category is memory expander. So basically, when you saturate your memory bus with DDR and you want more memory, you add a CXL module. So that's the CXL expander. That's the first category of products and that's the first products that we're going to develop because that's an ecosystem, we understand the same customers, same partners. We know the memory side of things and the use case is very simple. We just need more memory on the same processor. The software load is easier to manage because it's one processor one module. That's the first category. The second category of CXL products is what we call CXL pooling products and the idea here is, if you had several processors and several banks of memories, how can you ensure that each processor can access any type in any one of the memories, even if it's not your own memory. And that addresses the question of stranded memory in data centers. So that's the second type of products, memory pooling products and we're going to also develop these type of products. The 2 other categories are switching devices, we look at that later. We have enough on our plate with the first 2 categories and retimers. Now you're talking about competitors there, we're pleased to see competitors coming to that market. We started about 18 months ago on our side. We've built a team with both IP by buying companies and we're well ahead in our development for the product launch in '24. We see companies like [ Microchip ] playing in that space, stemming from past acquisitions that they had with PMC in particular. We see companies -- we see private companies like Astera Labs, they're very open about this developing retimers, that category of products and Marvell is going to play in that market. They're coming from the standpoint of -- they're an SoC company. If they want to play in the data center, they need to have CXL capabilities, so they need to build that capability internally. So we're all coming from different angles. But I think there's room for everyone there. There's so many categories of products and within each category, there's so many potential products. On our side, as I said, because of the inroads we have in the data center market for that memory interface, we're going to start with memory expansion devices for CXL. Question Christopher Caso (Analysts) Right. And as CXL takes hold, it doesn't have a negative effect on the existing DDR4, DDR5 environment, that buffer chip still becomes required for direct DDR5 system, CXL is expanding that market for you. Answer Luc Seraphin (Executives) Yes, we don't -- yes, that's a very good question. The question is, is it going to cannibalize the current business? And it's not actually -- what's driving CXL is that people are asking for more memory. The limitation is the process of technology has evolved faster than the memory technology. So the processing capability has grown faster than what the memory can support. So people are asking for more memory. And once they saturate the standard memory bus, they have to add memory elsewhere. So it's additive if you wish to that. So we see this as being completely additive. Question Christopher Caso (Analysts) Yes, right. And in terms of content opportunity, I think you said it was about another $800 million on top of your existing TAM, is -- what products -- because there is 4 product categories within there? What product categories does that include? Answer Luc Seraphin (Executives) Well, the first thing I would say is that this is our estimation today, the market is starting, so these numbers fit as we go. But when we look at this, this is mostly for the first 2 categories of products. Again, it's a bit like DDR5, this market is going to be based on the introduction of CXL 2.0. And CXL 2.0 is going to be dependent on the processing platforms being ready on time for that, it's just like DDR5. If the DDR5 platforms or supporting platform are late -- you're late for the corresponding modules, the same is going to happen for same itself. Question Christopher Caso (Analysts) And mind which processing platforms are referring to the... Answer Luc Seraphin (Executives) So CXL 2.0 is going to be introduced in the follow-on platforms from both Intel and AMD. So they're introducing Sapphire Rapid and Genoa now. The next one is going to be supporting CXL 2.0. Question Christopher Caso (Analysts) Right. Okay. Maybe take a pause here if there's any questions from the audience. People have been generally shy today except for the Credit Suisse people in the audience, but we'll just continue. What are the IP portfolio outside licensing, maybe you could speak to that a bit. And you referred to, there were some impact on that on probably some of the smaller companies that would license from you obviously impact a little bit worse in the macro environment. Maybe you could expand upon that a little bit? Answer Luc Seraphin (Executives) Yes, so our silicon IP business, as we said, is going to be between $120 million to $130 million this year, which is to size it. We are very focused in our offering. So because of our heritage, we offer high-speed memory interfaces and high-speed serial interfaces and the associated controllers. And we are -- we've built that business through organic growth and acquisitions. The acquisitions that we had last year were PLDA, who developed CXL and PCIe controllers. AnalogX who develops CXL 3.05 technology. And we also bought a company this year that helps us accelerate our product development. It's more an acquisition of a team that can do SoC. So high-speed memory interface like GDDR, HBM type of technologies, high-speed serial interface, like CXL 2.0, CXL 3.0, so this is our IP portfolio. And then we also have security IP and security is becoming pervasive in every semiconductor device, especially in the current environment a lot of customers are asking us to make sure that they can identify every piece of silicon that they use. They can trace every piece of silicon that they use, that the data moving between different parts of the system is secured. So that creates a lot of demand for security IP, but security IP goes into broader markets than simply data centers. It goes into IoT, 5G, automotive, defense type of application. So very narrow, but very specialized IP portfolio that is in high demand that's what we need. With respect to the environment, this is a different business than the buffer chip, buffer chip, there are 3 customers in the world, they're large customers. In the IP business, the business model is a license. So we license the IP for use of that IP. So we have a much broader set of customers. The sales cycles are much shorter, but a lot of these companies can be start-up companies and because of funding is starting to slow down in start-up companies, what we've seen is that there are not so many design starts as they used to be. So that's one of the dynamics that we're watching as we enter 2023. The other dynamic that we're watching, although, it impacts only a small part of our business, excuse me, is what's happening in China, whether we have restrictions in selling that IP to China or whether even in China with growth rates going down, there's a slowdown of design starts as well. So as much as our IP business grew about 25% this year over last year, we expect the growth next year to be more in the 5%, 10% range, 5% range. In the long run, this is a business that has the potential to grow 10% to 15%, but we expect that because of the macro environment and the geopolitical environment that the growth is not going to be that high in 2023. Question Christopher Caso (Analysts) Right. And the licensing of that business, is it a function of upfront licensing plus volume? Answer Luc Seraphin (Executives) Most of the business is upfront licensing and reuse licensing. We have almost -- yes, none of the business based on royalties. Question Christopher Caso (Analysts) Okay. Answer Luc Seraphin (Executives) So we don't have to wait for products to ramp to get revenue, we get the revenue as soon as we actually deliver the IP for companies that develop those products. Question Christopher Caso (Analysts) Right. So it kind of leaves in -- as we're wrapping up here, in terms of the near-term market conditions and how that affects your business. And I think we just said before on the licensing business to the memory suppliers, it doesn't really matter, that's a fairly stable one. With regard to the product business, certainly, the biggest effect is transition to DDR4, DDR5 and CXL. What effect has the slower market conditions in memory had on that business? Answer Luc Seraphin (Executives) Very little because I think the -- what we see in the market is that some markets have softened much faster than the data centers. When you look -- when you talk to memory companies, I think they suffered from the mobile phone market, the PC market, consumer markets, we're not in those markets, we're only in data centers. And there's pent-up demand for, I would say, capacity and bandwidth. So directionally, we think we're going -- we know we're going to continue to grow in 2023 faster than the market growth. This is a market that typically rose only 5% a year, as I said, we were at $78 million in '18. We're going to be at $225 million this year, so we grew much faster this year over last year was about 57%. So we'll continue to grow much faster than the market. What we have to watch is, how that market is going to transition from DDR4 to DDR5 and the impact in the short run on our product margins because we have better margins on DDR5 than we have DDR4 for obvious reasons, yes. Question Christopher Caso (Analysts) Right. And that leads to my final question is regarding margins and how do those transitions happen? I mean in one, if product -- as product revenue grows, is that accretive to margins? And then DDR4, DDR5, CXL how does that affect margin as those transitions happen? Answer Desmond Lynch (Executives) Chris, it's a great question. We've been delighted with the progression of our financial model, as Luc mentioned, this year on a adjusted 605 basis, we'll grow about 25% as market revenue growth year-over-year. From a gross margin perspective, we are operating at 80% to 85% on a gross margin basis, patents are 100%, Silicon IP is 90% to 95%. And on the product side, we are 60% to 65% is a long-term goal. And if you look on the product side, our products through the first 3 quarters of this year have been at 62%, so almost at the midpoint of the range that I gave you from there. Going forward though, as products grow faster compared to the other revenue streams, we'll see that our total company sort of gross margin will take down, but we'll continue to see high fall-through to the sort of bottom line and nice leverage on the bottom line from there. We've been very disciplined in our operating expense management, we're about $55 million per quarter today. SG&A, we've made the majority of our investment, so you'll see nice leverage there. R&D, we'll continue to invest in the high-growth opportunities ahead of us and you see an operating margin that's in the sort of 40% to 45% range with really strong free cash flow generation as well. So overall, as products continue to grow, we'll continue to see nice leverage on this sort of bottom line. Question Christopher Caso (Analysts) Right. And what about longer-term targets for that, is that 40%, 45% operating margin consistent with longer term, what do you think the business... Answer Desmond Lynch (Executives) That's what I would say, yes, I think being in that sort of 40% to 45% with continuation of the strong cash generation is the right model to be thinking [indiscernible]. Question Christopher Caso (Analysts) Okay. We're just about out of time, is there any follow -- final comments you'd like to make? Answer Luc Seraphin (Executives) No. We're glad we've moved the company from being a patent licensing company to a product company and we're excited about the prospects of the data center markets where the potential is really nice. Question Christopher Caso (Analysts) Well, great. Well, gentlemen, thanks both for attending. Answer Luc Seraphin (Executives) Thank you. Answer Desmond Lynch (Executives) Thank you, Chris. Question Christopher Caso (Analysts) Thank you, everyone. Thank you. Answer Luc Seraphin (Executives) Thanks. Copyright 2022 © S&P Capital IQ |
return to message board, top of board |
Msg # | Subject | Author | Recs | Date Posted |
795732 | Re: 11/29 | hudsonhawk2u | 2 | 11/30/2022 7:10:58 PM |
795733 | Re: 11/29 | valuationguy | 9 | 11/30/2022 9:03:25 PM |
795736 | Re: 11/29 | dfwl28 | 0 | 12/1/2022 8:31:33 AM |
795796 | Re: 11/29 | twobytebus | 7 | 1/12/2023 4:32:27 PM |
796636 | Re: 11/29 | twobytebus | 4 | 10/28/2023 2:31:46 PM |
796649 | Re: 11/29 | twobytebus | 2 | 10/30/2023 12:09:32 PM |