Okay. All right. I think we'll get started. Welcome to the Silicon Motion Fireside Chat. I'm very happy to have Thomas Sepenzis, who is the Vice President of Strategy and Investor Relations. SIMO is one of the leading merchant controller suppliers in the NAND flash ecosystem with longstanding positions across client solid-state controllers, eMMC and UFS controllers, and a growing set of embedded and enterprise storage solutions. Thomas, thanks for joining us today.
Thank you for having us. We're very happy to be here.
Great. Would you care to make some kind of introductory comments, or would you prefer I just jump into Q&A?
I think we can just jump into Q&A.
Okay. For investors who still think of Silicon Motion mainly as a merchant client solid-state drive, a controller supplier, what is the right description for SIMO today?
Yeah. We're the leading SSD, UFS, eMMC controller maker in the world. Now increasingly, we're starting to branch out this year, starting in the December quarter with our boot drive solutions, and now in this quarter with our first MonTitan enterprise AI-class controller products. We're heavily moving into the enterprise world. We've been the leading third-party controller maker in consumer, and now we're actively trying to engage with enterprise customers and become a big part of that market as well.
One of the most common questions that we get today for semiconductor companies is. Are you an AI play? Are you a data center play? Can you talk about how to think about AI and the data center business driving your company?
Sure. About four or five years ago, we decided that we really wanted to be in the enterprise. I think, at that point, AI still wasn't such a big deal, but we wanted to be in enterprise storage. We did see that as NAND advanced, and you went from SLC and MLC, which are very expensive technologies, to TLC, which is what you would find in a phone or a PC, and then now with QLC. QLC would be an opportunity for us to become an alternative to a hard drive in an enterprise storage solution, because the cost would come down. HDDs are still very efficient and inexpensive relative to NAND. We needed to get to a point where there was a technology that could at least be similar in terms of total cost of ownership in the enterprise.
We recognized that, and we went out and we worked with potential customers, asked them what it was they wanted in a controller for enterprise solutions, and that was the genesis of our MonTitan product, which we have been developing now for 5 years. A couple of years ago, we announced our first two customers, both Tier 1 customers, that were using our hardware and developing their own firmware. We decided we would also develop our own firmware for customers that maybe didn't have the resources to do their own, which we just started shipping our TLC version of that, and QLC is literally any day now out and ready to go, and certainly will be shipping by the end of the year. MonTitan, it's been a four or five-year development. Lots of cash, lots of time, lots of engineering resources.
We actually have our first two customers that are shipping this quarter for the first time in Q2 in production. We've been shipping and sampling, but in production. Production started this quarter, and we'll ramp second half of this year and next year. We've got a total of seven customers that will be shipping by the end of the year.
both for QLC storage and for what we call compute or KV cache, which will work with a GPU or the CPU in a data center or an AI data center to provide instant or close to instant backup for larger datasets than memory like DRAM or HBM can handle.
Multiple terabytes instead of one terabyte. We're shipping both TLC and QLC MonTitan products by the end of this year, and we're very excited about that. We've also got a whole separate product line that has emerged over the last, let's call it 18 months, and that's called boot drives. Boot drives are full SSD solutions. We're not just making the controller here. We're actually buying NAND. We're having a full SSD made for our customers. Our first customer is the largest GPU maker in the world, and they'll be using it with their next generation Rubin platform, where we will be providing boot drives for their BlueField-4 products as well as their NVLink switches and Ethernet switches. We started shipping to the prior generation, BlueField-3, this year, but we did not have the switches before. We're going to get some market share there.
This is a very big opportunity with them. We announced on our most recent call that we'd also secured one of the leading telecom infrastructure makers in the world.
Is going to be using our boot drives, and that's going to start later this year or early next year. We are working with the world's largest internet search engine company, where we're already selling the controller into their boot drive to convert them to a full boot drive customer. As well as other customers that hopefully we can convert over the next couple of years. This business started out very small, even 18 months ago, and now is looking like this could be a very, very big opportunity for us in the enterprise as well.
How do we quantify that? What percentage of your revenues are data center or AI oriented? However you want to qualify it, what is it going to be at the end of the year?
Yeah. I can certainly speak to MonTitan. We've said that we'd be exiting this year, and MonTitan would be 5%-10% of our revenue. We've been saying that for a couple of years, and obviously, a couple of years ago, our revenue was not where it is right now.
In terms of the expectations for 2026. that contemplates an increased expectation in MonTitan.
Over the last three to six months.
We have not really quantified the boot drive business quite as accurately or quite as directly. Again, originally, we thought that would be relatively small, and now it's becoming a much bigger opportunity. I think we'll try and attempt to give you a better idea of what that's going to look like next year when we report our next quarter.
MonTitan, 5%-10% of revenues exiting this year.
At a run rate of 5%-10% exiting the year.
Got you. Run rate of 5%-10%.
Not quite 5%-10% this year, but 5%-10% next year.
Right. Got you. growing, has that business been, to get there, is it doubling, or how do we think about the growth trajectory there?
Well, this is our first quarter of sales in MonTitan, so it's more than doubling.
Right. Got you.
It's going from zero to hopefully 10%-
Right
by the end of next year. It has very strong growth.
The other side of the question on AI is, how are companies using AI to help their design process or operational process? Can you share with us any thoughts about how you're using AI internally?
Well, I can certainly speak to our engineers are using AI in their programming-
just creating better efficiencies interdepartmentally, and deliverables in terms of updating the executive team on what it is that they're doing. I think certainly we've implemented a number of AI solutions that can help with the coding on next-generation products. It's certainly helped me. I have to track many companies, as you do, and it's much easier to track 100 semiconductor companies when you have access to AI and can pull the data from Claude or ChatGPT, whatever you like, to get a better sense of what the rest of the companies in your industry are doing. It's been very useful for me personally.
I'll certainly say that. Not that anyone cares about that. It's probably more about the engineering.
Yeah. No, fair enough. when we talk about the NAND memory cycle, I think on your earnings call, you highlighted that memory tightness could potentially last through this year and into next year, and historically, NAND tightness is typically thought of as a headwind for controller companies. that seems not to be the case. what's going on? Why is this cycle different?
Well, you never want to say, "This time it's different," so I'm not going to say that. I certainly will say that NAND prices have been going up very dramatically, and for the NAND makers, they're more focused on just selling NAND, right? If they're going to sell a full SSD or a full multi-chip package for a phone or for an automotive product, they're actually going to be bringing their margins down, which is not something they really want to explain to you, right?
They're also much more focused on HBM and DRAM, which are higher-margin products, and because the cost of making these chips is going up exponentially as you go from 12 nanometer to 6 nm to 4 nm , you have to really start to rationalize your resources. What we're finding is the NAND makers are actually moving out of NAND first, right? Especially automotive, mobile first, PCs second.
We'll see what happens in the enterprise. They're all still trying to stick around there. You saw one of the large U.S. NAND makers publicly announce they're getting out of mobile and PCs in December. The other one actually just announced on their most recent call that they're going to de-emphasize mobile. What we're seeing is market share shifting to us, right?
To other companies that have at least NAND that can be sold into these markets, particularly Yangtze Memory in China. A lot of times what will happen is a module maker will buy Yangtze Memory NAND, and then they need a controller because Yangtze doesn't make controllers, so they really come to us because we're the 700-pound gorilla in the controller space. If you need a controller for mobile or for PCs, and you have the NAND, then chances are we're going to get a call, right?
that's what we've seen, certainly in mobile. Mobile has been incredibly strong for us over the last six to 12 months, and continues to be. It was part of our strong guidance for this quarter, and we expect it to grow throughout the rest of this year, along with PCs. We're seeing that market share shifts are really helping us as much or more than anybody else in the industry.
No, it's a fascinating dynamic, I have to say. I want to ask you about MonTitan. It's an investment decision you guys made about five years ago, I believe you said.
What is the difference between MonTitan for enterprise or data center environment versus what you might sell into a client product?
Um-
Why is it so hard to do this?
Yeah. if you think about a client SSD in your laptop, it's getting requests from one person.
Right? there's no security layer, or not much of one, because there's no network involved, right? When you're in a data center, you could be getting requests from thousands of computers from that storage server. All that communication has to be managed, right? The security has to be there to make sure that whoever's asking is allowed to actually access that information. The speeds involved have to be much faster. Latency has to be lower. there's many different reasons why an enterprise controller has to be significantly more capable than a personal one.
Got you. Can you talk a little bit more about the boot drive opportunity? Again, it sounds like an investment you started making a while ago. Is there a way to think about what your content TAM per rack is, or how this business could expand over the next one to two generations?
Yeah. The current or last generation technology, we were looking at roughly $20-$30 for a boot drive.
That was a combination of our controller and the NAND that was involved. The NAND was typically 128-256 GB of storage capacity. For the next generation Vera Rubin platform, we're looking at 512 GB to 1 TB . You've also had NAND ASPs going up 50%-70% per quarter over the last couple of quarters. Prices have gotten well north of $100 for the boot drive.
Mm-hmm. Got you. Okay. On this, it seems you have a lot of capabilities you're pouring into this. I'm wondering about the moat that you create with the product. What's the defensible position? Is it security firmware, time to market, NAND sourcing? How would you rank order how you defend that business?
Sure. It will depend on, obviously, whether we're talking automotive, mobile, PC, MonTitan, or boot drive. Let me, I'll just touch on a couple, just to give you an idea. For the boot drives, there's a couple things. The main one being our security. The security in our chip, in our controller is very strong. In fact, the way we ended up with our first customers is we approached them for BlueField 4, and as we were going through the sampling of that, we all realized that we were already in BlueField 3, that it was our controller that was being used by one of the NAND flash makers.
Uh-huh. Mm-hmm.
They just said, "Well, why don't you just go ahead and make the drive for BlueField 3?" They were using our controller because our security was better than the NAND flash maker. We've found that to be a similar situation with the leading internet search engine company, who's also using our controller in their boot drive solution today. People have been choosing it for a while through distributors that we weren't even aware of. I'm sure we were aware of it. I wasn't aware of it. Now we're effectively converting those customers to our boot drive because they're already using our controllers. Now we're starting to gain a little bit more recognition in the industry, so other people are coming to us and asking us for the boot drive and the controller just because of the security.
very important second thing there is the ability to source the NAND, as you mentioned. When we reported the Q4 and gave guidance to Q1, we still weren't sure we'd actually be able to source NAND. we didn't know what kind of market share we were going to have with this customer, if any.
obviously the situation is pretty dire out there in terms of sourcing NAND or DRAM. we were able to work out supply agreements with three NAND makers, so that's become a very big differentiator for us. We actually have customers in other markets, like automotive, coming to us and asking us to help them get NAND now.
Because we've been working with these suppliers for 20 + years in many cases, and have very good relationships with all of the NAND makers. That is becoming a real differentiator. Certainly, in boot drives and automotive.
Yeah.
In MonTitan, we have greater QLC expertise than anyone in the business. We've been working with QLC from all of the NAND makers. When you use a MonTitan product, you can use anyone's NAND. Whereas if you choose Samsung's product, you have to use Samsung's NAND. We've learned little quirks with everybody's QLC that we can apply across all of them. We think we have a pretty good technology advantage in terms of our understanding of QLC for MonTitan. That's translated over, obviously, to our TLC compute and KV cache, where we're already seeing very strong uptick in terms of customers.
We only have a little bit of time left and I have to ask about the margin targets that you set. Where are you right now, and how should investors think about gross margins and your operating margin model? It seems like the target is right here. How should we think about that?
Yeah. From a gross margin perspective, we've always targeted 48% to 50%. I think most people are right at 49%, in the middle of that this quarter.
The puts and takes there are the boot drives and MonTitan. Boot drives and automotive, more often than not, will be kind of in the 40% to 45% range. MonTitan is 60% +. If MonTitan grows more rapidly than boot drives, then we could see our gross margins actually go above the 50%. If the boot drives grow faster than MonTitan, then maybe we just stay in that 48% to 50% range. Right now, we've told people that we're not changing the target.
We want to keep it there at 48%-50%. because we now expect much higher revenue coming from this whole enterprise group, and through the market share gains, and through ASP increases with our legacy products, that should translate to higher operating margin over the next couple of years.
Got you. All right. That has to be the last word. Tom, thanks for joining us.
Thank you again
in San Francisco this year.
Really appreciate being here.
Thanks.