Silicon Motion Technology Corporation ($SIMO)

So good morning, everyone, and thank you for joining JPMorgan's 54th Annual Technology Media and Communications Conference. My name is [indiscernible] SMID-cap analyst at JPMorgan covering US semi and semicap equipment companies. Really pleased to have Wallace Kou, President and CEO of Silicon Motion; and Jason Tsai, Chief Financial Officer, joining us here today. Wallace, Jason really appreciate you both being here and looking forward to the conversation. Before we dive into Q&A, we'd like to give you, Jason, a few minutes to introduce Silicon Motion for the audience. The story has evolved meaningfully over the last 18 months or so. So this, I think, would be a great way to set the stage for the conversation.

Yes. So our business has scaled quite meaningfully over the last couple of years. If you take a look at our expanded portfolio, our core business historically was an eMMC -- embedded eMMC and UFS, primarily for smartphones and now increasingly more important for IoT and other connected devices. In edge computing for SSD controllers, that's been an opportunity that we've been capitalizing on as our share gains here have significantly accelerated with the latest generation. And then you look at some of the new opportunities that we've been going after pretty aggressively, right? On the enterprise side, we've got enterprise boot drives that has been scaling -- beginning to scale [indiscernible] at the beginning of last year and scale more meaningfully throughout this year and going forward. We have our enterprise MonTitan SSD controllers that we've talked about winning with 2 Tier 1s and 4 additional customers. Now we're seeing sell-through with that. We have 5 Tier 1 CSPs that will be ramping throughout this year and then ramping more meaningfully into next year. And then lastly, our automotive Ferri business where it's a business that we've been investing in for the better part of 10 years. We've got a complete portfolio here that we're able to serve an automotive market that is increasingly having a harder time to secure opportunities sourcing. And so being able to step in, provide automotive-grade solutions that we've been working with them on for the last many years has created this really strong inbound opportunity that we're seeing where there's a lot more carmakers, we service virtually every carmaker around the world, whether it's internal combustion, battery electric, autonomous or not, we're seeing -- we've got business across the board. And so I think as you -- one of the things that have resonated well with investors and one of the things certainly that we've been really proud of is that across each of our businesses, we're scaling with the top customers around the world. This is a multifaceted growth driver where we're seeing different end market catalysts, but all of those catalysts are creating a significant tailwind to our business. And so I think we're really well positioned. Obviously, this year, we're looking at pretty significant growth. If you look at what consensus expectations are. But some of these big opportunities in enterprise and boot drives an automotive are really just beginning to scale towards the latter part of this year. That will continue to fuel long-term growth into '27, '28 and beyond.

Got it. If we could maybe touch on the broader memory environment. Can you perhaps share your views and the outlook from here? And specifically, whether you see the uptrend in pricing that's been ongoing for a bit now continuing into next year. And related to that, how is Silicon Motion navigating the current memory supply tightness. And [indiscernible], this kind of tightness has been historically a constraint on your business. But so far in this cycle, you've been thriving. So what's different this time?

I think -- because AI is -- a strong demand for AI, definitely both DRAM and NAND will be in shortage continually for entire 2026 and also 2027. You won't see any meaningful improvement until 2028. As interim probably late '27 have some meaningful improvement. Because this is very unique. It's not like in the past 10 or 20 years, cyclical for NAND, DRAM, supply shortage and oversupply because this is a strong demand from AI and not just the infrastructure, but it's an implementation application go everywhere, it go to cooperate itself. So we won't -- I think nobody can see it's just a short-term shortage. So how you can leverage shortage and price up to gain market share, they become very critical for Silicon Motion. That's why we leverage the trend. Our 5 product line versus [indiscernible]. And second is our embedded mobile for [indiscernible] and service enterprise controller and enterprise [indiscernible] solution and automotive Ferri solution will grow sequentially. I think we'll continue to grow [indiscernible] or beyond. We have a strong leadership with NAND maker, where the only company have active projects with our 7 NAND maker simultaneously today. So that's why we can leverage NAND supply and supporting our end customer.

So maybe just touching on smartphone and PC demand specifically, do you expect end market demand to deteriorate further into the second half of this year and that's on the back of these additional memory pricing increases that we talked about. These end markets have outperformed so far in '26. So do you view the incremental price pressure as more of a risk or an opportunity?

I think the impact regarding the NAND supply shortage and the price up, it definitely impacts both smartphone and PC, a top about smartphone. I seen the major impact really for value line smartphone model [indiscernible] value line player going to suffer most in 2026 and partly '27. We see the most of analysts prediction will be 10% to 13% decline in 2026. However, definitely, iPhone going to gain market share is through the strong procurement capability. I think Samsung will also gain market share. But to say that, Silicon Motion will continue to grow our mobile embedded eMMC-UFS. First of all, as dynamic outsourcing their project in terms of port to Silicon Motion. That's what we gain with the NAND maker. Second is because several NAND makers, they walk away from more business. So they're selling wafer. We own 90% of the module maker who provide solutions to smartphone makers, especially in China, like for Xiaomi, for Oppo, Vivo, and Transient. That's why we also gained market share from there. We believe we have very, very strong growth for both eMMC, UFS. In addition, I've seen so many new IoT devices. We see the smart glasses [indiscernible] be 60 million units this year, Smartwatch, smart TV, new [indiscernible], smart door lock and so many new [indiscernible] devices we're probably going to ship more than 400 million unit eMMC this year. So this is very great to be in today and thank for the effort in the past 10 years we will continue to gain market share during major smartphone market declined 10% to 15%.

So let's turn to Blue drives. We've seen faster adoption within the NVIDIA Black Well platform since the first quarter of 2026. And how should we be thinking about Blue Drive revenue for Silicon Motion this year? And how does that picture evolve into next year as [indiscernible] begins to ship in greater volumes?

I think we're starting [indiscernible] the leading TV company since 2023, '24, they come back to us. The reason is the current 3 suppliers [indiscernible] motion controller. In the past, we support quite a lot of buy controlling firmware to NAND maker and make sell into server and CSP company. I think now they like to provide total solution -- that's why we qualify by the [indiscernible] 3 start to ship from second half last year. Now we can really more from [indiscernible] we own majority for NVLink Switch [indiscernible] as well as the Ethernet [ 6967, ] [indiscernible]. So this has become tremendous going option to us. To say that we changed our business model because we are able to secure the NAND. So for the #1 search engine company, we also start to support the boot drive solution instead of controller, we're also winning the #1 telco company for Puja as well as additional new server maker. So I think boot drive solution for enterprise would be meaningful significant business for us to grow in the next few years.

I think 1 of the things to point out here as well is that as well as pointed out, it's a very diversified range of end markets, customers we're not -- our business isn't dependent on 1 customer or 1 product and hoping that things work out well, right? This is across a wide range of end market with multiple different factors affecting each of the demand factors there. And so we're really excited about the opportunity here as we further diversify our business. We're also engaging with other customers in this area that will -- we're excited to talk to you guys about longer term. But again, the expectation here for us internally is that investing in this business means a very diversified, strong revenue stream longer term. .

Right. And Wallace, you touched on the boot drive opportunity at the leading search engine company, when do you expect revenue contribution for that particular product to begin? And how do you think about the scale of that opportunity relative to to NVIDIA, I mean, this company's AI, AI are essentially the second largest deployment of AI accelerators in the market. So the natural question is whether the revenue contribution could be comparable to the NVIDIA platform over time.

First of all, we did not provide the full year guidance, but I believe our boot drive solution business will become very meaningful and will become significant in the company in the next few years. So first of all, we don't compete with a NAND maker and our boot drive unique. Most of the next to CPU boot drive with DRAM. I believe NVIDIA [indiscernible] is really produced by Samsung. But out boot drive is direness with a controller with NAND with our firmware, particularly fine-tune the security portion. That's why we are very unique, provide solution in the [indiscernible] form factor setting a very small M2 board and selling to major customers. NVIDIA is just 1 of them, right? So -- we see the capacities from 256 gigabyte all the way to 1 terabyte and with a price increase for NAND. So the ASP also increased dramatically. We are able to maintain 40%, 45% margin in the next 2, 3 years. We see there will be a very meaningful contribution for the company for growth.

Right. So as that boot storage TAM expands meaningfully, how do you assess the competitive landscape? Who else is showing up here? And what gives you confidence in sustaining the market share outperformance that you've built?

First of all, for NAND makers, this portion might not be the most important product for them, right? So -- and then using the R&D resource to put in. So the only probably 1 or 2 NAND maker participating, but most of them don't have durab;e solution. And provider for NVIDIA BlueField 4, all use Silicon Motion controller. So we dominate for this market even for search engine, Google, I think the other 2 companies, all use Silicon Motion, different controller [indiscernible] different NAND. And so we feel very comfortable to grow this market with our unique technology and serve our major customer.

And then on SSD solutions, which is now approaching about 20% of your revenue mix, and that's up from low to mid-single digits and in prior years. As that weighting continues to grow, how should we think about the gross margin trajectory going forward? By our estimates, boot storage solutions is running around 35% to 40% of margin today. So the natural question is whether that mix -- the shift in mix dilutes the 48% to 50% corporate gross margin target as that weighting keeps moving higher.

In the past few years, we are not aggressive to engage with the automotive solution business and the [indiscernible] solution because the margins were too low, right? So there's will jeopardize the corporate average gross margin. But now because NAND supply shortage, we are able to secure NAND supply enterprise customer willing to pay for. We can pass through the cost incremental to the end customer. So we want to maintain, we believe, as I said, [indiscernible] margin for [indiscernible] as our motive is comfortable. And because our motion family had much higher gross margin is offset the [indiscernible] Solutions margin. So our copac average go we will maintain 48% to 50% gross margin for the next few years.

Okay. And then turning to the enterprise SSD controller business. With rising inference demand, which includes dynamics like [indiscernible], offload, how large do you currently estimate the enterprise SSD TAM to be? And how much of that is addressable for Silicon Motion over the next, say, 3 to 5 years?

This is a very exciting moment for a storage company today. I think the growing opportunities is unbelievable. You're looking forward the tier for the storage, the lowest the co data is HDD, but second tier is a [indiscernible] storage right? This is for QLC for high capacity. But due to -- because the price increased a lot. So we see them really ecosystem for the data storage capacity the [indiscernible] from 12 terabyte down to 64 terabytes, but the [indiscernible] opportunity. But so far, because the QLC output from the total [indiscernible] lessen at 25%. So I think the demand is not as strong as expected due to the NAND supply limitation. But the next, I think it's a computer -- this is next to CPU. Normally, the density is about 4 to 8 terabyte primarily is NAND maker provided to all the server maker as well as NVIDIA and Google. Now through the March [indiscernible] NVIDIA, Janssen, they announced CMX is contact memory storage. This is new because it is directly through the for inference, right? So this has become very, very interesting and high demand for the customer. We see each of the GPU neither on GI and defined by NVIDIA [indiscernible] terabyte. So MVC72, which means each of the server rack need 72 drive, each have a 16 terabyte totals 1.16 patabyte. So this is very new and they're going to consume a lot of NAND output. I think about how many [indiscernible] going to ship verruben server in this year and next year. So this is a new opportunity for Silicon Motion. So our MonTitan, not only for the WAM data for the QRC, but also suitable for PSD, neck to CPU as well as MX. And we haven't mentioned the storage next which were ramped up in 2028 for the low latency drive for the contact less than 4K by or 512 [indiscernible]. I saw so many new opportunities. I believe Google, AMD and others, their architecture have a different, but they all need a similar TVCs story, too. So the tremendous new opportunity coming for the controller maker as well as NAND maker.

On the MonTitan road map, you're now expecting more TLC NAND adoption in the near term. How large can the TLC TAM be for Silicon Motion? And when do you expect more meaningful QLC adoption to materialize? And how should we think about quantifying the relative QLC versus TLC opportunity?

I think the QLC the output today is less than 25% worldwide for NAND maker. It takes time for QLC to be qualified by CSP and Tier 1 service maker. So today, under supply shortage on the TLC a lower capacity become more favorable. That's why majority of our design win and also shipment customer use TLC, but some use QLC. I believe when Samsung and SK Hynix, V9 QLC become qualified and ramping from second half '27 to '28 and QLC will become mainstream since 2028 and will be more demand for high-capacity SSD during that time.

And then on HBF, can you maybe walk us through the projects you're working on? And how does the HBF ramp change Silicon Motion's opportunity set?

HBF is high bandwidth of flash. It's a new concept, new I think initiated by [indiscernible]. This is a purpose to try to support [indiscernible] demand, reduce the capacity for HBM, right? But I think it's an initial stage. We believe this should be developed by NAND maker choose experimental process. And if you look at technology fundamentally, it's very similar to [indiscernible] years ago from Intel. You need a very, very fast nonvolatile storage to feeding DRAM as well as [indiscernible]. But I also need system software chain for memory mapping. We are invited by some NAND maker, but due to resource constraints, we [indiscernible] turn down during time. We will watch out the market trend continually. If -- but I don't see HBF will being NVIDIA platform might be in other CSP platform, if they needed, right? So we watch it closely. We want miss opportunity to become real.

Okay. Let's turn to PCI Gen 6 enterprise controllers. What's your sense on the time line for introduction? And how does Silicon Motion's time to market stack up against the competition?

PCIe driven by NVIDIA is the first for [indiscernible], especially for [indiscernible] also will be used for CMX later. Initially, [indiscernible] will be PCIe. Silicon Motion, we will tape out very soon. We're sampling in Q4 this year and production in maybe Q4 '27 and '28, but at the same time, we also have separate in develop PCIe 7 because NVIDIA driving the momentum every year changing. I think we'd like to become leading provider [indiscernible] during the time. In [indiscernible] we have much more design win from both U.S. and China CSP as well as 2 NAND maker, they're all major [indiscernible] controller.

So let's move on to client SSD controllers. Do you expect market share momentum to continue from here? You're already above 50% in PCI Gen 5 share? So is there room to push higher? And is there perhaps a market share ceiling you'd flag that customers typically where customers typically would want some level of supplier diversification. With PCI Gen 5 adoption accelerating this year and next year, how should we think about then the ASP uplift as well. And in addition to that, do you have any early updates on progress with upcoming PCI generations.

For the PC industry, it's very interesting this year because the NAND maker have less allocation to PC OEM due to the supply constraint. We benefit from our market share gain because we had about 33% market share with [indiscernible] in the past, but we have more than 50% market share gain in [indiscernible] when PC OEM start to ramp up, see a channel for channel, primary for general in the second half of this year, we're continue to gain market share. So it doesn't matter see the overall PC market will decline 10% to 15% from unit shipment. But I personally think is much better than 10% to 15%. I think probably 10% or even less because Apple's MacBook and go and gain market share and several other NAND PC OEM going to produce new lower cost and very same model coming in the second half of this year. So overall, the demand is lower capacity for SSD, but also performed well to meet some consumer demand. So I think we will continue to grow market share for both [indiscernible] Gen 4 and Gen 5, and we continue to benefit from NAND maker outsourcing to SMI because I think Samsung [indiscernible] move a majority of our their NAND solution team to HPM. I think as everybody knows, [indiscernible] will be customized. So it's not a standard product selling to everyone, all the CSP and the CPU maker won't have customization to meet a certain demand. So we believe we're going to benefit from outsourcing project from NAND maker in both mobile and clients.

And before we continue, are there any questions in the audience? Yes.

In terms of QLC, I want to ask you guys about your road map for optimizing latency. And I think you guys have some interesting research that's probably going to be coming up in terms of how you're handling that. So I would love to get some insight.

We cannot touch specific technology, right? The QLC, we're being worked with our NAND maker in the past 10 years with [indiscernible]. Their first sample come to Silicon Motion. And it's very interesting, you're looking for inference. You're thinking about the KV cash concept, key value cash for enterprise, very easy to understand. But how that can transfer to mobile and to PC, it become a totally different technology because limited DRAM and limited SSD, the storage space. So you think about how the layer data transfer between from CPU and to DRAM to the storage device like UFS mobile, SSD in the PC. You need a certain technology, work with a system provider, and they can optimize within regarding data arrangement to make sure the performance and meet the consumer end user demand, right? So meaning, it's not a unified memory architecture. It's something combined between SLC, QRC, TLC, certain cash mechanism make user sale, you are running at within the cash. I don't know whether I answer your question?

Any other questions?

Thank you. So we talked about the margin consistency for the foreseeable future, but I think something that needs to get a little bit more credit is like the operating leverage that we've experienced over the last year and change. We've gone from Q1, it was like high singles to now we're guiding north of 20% operating margin. So how do you balance invest in new technologies, R&D and also memory price through withdrawal variable costs while also leveraging a certain OpEx base to continue ramping this...

So the cost of NAND is going to the gross margin. So our ability to maintain that gross margin line consistently is balancing growth in SSD solutions with ramping on Titan, and we've been able to manage that pretty well. I think there's a lot of value that we're providing and building a solution that enabling us to get better margins on the solutions side than ever before. But we're also -- part of the value is being able to procure that NAND, having that security of the supply chain. . On the operating expense side, we'll continue to invest, right? I mean from an OpEx dollar, certainly, we expect to grow pretty meaningfully this year, but revenue growth is growing a lot faster. To your point, operating margins are getting back to where historically we've been, our normalized range of 20%, 25%. Longer term, we're confident that we can still get to that 25% plus long-term operating margin target as we continue to scale revenue but continue to invest alongside. There's going to be additional projects. We talked about 4 nano project that's for PCI 6 that will tape out in the third quarter of this year. We'll do more of these advanced process on, which is whether for PCI 6 or 7 for UFS 6. There's new technologies that we're investing in today that will require us to continue to tape out. But we see significant revenue opportunity with these new products that help drive that operating margin leverage.

Yes. We and the [indiscernible] centers, right, gross margin more than 80%, but our controller cannot increase margin quickly because we try to retain and maintain the customer. It's a great opportunity for us to see our fee for automotive as well as [indiscernible] solution gaining market share. And we want to leverage our technology controller and firmware as well as our ability to secure the NAND to maintain 40% to 45% gross margin. So product mix will maintain 30% corporate gross margin. But because our goal is to grow top line aggressively. So way for our Q2 earnings call and with the Q3 guidance, you can probably understand how fast we can grow in 2027.

Any other questions? So maybe touch on the mobile controller side. How do you see growth playing out here? And do you expect increased outsourcing opportunities from the NAND makers? And on eMMC specifically, where what's the current composition of the market? And how does it split between mobile and nonmobile applications today? And are there additional growth drivers beyond mobile that we should be watching out for?

We see NAND maker beside YMTC in China because they have obligation to support the local smartphone maker. All other NAND makers, they probably only focus on high-end. So they're going to outsource the mainstream [indiscernible] to control the [indiscernible] because the current solution [indiscernible], their NAND is a legacy generation. When they move to from B9 like Samsung [indiscernible] to V10, they have to use additional solution. If they can continue to play the UFS 4 and 3 in the market, right? So we see great opportunity to Silicon Motion. For EMC today in the smartphone is less than 10%, and I think they continue going down. But EMC great opportunity, as I said, for the IoT devices and other smart devices. So we own almost a 90% module maker who is supporting for that sector. Doesn't matter Google, Meta, Amazon, Xiaomi, their smartphone, smart glasses, all using SMI controller, right? So I think in that portion, we see growing very well, especially new generation set-top box, [indiscernible], Smart TV and Doorlock. It just gave a tremendous opportunity to expand. But that portion, we don't want to touch solution. We only focus solution controller, provide for all the module maker and some NAND maker to grow. We'll continue to grow our embedded EMC controller, especially this year, growing very, very fast, even in the smartphone decline.

Great. And then finally, on capital allocation. Can you maybe outline for us how you're thinking about use of cash, given your quite strong balance sheet position. Should investors expect the cash dividend to scale meaningfully alongside your strong earnings growth trajectory. And -- then maybe on M&A, are there any key areas of focus that you'd want to highlight?

Sure. From a capital allocation standpoint, as you know, we've been paying a dividend for the better part of 10 years. We've raised that periodically as the business is scaled. -- and profitability scale. One of the things that we look at more closely now, though, is certainly as our revenue is ramping much more meaningfully, working capital needs have also gone up. Inventory has gone up. Inventory days have stayed relatively stable around kind of the 7-plus month range. However, dollar inventory has gone up quite a bit now to north of $400 million. As our business continues to further scale as we have, obviously, ambitions to continue to grow our SSD solutions more, that's going to require, again, continuing purchases of NAND, continuing increases of inventory. So we're trying to balance certainly, capital allocation, capital return to shareholders as well as our working capital needs. .

Got it. Well, we're almost out of time here, gentlemen. So thank you so much for the time. This has been a great update.

Thank you.

Thank you.