Okay, great. I want to thank everyone for coming out as we get into Wednesday afternoon. I appreciate everyone coming. And for this next group, we'll have Silicon Motion. And we're pleased, just only a few conferences a year, we get Wallace Kou, founder of Silicon Motion. We also have, as many of you know, Jason Tsai, the CFO. And for the discussion today, we'll keep it pretty informal with a fireside chat so we have a chance to go through a lot of the topics. And for Silicon Motion, it's been quite an impressive year, strong revenue growth. And I think coming from founding the company to now reaching $1 billion revenue run rate with quite good growth across product lines. I think, Wallace, maybe to kick off, to kind of bring everyone up to speed, how does the business mix look?
As we build and get to this billion-dollar run rate, how is the business cut across the applications and the key areas you're in? And then we can kind of move forward and start to talk about the next step, the next billion.
Yeah, thank you, Randy. We have four major product lines. The first is Client SSD, is about 50%-60% of our total revenue. Today, we have about 30% global market share. Second is a mobile controller, eMMC UFS, about 30%-40% of our total revenue. We have about 20%-23% global market share for smartphones, and also including the IoT devices. Third product line is our Ferri Automotive, and we have around. We are ramping quickly. Last quarter, we saved about 5% of total revenue, but in 2026, we'll be growing to 10% of total revenue. We do have a very strong design wing, and the ramping, we saw the leader, including Toyota and Tesla, BYD, and several others. The fourth one will be Enterprise MonTitan product line. This is our PCIe Gen 5 with both 16 channels and 8 channels.
We have mentioned we have two tier-one customer design wins that start to ramp from 2026, and a total of six customers today, and we'll be 5%-10% of total revenue by 2026, 2027, and we also have an enterprise BlueDrive design win and start to ship to BlueField-3 customers. They're going to continue quite a lot of interesting storage for BlueDrive, including Google, so this is our major growth driver going forward.
Okay, that's a good overview. And we can dive a bit deeper into some of those product areas. Curious for enterprise, if we go back a few years ago, you had the Shannon business, which had ramped up and for a couple of years looked really promising and then ended up adjusting. But go through the history on your approach to enterprise. And as you come in with MonTitan, how's the approach? How has it changed? And how do you see the confidence to kind of bring that up and grow that into a bigger business?
Yeah, Shannon acquisition is 10 years ago. That's when we're starting to engage, expand to enterprise business, because Shannon is based in Shanghai. Most of our China customers want the solution, so it's a learning curve. Shannon has a total different architecture. Our initial engagement is with Alibaba and Baidu. It was successful, but it's just we don't have the expert know-how for NAND procurement, so we lost quite a lot of money. I think now MonTitan, we started four years ago, regrouped the team and focused on controller only, and then we said it's a firmware. We have a pretty flexible business model. We provide turnkey solutions with firmware. We provide SDK for tier-one customers that can develop their own firmware. Also, we have a SAMI customer, so we're based on different customers' workloads, and they can fine-tune based on the different NAND and FTL.
So this gives us flexibility to expand our business, especially QLC moving forward has become very important for high-capacity SSD. We are the leading controller for QLC, working with all the NAND makers. It gives us leverage and a position for the high-capacity SSD.
Yeah. You've talked, I think, about toward the end of next year, into 2027, 5%-10% of revenue. How is the maturity of the product, just in terms of technical milestones, hurdles to bring it up? And how broad do you see it? Because you mentioned having a good number of customers, I think six customers, eventually like two to start up. But how do you see the ramp-up into those customers and maturity of the platform?
So most of our firmware, Silicon already matured. Firmware is final qualification in the customer, in the customer site, because one of the customers is a NAND maker. So really, that's a secure supply. And also, we are pretty confident about the ramp of the volume. The other tier-one customer in the U.S. is also ramping in the second half next year. But all others are in the final stage for qualification. So we're pretty confident regarding the current guidance about the revenue. And I think if the customer can procure a NAND better, we may change our guidance in the second half of next year and probably better numbers.
Oh, that's good. Good to see better numbers. And we'll talk a bit more later about the constraints. How do you see the TAM for enterprise in terms of opportunity? And with this initial customer set, how much of it do you think you've cracked to go into it? Is there a way to think on units or value of these controllers, what it could be?
We are very excited about the enterprise business for our growing momentum because our PCIe Gen5 MonTitan is a very unique architecture. Their performance shape is made more suitable for the AI ecosystem because performance and the power can be scalable based on different workloads, and they can change dynamically. So most of our CSP customers and also NAND makers, they really like it because it's much more flexible. And as everybody knows, the warm SSD demand for high-capacity has become bigger and bigger. And it's all required to try to be QLC to make it cost-effective. So this gives us a strong position to focus on the warm storage. Although suppliers are still limited, but I think by the second half of 2026, we're going to see more NAND makers provide the output. In addition, compute storage is TLC-based, 4 TB- 6 TB .
I think we're already in the initial ramp for next year, TLC. Second half will follow by QLC. Even, we haven't talked about Gen 6. I think the Q2 earnings call next year will give us good guidance and a picture about the Gen 6 development. Gen 6 design momentum is much stronger. Even our Silicon won't be available until the second half of next year, but the Gen 6 design will have multiple tier-one, including our Gen 5, but we have additional couple of tier-one coming for Gen 6 controller.
Yeah. And if that silicon is available second half of 2026, how long does it take to go from the silicon-ready to solutions? Because it sounds like.
It's about nine months to one year for qualification, especially in the end customer side, but including two NAND makers.
When this comes up, so it sounds like that, as you move beyond 5%-10% towards second half of 2027, 2028, it sounds like it scales up a bit more from there.
We didn't give a guidance yet.
Yeah. So my goal here, our goal here is to first provide the first waypoint, right, which is 5%-10%. And we started talking about this back in 2024. Once we achieve that waypoint, we'll talk about how that inflects going forward with new products, new customers, how that scales. Certainly, there's a lot of demand around QLC high-capacity SSDs. Right now, the gating factor being availability of those dies, but as that opens up, we should be able to give you guys a better understanding as to how that scales longer term.
Yeah. Does the hard drive constraint help that? I mean, NAND is tight, but there's also the constraint on hard drives. Is that accelerating some of the activity where they're looking at this solution that could help this business ramp up?
Exactly. I think the enterprise SSD is more suitable than HDD in the AI, especially inference. That's why the warm storage has a very high demand. But now the major challenge is really NAND supply shortage. If NAND supply can be stronger, our dependent customers we pick, we engage with, I think we're going to see a very strong ramp-up.
Okay, good. Actually, the other one I want to address, I think boot drive. You introduced it a bit, and that's one where you talked about being on BlueField, so NVIDIA's GPU product line. Could you talk about traction both for GPU and also for ASIC market with what leader in ASIC? Just your position there. And how's the business growth for that as we start moving to new platforms like Rubin or we move to next-generation ASIC designs, if that business actually looks more promising to scale up?
So we started engaging with NVIDIA two years ago for BlueField-3 design. BlueField-3 is a GPU. There's a network accelerator they use for the NVIDIA storage ecosystem. Initially, I think we did not see there would be a very big volume or will be meaningful revenue for Silicon Motion. But I think lately we start to ramp in September. The current forecast in the PO is quite large. We have to work closely with the NAND supplier to make sure to secure the supply because this is a sole solution. The first time we put a boot drive solution, initial range about 128-200 GB, and the margins align with the corporate average margin. And we're also in the process for qualification for BlueField-4, and NVIDIA gave us four additional switch projects for boot drive. This is also range from 256-512 GB.
BlueField-4 will be 512- to 1 TB range. Now we see this business become more interesting. We haven't given guidance yet, but in addition for MonTitan business, although it's part of MonTitan, but we don't want to miss the investor. Originally MonTitan focused on controller only. The boot drive will be upside for our revenue growth. I cannot tell the number, but I think it will be significant. But the dollar range per unit will be around 20-something to about $100, depending on the capacity. This will be a significant contribution to our top line as well as the profit. Besides NVIDIA, I think, as we mentioned, we have supported boot drive controller for Google since five years ago. I believe we also be part of a TPU, but just this is a controller only.
We also support to the other NAND maker who support to Google directly. We have other boot drive for certain other tier-one CSP, but I think this looks like a very interesting business model. We had to rethink it maybe in the future, long term, but NAND is now in supply shortage. We might provide solution for other boot drive.
Okay. Actually, and clarify for boot drive, is that one for the NVIDIA solution you have to pass through the memory? And as pricing moves up, can you pass through so it doesn't affect the margin? It's a pass-through, but it's higher revenue.
Yes, that's correct.
Okay. How do we think of the application? Is it scaling up more that you're seeing more NAND attached to these AI servers? So this is more about the inference. And as the reasoning models pick up or the compute and agentic, it ends up expanding. I'm curious where the upside is. Is it just that we're selling more AI GPU systems?
So the BlueField is attaching, not have to be selling with an NVIDIA Blackwell GB system. It can be attached to Intel or AMD because the interface, InfiniBand or Ethernet. So quite flexible. We see this architecture can be scalable. It's not only NVIDIA have a GPU. I think AMD all have GPU, but it's non-popular today. We think this is a great opportunity, and we not only want to expand for GPU, hopefully we'll also expand to the CPU and the GPU boot drive. So this is we are looking for the market TAM because this is very interesting, high momentum, and can grow faster in the near term.
Okay. How do you think of the, yeah?
I think what we've seen here is that when we first started with the first project, after the success there and the win there, we were invited into qualifying for additional projects. Those additional projects bring to us additional volume, but on top of the additional volume, it's higher ASP because they're higher density. So you're going from products that are in the 20s ASP. So when you scale up to upwards of 1 terabyte for some of these boot drives, you're scaling up to north of $100. So not only are we getting the benefit of both the volume increase because of the additional projects that we're winning, but we're also benefiting from the ASP increase from that higher density. So we're very excited by the opportunities.
To Wallace's point, I think over the next few years, this is a business that can scale to become much more meaningful.
Yeah. Our initial intention for boot drive is really to enable the high-capacity drive. Because through NVIDIA, we understand the ecosystem for the software partner, like DDN, like VAST Data, WEKA, and go through the Dell, EMC, go to NetApp, go to Cisco, go to IBM, go to HPE. So this is the customer we're talking for the high-capacity SSD beyond just CSP. But suddenly, the boot drive itself becomes a very interesting business. So there's upside for MonTitan.
I think, Jason, you're saying it's single digit. It's still single digit contribution now, but it's.
We haven't commented on what the contribution is. So once we get into next year and we talk more about kind of next year, we'll kind of frame the opportunity a bit more.
But it sounds like even with all the NAND pass-through margins and what you're running this business, it's good margin.
It's good margin.
Yeah. Okay, good. Actually, I want to switch to a couple of the traditional applications. Smartphone first, I mean, there's two sides. One would be just opportunities. And I feel like there's always been in the mobile, the higher end, these new UFS standards. And in the past, it was kind of tied to individual memory companies, whether it was Hynix and then Micron. How do you see the opportunity now for those higher-end controllers, like the mobile business? Do you think it's going to be driven by standard upgrade to UFS 4.0? And are there areas to gain? I think you were saying low 20s market share.
Yeah, we're about 20%-23% market share today from global market, including iPhone, which we don't have. But I think the UFS 4.0 becomes very important for the company. Not only we can grow with the existing NAND partner, but also we have opportunity to catch the outsourcing from a Korean company because the UFS 5.0 will become high-end in late 2026 and 2027. I think the NAND makers, they are not going to redo the UFS 4.0 become mainstream. So they need a new controller and supporting the upcoming new generation NAND to consume the NAND output. So they give a tremendous opportunity to expand not only in the mainstream, the low end, but also expand to the high end.
Okay, that's good. So it sounds like 2026, 2027, UFS 5.0 might be the time your market share.
We will also get an opportunity to invite to be participating in UFS 5.0, but I think my play is low key. We don't want to compete with Samsung, Micron, and Kioxia, but there's opportunity. We catch the UFS 4.0. We in parallel, we have UFS 5.0 coming in 2027.
Okay. Certainly, had when your U.S. customer, I think they exited some of the managed NAND, how do you see the business with that customer continuing on?
We continue to support the same controller for the automotive. But I think the Micron probably already announced they work with us to continue selling the NAND wafer to support the multi-maker. So I think that will help us to grow for multi-makers, especially during the NAND supply shortage. We feel very comfortable for the high growth for 2026 and into 2027 for our mobile controller.
Yeah. Could you talk actually broadly about the supply shortage? I mean, in the past, there were certain cycles that impacted the unit demand. Just when it gets very tight and there's pressure on the bill of materials, pressure to spec down. How does the impact from rising NAND price, how do you think it impacts the business next year?
I think this is the first time in the past 30 years you'll see the HDD, DRAM, NAND all in short at the same time. And because it's short at the same time, which means it's not because supply production cut, it's really demand is very high, right? Although maybe some double booking, but really the gap is so big. Today, 50% of our business, 50+% business is related to NAND maker project. And 70% of our business related to OEM. Only 30% related to retail as well as the embedded industrial. So I think most of our customers, even module maker, they have a stronger buying power with a NAND maker such as Kingston and ADATA. So I think the impact from NAND shortage to us is rather small compared with others.
Because we have so many major developing pipelines start to run on PCIe Gen 5, eight channel, four channel, second half for UFS 3.0 and new generation UFS 2.2 and 4.1, as well as automotive, we have so many new designs with Toyota, with BYD, with Xiaomi and Mercedes and General Motors, and even including Ford. So it's a tremendous new opportunity for us to grow and to offset some small impact from multimaker if they cannot secure the NAND. So overall, we feel very comfortable to move into 2026, waiting for our earnings call for Q4.
Okay. How severe is this shortage? Because it's unprecedented when you mentioned all memory types. But are some channels struggling to get parts where you could see certain segments of, say, tier two, tier three smartphone brands in China, or you could see PC brands a bit tight? Is it that level of severity, or is it more just they're managing through higher pricing?
In my personal view, the shortage will stay at least two years. I kind of comment 2028, but 2026, 2027 will be shortage. There's no way in the near term to change the situation because the equipment, it takes time, takes about 52 weeks, and I think until the NAND, the DRAM maker from Samsung, Hynix, and Micron, they start to ramp up their mega fabs in the second half of 2027. I think we'll have to see some relief from DRAM, then their K-pack will move into the NAND, will be more. Today, some of the module makers are very hard to get a supply from the NAND maker due to NAND price really went up sharply in the last three months. You look at October, they went up 25%-30%. November, they went up again for 50%. It's just incredible.
It's very hard for module makers to deal with this current situation. I believe when NAND price goes crazy and for spot market, I think we'll scare away some module makers, but it will cool down a little bit. So I think we will see natural balance because the OEM price is not like a spot price. But today, I think the channel pricing and even automotive makers, there's a loss group to willingly accept the price up, but the price really up. So I see that I feel sorry for some small players for module makers, but that's the way it is. Hopefully, NAND makers, they can be more balanced about the situation and moving forward to 2026.
Remind us that module, the ones you think are a bit more at risk, how much exposure do you think you have in that business? Is that more kind of some of the lower end, like eMMC, or would that be SSD? What parts of the business could you see some impact from some of those smaller module houses?
Smaller module maker, probably some is not. Most of our customer module maker is a much bigger size because it's a public company, right? They have a more financial position to bargain and purchase the NAND. And most of our customer module maker, they have 8- 12 months inventory already in hand. So I think it should be okay for 2026, but maybe we'll start to face the challenges for second half 2026, 2027. But according to some of our customers, I think look like they negotiate with NAND maker, they're bid supply 2026, 2027. 2025 and 2026 are similar. But I cannot really comment for all the different customers because everybody has different positions. Hopefully, we feel pretty strong because many of our OEM projects are from NAND maker as they start to ramp up. So they will offset some small impact from module maker.
Yeah, it's good to have those share opportunities in these new businesses like the boot drive, and core SSD. I think in the introduction, 50%-60% of revenue still. When do you see the share gains coming in? Because I think of that market, it might be a little bit capacity constrained and somewhat more mature, but if market share is a potential there, how do you see your share trending?
For consumer clients today, as we state, we're 30% market share for PCIe Gen 4. But Gen 5 controller is even stronger for high-end 8-channel PCIe Gen 5. We have more than four NAND maker design wins and almost 90% multi-maker design wins. So it starts to, currently, I think last quarter we mentioned Gen 5 ramp up is already 15% of total SSD revenue, but it's not OEM ramp yet. OEM will start to ramp from Q1. Our four-channel mainstream controller, DRAM-less, also have four other combination design wins, including two Korean customers. This is also including a 90% multi-maker. So our Gen 5 will be more than 50% for Gen 5 market. We're seeing from Gen 5 start to ramp next year, and our market share is going to go up. So it doesn't matter who is going to win.
I don't know which NAND maker or multi-maker, but our overall market share will continue to work 40%.
Okay. So to clarify, you were saying Gen 4 was about 30%, but Gen 5, based on design win, seems like closer to 50%.
50% or higher.
Okay, 50 or higher. Okay. And then for cycles, are they moving faster? Do we need to start thinking about PCIe Gen 6 for the generation on?
I think PCIe 6.0 in a PC is going to take a longer time. Maybe we'll be 2029 or 2030. We'll see in the high end, probably one in the mainstream. But enterprise is much faster. Enterprise will start to see Gen 6 transition from late 2026, 2027. So that's why we try to speed up and to get a more R&D resource working on the Gen 6 because we have much bigger momentum design win pipeline. And in parallel, we start to prepare for Gen 7. We do have a tier one customer wanting to customize the chip. So the R&D resource are our major concern. We need to invest more and keep a good execution and focus to see that happen. And hopefully, it will be a much bigger pie for our company.
Yeah. Actually, maybe a segue, a couple of questions for Jason. First, it came up on the call. I'm curious, the inventory, because there was actually a huge increase. It sounds like you have a lot of design traction, but when I look at that balance, is that just you're getting ahead of it to prepare parts?
Yeah, that's right. We've got a tremendous amount of visibility towards order flow for the next 12 months. And certainly building up inventory for NAND as well to supply some of these new opportunities in automotive, in boot drives. We've been procuring NAND as well. So it's a combination of all of these products, both controllers and NAND, to support, again, order flow that we've got high visibility towards.
Okay, good. And profitability, it was a clawback. You're back to gross margin, almost like kind of 49%. Operating margins, 19%-20%. That's pretty close to the target range. You probably get asked now that you're here, but is this the level you think based on the mix, the design wins, is this the level or is there room for leverage?
Look, I think for now, for gross margins, 48%-50% has been our historical run rate, and I think we're smack dab in the middle of that right now. For fourth quarter, we're guiding right in the middle of that as well. I think longer term, as we see MonTitan and enterprise business scale beyond 5%-10% of revenue, we will talk more holistically about how that affects long-term gross margin run rate, how that uplift looks like. From an operating margin perspective, we've been spending a fair bit of money investing in our products over the last two years. We had three six-nanometer products. We've been investing in enterprise, and the reality is that those investments are really just coming to ROI, seeing ROI, early stages of ROI for just one product. That's the eight-channel PCIe Gen 5 controller.
Enterprise, we're starting to see a little bit of revenue, but we haven't seen that inflection yet, and certainly the other products, they're coming, but it's still not there yet, so we'll start seeing significant revenue scale, starting to build momentum going into next year. We'll continue to invest. As Wallace pointed out, we've got a 4-nanometer investment coming next year, but we will be judicious about how we spend and how we manage our margins going forward. Normal operating margins for us is 20%-25%. Our target longer term is 25%+ . We're confident that we can still achieve those numbers.
Okay. So it sounds like even with the tape outs, even going to 4-nanometer and the R&D costs, the engineering costs, your sales could grow faster than even this R&D investment you're making.
Yeah, we're going to, well, we'll certainly talk more about what next year looks like next year, but our goal is certainly, again, to be responsible.
Okay. Curious to the cash. I know you have a good dividend where you're paying out steady and it's been increasing a bit, but over $270 million cash. And you have MaxLinear, you have that arbitration that could come through. If just say you were to ramp up and get additional cash, what's the intention for that if it's different ways you reinvest, different cash return, M&A?
I think for us, capital allocation, capital return has always been a three-pronged strategy, right? So first one being the dividend, we've been paying that for the better part of 10 years. It's increased gradually over the last 10 years from $0.60 a share to now $2 a share. The second part is share buybacks. Over the last three to five years, about half of our free cash flow has gone to periodic shareholder purchases. And thirdly is acquisitions. While we have not been very acquisitive, we continue to look for opportunities that are additive to what our core capabilities are. How do we accelerate next generation technology? How do we accelerate capabilities? So that's what we continue to look at.
And so whatever happens with the MaxLinear thing, which is not something we're commenting on, but whatever influx of cash we get, we'll continue to look at all three as priorities and allocate appropriately.
Okay. Maybe I'll squeeze one last one just on the input costs, because there's been more talk of steady foundry price increase and substrates are a bit of a cost. How do you see the cost pressures in the business going into next year?
I think the TSMC cost increase, wafer increases has been at 5-nanometer and below. So we're 6, 12, 28, 40 nano. So we're largely insulated from that. From a substrate basis, I think our operating system team has done a good job in maintaining those relationships and securing enough supply to make sure that we can supply our customers, right? Part of the inventory build is because we have good order visibility, so we're able to make sure we allocate our operations to have them build up enough supply.
Okay, good. I think with that, we have to wrap up, but it sounds like a good story. A lot of growth drivers managing this NAND tightness well and seeing a good way to mix up into some of these opportunities. So look forward to hearing more as we go into next year. Okay. Thank you.
Thank you, Randy.
Thank you, Randy.