Okay, we'll go ahead and get started. Welcome, everybody, to Needham's 27th Annual Growth Conference. My name is Quinn Bolton. I'm a Semiconductor Analyst for Needham. It's my pleasure to host this fireside chat with Silicon Motion. The company is the global leader in supplying NAND flash controllers for solid-state storage devices, including SSDs and embedded storage. Silicon Motion supplies more SSD controllers than any other company in the world for servers, PCs, and other client devices, and is the leading merchant supplier of eMMC and UFS embedded storage controllers used in smartphones and IoT devices. Joining me from the company today is Tom Sepenzis, Senior Director, Strategy and Investor Relations. Tom, thank you for joining us.
Thanks for having me, Quinn.
Maybe let's start off. The company issued its date for the earnings conference call for the fourth quarter, but in that release, you provided some preliminary fourth quarter results with revenue towards the lower end of the original range and gross margin near the midpoint. What are you seeing in the, you know, sort of what did you see in your end markets through the fourth quarter?
Yeah, so you know, we saw weakness. Seasonally, we generally would see December being up over the September quarter. We saw softness coming into the quarter. Obviously, it was a little bit more than maybe we expected coming in. We were able to still get within our guided range. So, you know, it wasn't exactly a disaster, but certainly would have loved to see a little bit more strength in the end consumer market than we did.
Okay, so was it kind of spread across both smartphones and PCs, kind of more just general macro or consumer spending related rather than focused in one end market?
Yeah, so I think that mobile actually held up pretty well. So it was, we saw a little bit of weakness there, but mobile held up a little bit better in the PC market. We're not sure if it's really just destocking inventory that's going on with the OEMs or if perhaps they're maybe, you know, waiting to introduce higher-end solutions here in Q1, which, you know, they were waiting on some CPU chips from AMD and from Intel. Those will be integrated with our new PCIe 5 controller product at four of the six major flash vendors. So we're actually, we have a really much better market position in PCIe 5 coming into 2025 than we ever, ever have in PCs. We do expect to get 50% plus share of the high end of the PC market starting in this year. We've never really played there before.
So that's 10%-15% of the market we've never had. So we actually do expect things to change pretty rapidly as we go through 2025, but it just was a little bit of a, you know, a blip here at the end of the year.
Got it. Okay. I know there's not much you can say on the MaxLinear arbitration, but maybe just to remind folks, you know, you had, you know, filed for arbitration against MaxLinear for willful and material breaches of the merger agreement. You're seeking termination fee of $160 million plus further substantial damages interest. What can you say about the process? I think we've now selected the panel of three judges with a hearing date in October of this year.
Yeah, so I got to be very careful since it's a legal situation.
Yes, understood.
Generally speaking, we are in arbitration, which is taking place in Singapore in mid to late October. It begins, and we would expect an outcome sometime within the next month, certainly by the end of Q4 of this year.
Okay. And this is a sort of a binding process. There's no chance for appeal on either side.
From my understanding, it's a binding process.
Okay. Okay, perfect. I assume there's not much else you can say, but I'll.
There really is.
Okay, there is. All right.
Sorry.
The other sort of question we're asking many of our companies is with the Trump administration coming in next week, lots of talk about tariff, potential higher tariffs on import, China imports, to the extent that those tariffs are implemented. Would there be any direct impact on your business, or would it just be more of an indirect impact if, you know, consumer devices cost more money coming into the States?
Yeah, I think with that question, it'd be more of the indirect cost. We're not typically the component that is directly impacted by these kinds of tariffs and/or, you know, we're not disallowed like a CPU for AI might be or a GPU for AI or for computing. So typically, we're not the direct recipient of some kind of tariff or fight between the U.S. and China, but we can certainly be indirectly impacted as can the broader economy by an ongoing trade war between the U.S. and China.
Got it. Okay. We talked a little bit about the PC market sort of being sort of softer in the fourth quarter. I think you would again sort of said that going into the quarter on your third quarter conference call feels like things have been a little bit softer as you look into 2025. What are your expectations for just overall PC demand, and do you see it being sort of second half weighted?
Yeah, we haven't really given guidance to 2025. I would say that typically we do expect the way the years typically run is they're seasonally weak in the beginning, certainly in the March quarter, and things typically build from there with December being our highest quarter from a revenue basis. I wouldn't expect, I think coming into 2025, we would expect it to play out like that. So the second half definitely should be seasonally stronger and therefore heavily weighted towards the second half from the first half.
Got it. And we'll get into it, but obviously lots of new products, you know, ramping in 2025. So I'm sure that that's going to drive a lot of the company-specific growth.
Yeah. In our case, we have multiple new products or product transitions that are going on in 2025 that are unique to Silicon Motion. Again, I mentioned PCIe 5 and the PC market at the high end of the PC market. We're entering that market for the first time. That has a 100% ASP increase over our PCIe 4. We've got a new product for mobile in UFS 4, and that's got a 50% ASP bump. We've got our entirely new enterprise-class product, MonTitan, which is expected to ramp in the second half. We've just started shipping that in Q4, but really don't expect our customers to ramp significantly until starting in the second half of this year. And that is a much more expensive product than any of our consumer products and significantly higher margins as well.
A few more just kind of on the sort of near-term, you know, kind of just industry conditions. To the extent PC business was a little weaker, do you think we exit 2024 with excess inventory of either PCs, you know, in the channel or excess inventory of controllers at your customers if end demand was a little softer than expected?
You know, we actually thought exiting Q3 and entering Q4 that the PC market and mobile markets both were actually the inventory levels were in line with end user demand. So we just saw, you know, slightly weaker demand than we were expecting in December. And, you know, hopefully that will start to take care of itself as we go through 2025. You know, a lot of consumer demand uncertainty can be directly impacted by the political unknowables, which are still in effect until the Trump administration takes office here in a couple of weeks. But hopefully things will get resolved quickly once that happens and we understand whether or how tariffs might emerge and, you know, let the industry digest that and move on.
Got it, so it's more sort of uncertainty around tariffs and geopolitics rather than it doesn't sound like you think the channel is sitting on a ton of inventory or customers are sitting on a lot of inventory.
No, we don't. We certainly didn't see that at the end of Q3, so I obviously will reserve my right to update that as we get closer to reporting earnings in the next few weeks and we go through what our sales teams are reporting in to management this week and next week as we get ready for earnings, but from my understanding, things aren't that different than they were at the end of last quarter.
Last sort of big picture question. NAND spot pricing has been, you know, pretty weak in the second half of 2024. I think it started to see some stabilization in December. But what's your outlook for NAND pricing and supply-demand equilibrium next year? And more importantly, how is SIMO affected by either rising or, you know, lower NAND prices?
Sure. So, you know, NAND prices have come down, particularly on the consumer side over the last, let's call it three to four months, and that's directly in line with the demand, the drop in demand. So we think that's likely going to be short term and can turn around pretty quickly. Enterprise demand has been much stronger thanks to AI. So we did see some small price declines on the enterprise side, but not significant. Maybe that's already stabilizing to your comment. So, you know, we'll have to take a better look at that over the next couple of weeks where that's going. But our CEO stated on the last call, we expect that NAND price equilibrium getting back to kind of a more normalized market will likely happen by mid-year.
Mid-year. Okay.
As far as where we're concerned, NAND pricing typically does not impact our business that much, even though we tend to trade with the NAND flash makers. We don't really share in their fortune when prices are going through the roof, and we don't normally suffer when their prices are going down because we're a unit volume-driven company, so all we care about is unit growth. We have seen a couple of times in the entire history of the company, 2008, 2009, during the meltdown, and then between 2022 and 2023, we saw a NAND flash meltdown with consumer electronics. Those are the only two times where we've actually seen our pricing, you know, impacted by a broader market correction.
Got it. That's more the exception than the norm.
Normally, the only change for us is that when prices come down, we would typically see a greater mix from our module maker customers, and when they go up a little bit too high, it makes it more difficult for our module makers' customers to compete, so we'll see a greater mix from the NAND flash makers.
Got it. Okay.
But end result from a revenue and unit perspective for us amongst that mix of customers is negligible.
Okay. You know, coming into some of the new products in the PC and the smartphone market, you talked about your new eight-channel PCIe Gen 5 controller. You've got design wins with four of the six leading NAND vendors. I think that product was supposed to go into production late last year. Did it sort of ramp as expected? And more importantly, how do you see the ramp through 2025 and 2026?
Yeah, so we started shipping to customers in Q4. Again, we have four of the flash, four of the six flash makers. We've got many, if not all of the, or nearly all of the module makers as customers for that product. Obviously, it's not going to just start shipping all at once in December, so it's got to ramp up. And so we would expect to see linear growth as we go throughout 2025. And eventually, we do believe that, you know, again, that's 10%-15% of the market that we haven't addressed before, and we believe we'll have 50% or more market share there.
Getting to that 50% market share, is that a year? Is that a couple of years? How long do you think it takes?
It's probably two years. So PCIe 5, so some high-end products in 2025 will still actually use PCIe 4. So not all of them will ship immediately to PCIe 5. So we expect PCIe 5 to be about 5% of the market in 2025. We do then expect that to grow to 20% plus in 2026 of the market.
Got it. And then you also have a four-channel PCIe Gen 5 controller, which I think you've said, you know, goes into production probably early 2026. That targets more mainstream applications, I assume. Talk about, you know, expectations for that product.
Yeah. So PCIe 5, again, in the early going first year, almost two years, is going to be really high-end focused. But we do have a PCIe 4 controller that comes out early next year that will target more of the kind of upper middle end of the market. So that's how you get to 20%-25% of the market.
So you're looking at four-channel and eight-channel to get to that sort of.
That's correct.
Okay, and I don't know if you've disclosed, but can you say how many NAND manufacturers you have wins now with a four-channel PCIe Gen 5 controller, or is that something that you guys haven't yet sort of talked about?
Yeah, we haven't really talked about that. I can say that we are engaged with, you know, most of our current customers, which includes all six and most of the module makers. So hopefully we'll have some positive announcements as we go throughout the year about design wins with each of those customers.
Got it. And I think you said for the eight-channel device, you see an ASP uplift of about 100%. I assume you still get some uplift on the PCIe Gen 4 channel controller, but, you know, what's kind of the blended ASP lift that you get going from Gen 4 to Gen 5? And perhaps more importantly, what kind of gross margin impact do you see?
We haven't really talked about the pricing of the Gen 4, I'm sorry, the four-channel products. I can't really discuss that, and I'm sorry, the second part of your question was?
Oh, just gross margin. ASPs are going up nicely. It sounds like at least with the eight-channel that you've talked about, I assume there's probably some gross margin lift there as well.
Yeah, so definitely some gross margin lift from both the four-channel and the eight-channel controllers, along with our UFS 4 product for mobile and along with MonTitan. So we would expect to see our gross margins continue to increase. We just announced for the fourth quarter that gross margins were around 47%. 48%-50% is our normal range that we'd like to get back to, and we do expect to get back there in 2025.
Perfect. And then on the UFS 4 side, can you update us on status of the engagements with customers and NAND manufacturers for UFS 4?
Yeah, so we have announced that we have one flash maker partner that we're going to market with already. We expect that that will, you know, we hope to increase that as we go throughout the year. We also have multiple module maker customers that we have design wins with that'll be entering the market here mid-year in 2026. We got a nice ASP uplift in UFS 4 over UFS 3 of approximately 50% there as well. So that should help, you know, our margins grow from 47% today to hopefully towards 50% as we go through 2025, but yeah, we have multiple partners there, including one flash maker that we've announced and expect to add more as we go through the year.
As you look at sort of penetration of UFS 4 in the market, where do you see penetration going, you know, for 2025, 2026 for that high end of the market?
Yeah. So UFS is right now really just the high-end flagship. So Samsung Galaxy S series, the Apple iPhone, obviously, and maybe one or two of the Chinese handset makers, and they're very high-end flagship phones. As we go through this year and into 2026, we would imagine that to kind of move down into the mid-range, the upper mid-range of the market. And that could be, you know, now you're going from maybe the top 20%-25% of the market, you're actually opening up another 25% to even 50% of the market that will start to use UFS 4 as we go through 2025 and 2026.
For those premium tier phones, are those typically using UFS controllers that are internally designed by the NAND manufacturers? Is that where they try to focus their controller efforts at the sort of highest end of the market?
Yeah. So when a lot of times one of the flash makers will develop their own in-house controllers, normally on a leading edge product. So they want to be the first to market. They're competing with each other to be the first, to be the fastest, to be the best. But by the time it hits the mass market, which is typically a year or two later, they're moving on to the next generation to try and beat each other at UFS 5 or 6 or 7. So we typically step in in the second generation of a market. So UFS 4, which has been around for a year and a half at the very high end, just the Samsung and the iPhone. Now as that enters the larger market, that's why we have timed our entry there. Same with PCIe 5, right?
When it shifts from just desktops and just early adopters to now the mass market, we're there with our PCIe 5 product. So, you know, again, the flash makers typically go for that first 15%-20% of the units, and we then capture everything that comes after.
The second wave, or as it moves more mainstream.
Yeah. Because, you know, if you look at one of the flash makers, let's say they came up with a UFS product a year and a half ago, the flash that they made at that time could handle 1.6 to, you know, maybe 2 megabits per second, sorry, excuse me, gigabits per second of speed. The flash that's made today goes up to 3.6 gigabits per second. So their controller that they made a couple of years ago is behind the times. And the cost of producing a new controller has gone, you know, from $3 million-$5 million in, say, a 12-nanometer controller to $15 million-$20 million in a 6-nanometer controller. So, you know, how many of these things do you want to build for every generation?
Right.
Normally, you're just going to build one and then you're going to move on to the next generation of product, and so that's why we tend to get, you know, much of the market, 30% of the PC market or more, and now at 50% with our PCIe 5, it's because we tend to time our product introductions for the biggest part of the market.
Got it. And then just can you give us an update on your efforts to partner with handset OEMs on the QLC NAND? I think you had one handset manufacturer moving to production late last year, but maybe just an update there.
Yeah. So that's a very exciting and new area for us. We typically sell into the mobile market through the flash makers, but we do have an OEM relationship that we announced last year for the development of QLC NAND. QLC being the latest type of NAND. It's four bits per cell. You can get a very high-density amount of memory for much lower cost than you can in TLC or SLC or MLC. So we've been developing a QLC product for a handset OEM directly. One of the advantages they get is that if they were to use our QLC controller, they can then go and pressure all of the flash makers to give them their flash at a lower price in any given quarter because our controller actually works with all of the flash makers' NAND, which actually no one else can say.
We're the only one that works with all six. So that's going very well. We're in trial right now. We believe that, or we've been told that if successful with this trial, they're likely to start shifting all of their handsets to our QLC controller moving forward. Because QLC, the cost is so much less for the NAND that it enables them to actually offer a phone that would have the same amount of memory as a flagship phone for half the price or a third of the price. And you just can't do that any other way.
Coming back just to sort of thoughts about refresh cycles, I think in the past you've talked about a PC replacement cycle potentially coming in the enterprise market in 2025. What are the sort of milestones you're looking for internally to, you know, sort of gain confidence that we may indeed see a refresh cycle?
Yeah. I mean, I think everybody was hopeful that we'd see a refresh cycle in 2024 coming into it. People were focused on AI at the edge and PCs and smartphones. And also with COVID, everybody went home and built home networks, bought new home computers, and had to work from home for a few years. And typically you'll see a refresh cycle from, you know, PCs and mobile phones at least every three to four years. So everybody went for COVID back home in 2021. There was a belief that 2024 we'd start to see the replacement rate. Obviously, that was not the case. So I think everybody's carrying that hope over into 2025. I do think that in 2025 there is yet another reason that does suggest that certainly by the second half of the year we will see a replacement cycle in the PC market.
And that's the sunsetting of Windows 10. There's a large segment of the enterprise market that really hasn't thought forward enough on that and is going to get caught, and they're going to have to upgrade their PCs as we go through this year. So we are hopeful, but, you know, I think over the next few weeks we'll have a more direct number as to what we think the PC market can grow. But certainly, typically we would think it'd be low single digits, but there's cause for hope that it could be more than that.
There might be. Okay. And then, you know, I know this idea of an AI refresh cycle or AI PC, AI smartphones, you know, folks may not be entirely convinced, but to the extent we do see an upgrade cycle to AI-enabled PCs and smartphones, you know, is there a benefit in terms of higher flash capacities, faster flash? I mean, is that a positive for the controller business if we get, you know, sort of an AI cycle in either of those trend margins?
Yeah. I mean, if you're looking at AI capability at the edge on the actual device, what that's going to need is higher computing power and higher amounts of memory, so higher density memory. So that really translates well into needing QLC, you know, at least TLC, because other technologies are way too expensive in a phone and likely a PC. But QLC is ideally suited for AI applications, particularly when you're just doing the reading of the AI or the inference, getting the answer, right? Not when you're training your AI, maybe then you want to use TLC. But QLC is improving there. But when you're just using like ChatGPT and you're typing in a question, QLC is ideally suited for that because it's so much less expensive than the other technologies.
And we have been working with QLC longer than anyone else, and we've been working with everybody's QLC, so we have a real advantage there. Excuse me, we think going in, you know, as AI does end up in edge devices, we think that will benefit us greatly.
Good. You know, one of the biggest new opportunities for the company.
I'm going to grab a water.
Is the company's introduction last year of the MonTitan solution for enterprise SSDs. You know, you secured two wins with tier one OEMs, one in U.S., one in China in Q1. You know, how do you see those designs progressing? I think there were some expectations for initial unit volumes for MonTitan by the end of the year. Did that, you know, is that on track?
Yeah. So, you know, MonTitan is the exciting product everybody wants to talk about because it's an entirely new market for us, the enterprise. And, you know, MonTitan is again leading technology in TLC and QLC. So the industry is shifting away from SLC and MLC to TLC and QLC. And we're there at just the right time with MonTitan. We did start shipping to our first two customers in December, so everything's on track there. We have expected and continue to expect that that will start to really ramp significantly or more significantly than now in the second half of 2025. And then 2026 and 2027 should see very strong growth becoming 5%-10% of our revenue by then. And then hopefully between now and the end of 2025, we can announce new customers to layer in.
Right. And the 5%-10% revenue in 2026, 2027, is that just from those first two design wins, or do you need additional design wins to kind of hit that range of 5%-10% of revenue?
Actually, our CEO on the last call explained that any one of our customers could be 5%-10% of revenue because these are all very big, very big customers, and the enterprise market is, you know, very large and growing. We all know what AI has been doing lately, so that's going to be a major driver of end demand in the enterprise, particularly for QLC, so we're very excited about our prospects here and very excited about MonTitan, how well it's being received so far in the market, and yeah, we expect to see great success as we move forward.
Good. I know you talked about expectations for perhaps some additional design wins to be announced by the end of 2024. You typically don't press release design wins. And so should investors look to the February 6th call as the most likely next update on sort of your status of customer engagements on MonTitan?
Yeah. We typically give our status updates on what we're doing in MonTitan on our earnings call. So I would think that that would be a better time now.
Yeah. I didn't think I was going to get you to pre-announce anything here in the fireside chat, and then, you know, for folks in the audience that may not be familiar with the pricing, just give us ballpark ASPs for MonTitan versus the client business because it's a pretty significant number.
Yeah. So if you're looking at a mobile device, you could be talking $1-$5 for a controller for a mobile memory. For a laptop or a computer, you're looking at anywhere from, you know, $3, sorry, $5-$7 if you're talking PCIe 4, $10-$15 with our newer PCIe 5 product. MonTitan is $40-$70. So it's, you know, it's four or five X our normal product in terms of ASP.
Yeah. I was going to say it doesn't take a lot of units to start from the beginning.
No, and it's higher, you know, it's higher than typically higher than our gross margins for our normal portfolio as well. So it's a big opportunity for us, both top and bottom line.
You know, this gets into maybe a little technical, but are there differentiating features of MonTitan that you guys have built into the product that differentiate the solution from other SSD controllers, enterprise SSD controllers on the market?
Yeah, again, you know, QLC is very new. It's something that everybody wants in the enterprise. A lot of people have been talking about, you know. I look back in 2018, one of the flash makers was talking about having QLC. The reality is it's not really in the enterprise market today, particularly not in AI. We've been working with QLC for over a decade, so we know how this works. Our product is now ready for this market. As it transitions to QLC, we think we are in a very strong position competitively. A couple of things just by example that we do very well with our algorithms relative to the competition is we can deliver lower latency, higher speeds.
We also have better error correction than most of our competitors, which is extremely important, especially when you're talking about QLC, where errors tend to be the biggest inhibitor of people choosing QLC over, say, TLC or another technology. We also have in particular one technology called PerformaShape that allows you to dial the performance of the server to favor either writing with fewer errors or speed. Most other of our competitors, you actually have to flash that server once for either. Then if you wanted to change it, you'd have to take it offline and reflash it. Ours, it's literally an adjustment.
Yeah. I was going to ask Micron on its call, you know, started to talk about enterprise SSDs replacing nearline HDDs. I know that's always been sort of a thought, but it feels like that may be coming, you know, a little bit more of a reality over the next couple of years. I assume that QLC or having a good QLC, you know, solution makes that ability to replace nearline HDDs probably that much easier. And so to the extent that happens, sounds like QLC adoption would be, you know, good for you guys.
Yeah. I mean, we sometimes call our positioning nearline SSD to compete with nearline HDD. But the reality is we expect HDDs are going to have a long and happy life. There's always going to be a place for them because, and what's good for them is good for us. There's a growing, the data demands are, you know, the growth is unbelievable.
That's slowing down.
And it doesn't slow down. And, you know, so being involved in memory is obviously very, it's great for all of us until, you know, we just sometimes have bad moments. But over time, we do expect that we're going to find a nice place within the enterprise server market, the AI server market. But we do think that HDDs will continue to exist alongside.
Okay. Moving to the auto market, I know it's not as large an end market today as PCs and smartphones, but I think you've got an auto-grade certified PCIe Gen 4 controller. I think today that accounts for about 5% of revenue with expectations that auto could be 10% of revenue by 2027. You maybe talk about some of the growth drivers in that automotive business.
Yeah. So, I mean, the growth drivers in automotive are ADAS systems, infotainment systems. We've been building up a, you know, a group of wins over the last few years that are just now starting to ship. So while the automotive market was a little bit rough and tumble over the last year, for us, it's almost a greenfield opportunity. So we've been growing across all our products, eMMC, UFS, PCIe. You mentioned the PCIe 4, ASPICE 2 certified product. We also sell our Ferri products. And so actually automotive is the one market where we sell across every single one of our products.
Oh, so it's not just SSD, it's across everything.
It's across everything. It's generally higher ASP and higher margin because, you know, you have to have a much more robust solution when you're going into an automotive environment that can handle higher temperatures and bouncing up and down and, you know, all the rest of it. So it's a very good market, and we've been collecting a number of wins, and we expect to collect, you know, more wins as we go forward, given that we've been stepping up, you know, our abilities here. And because it's a little bit more complicated and has to be a little bit more robust, there's fewer competitors in this market, so that's why we think we'll get up to 10% of our revenue by 2026, 2027.
Some of the same competitors you see in the consumer markets, or is it mostly just the NAND manufacturers with internal controllers going after the automotive?
It's more the NAND manufacturers with internal capabilities than a lot of the third-party merchant suppliers of controllers just really don't have either the know-how or the, you know, the product depth to be able to attack this market.
Right. Okay. And then on the sort of related to auto, you know, your thoughts on the industrial opportunity. You've talked about industrial helping to diversify your exposure. You know, what kind of applications within the industrial markets do you see being the biggest opportunities?
Yeah. I mean, in terms of biggest opportunities, we serve a lot of industrial markets, so it's kind of difficult. Some of them blend into other markets. I think, you know, robotics and manufacturing automation, you know, these are only going to continue to grow. If you believe that there's going to be a Tesla robot in everybody's house, then it's probably going to have memory. So, you know, if there's a billion of those in the world, that would be great for us. We just, you know, you don't really know what the applications are going to be. They tend to be smaller than, say, a PC market or a mobile market, but more varied. So there's a lot of them that add up.
Yeah. I think I saw something coming out of CES that Elon's talking about: every person on the planet may have three robots at some point, so there could be like 20 billion robots. So it could be a big opportunity.
Yeah. Well, I mean, it's certainly good. We'll see how quickly it comes about.
I don't know if I'd want to bet against Elon.
No, that's true.
You know, I guess, you know, sort of another question we get, you know, especially from folks that are perhaps new to the name, you know, consistent concern from investors has been, you know, your NAND customers looking at insourcing versus buying merchant. And, you know, does that mix, you know, is it shifting to more in-house? Is it shifting to more outsourced? Just your thoughts on, you know, the competitive landscape from the internal development at the NAND manufacturing companies.
Yeah. I mean, we've seen really a continual shift. It's been interrupted for various reasons along the way over the last 20 years. A good example being during the MaxLinear acquisition, you know, some people kind of put on pause the outsourcing because they didn't know what would happen. But on balance, if you look over the last 20 years, we've seen outsourcing increase kind of in a, at least in an up and to the right. We think that will continue, and we think actually there are even more accelerated drivers right now. And what that is, is the cost. I mean, taping out these controllers is going up exponentially. The last generation for NAND flash controllers was 12 nanometers. It was $3 million-$5 million. Not that much for a Micron or a Western Digital or a Kioxia to absorb.
Now it's $15 million-$20 million at 6 nanometer. Arguably still not that much. But that's just for the tape-out. That doesn't include the R&D. It doesn't include a number of other costs that go into developing each one of these new controllers. So we certainly, with our discussions with our NAND flash controller customers, it does appear that they want to continually outsource more and more to us as we go along.
We're getting maybe a minute left, so I'll just see if there are any questions from the audience.