Go ahead and get started. Welcome, everybody. Thank you for joining us on the second day of Needham's 26th Annual Growth Conference. My name is Quinn Bolton. I'm the semiconductor analyst for Needham. It's my pleasure to host this fireside chat with SMART Global Holdings. The company designs, builds, deploys, and manages high-performance and high-reliability enterprise solutions across the company's computing, memory, and LED business lines. The company serves its customers by leveraging its deep technical expertise, custom design engineering, build-to-order flexibility, and best-in-class quality. Joining me from the company today are Mark Adams, President and CEO; Ken Rizvi, SVP and CFO; and we have Suzanne Schmidt from Investor Relations in the audience. Mark, Ken, Suzanne, thank you for joining us.
Thank you.
Mark, before we get into, you know, questions on the company, was just wondering, you know, I know you've been with the company for a while, but just talk about your background and what attracted you to SMART Global. What opportunity did you see with the company, compared to your prior role at Micron?
Yeah, great. Thank you. So yeah, I grew up in technology broadly. I was at NCR early in my career. Then I went to business school, and I came out, and I opened Latin America for one of the hotter startups in the NASDAQ IPO swing in the 1990s, a company called Creative T echnology. For those of you who are computer enthusiasts, this is a company that made Sound Blaster, and I was there for eight years. And then I was recruited to be the COO at Lexar Media, which is a flash memory company that was bought by Micron. That's how I kind of got into the Micron world, and when I got out to Micron, I was the one, there was 18 officers, big company, and I was the one guy who was the marketing person.
Then there was 18 people, 17 people from engineering fabs and what have you. And I actually found it funny at Micron when I was in the growing there, it was like, I would say something that I thought was just like a softball, and the engineers were like. And they talked 80% of the time about stuff that's going on in the factory I had no idea on, so I got to know the business through that way. It was great, great run there for 10 years. That's where I met Ken. And then I had a heart problem. I was named president, and then I think you all know what happened to Steve Appleton. He passed in a plane crash. That was kind of an interesting kind of time in our evolution at Micron.
I was about to be named CEO, and I had a heart problem that took me off the field for two years. So then I joined some boards. I'm on the board of Cadence Design today. I was also on the board of Seagate. I had to step off 'cause of the over-boarding thing with the kind of what you're allowed to do as a public company CEO. So in the midst of all that, I've known Silver Lake for years, and I thought I was gonna just stay retired. They called me and said, "Hey, could you help consult on an acquisition?" For those of you who know private equity, that's the old end around, like, "Let's sucker you into an advisory role, and then you're gonna become a CEO.
You just don't know it yet." And so that's exactly what happened. The thing for me that I liked about SMART, at the time, was, like Micron, very few companies in Silicon Valley, the company still is the first thing. The team's the first thing, and that's what I thrive on. I grew up around athletics, played athletics in college, and the team's everything, right? And so, you know, in Silicon Valley, you don't see that as much, or you see some of it, but not a lot of it, and the business these guys are in, they really work hard, and they're loyal, and people stay at the company. They don't lose a lot of people.
It's a culture thing, and, you know, I'm at a point in my career that I wanna be around people I care about and wanna work with. It's not something, y ou know, I'm not trying to just force myself into a situation. So the markets they do, the value add they provide, it was a platform to think about how we could do something with this company at the time. And I would just say, before I hand it back over to you, that, you know, what I learned along the way through a bunch of great leaders, you know, from other folks, is that, you know, as you know, strategies evolve.
When I joined this company back in 2020, we were 76% memory, and the strategy was, "Let's just go be a holding company." That was what I was sold on early, and I did a perception study. Actually, Ken got me to do it and said, "We should do this just to kind of level-check on what the world thinks about SMART Global Holdings." And it was the best money I've ever spent. But more than that, it wasn't that it was, like, earth-shattering in terms of the conclusions, but they came back and said four things. "One, please get out of Brazil. Two, diversify and keep getting away from memory. It's just too cyclical for the investor community to really value this company properly. Three, we don't need a holding company model.
You know, the world does not need another holding company, so we can diversify ourselves." So really, that led to four, which is, "You need a better strategy. You don't have a strategy." And that's what I walked into, and so, like, it was just really refreshing to hear from our shareholders what they thought about the company, and they, they'd said some nice things, but this is what we were walking into. And so that's the catalyst, is that I knew I was coming into a bit of a turnaround, and that's really been the catalyst to what we're trying to build here today. Sorry for the long-winded answer, but-
No, that's-
Just to give you some context.
That's great. Obviously, IPS is now an important part of the business, and, you know, we're seeing, you know, results. You just announced last week, some momentum, you know, in that business. Maybe since you did have results just last week, you know, just quickly touch on the outlook, you know, quarterly results and the outlook for folks who may have missed that.
Yeah. So the, well, we're in a good space, obviously, and so the evolution, I said we were 76% memory three years ago. Today, I think, the last quarter, we're in the low 30s in memory, and now compute's the biggest part of our business. Over the last three years, we've kind of have driven a new strategy for the company. We're more of a traditional model than we are a holding company, although there's remnants of that, and we're trying to tease our way out of that. The IPS world is really interesting for us because, I believe that 2023 was the year of GPU sales.
And what I kid around about with people in the market is, as much as 2023 was GPU sales, 2024 and 2025 is gonna be the year that we— people have to use them and figure out how to deploy AI. And we're not a IP technology systems company. We do that. We integrate well. That's who we are. But our core competency is 25 years in HPC, and a lead because of HPC, a lead into the AI world. Our core competency is on design, software, and managed services. The hardware is kind of like a you know, a means to an end for us. W e know it well. We've been installing systems for two and a half decades, but what we're really good at is helping people deploy these systems. Now, a data point for you.
Last spring, I believe it was, Meta, who is our largest customer, they issued a press release. I wish they would have told us, but they didn't. They issued a press release, and they said, through and we were brought in to build out the largest AI infrastructure at the time ever deployed, 16,000 GPUs, and we were brought in because their in-house team couldn't make it work, and we kind of came in. That's what led to a really good first half of last year. But the key point is, they issued a press release and said, "Through our integration partner, Penguin, we've achieved 95% uptime in our AI systems." The industry standard right now for AI systems is somewhere in the sixties, and it's that know-how that's really hard to copy.
I mean, it's really hard to just take two and a half decades of a culture that's been deploying these systems. And by the way, it's not as much as what we're good at. Sure, it's also what we know is gonna be a problem that we go into conversations with customers, we tell them stuff is gonna happen before it happens. And as part of that, that's where we're more of a trusted advisor. So if you look at my margin model in that business, we don't really break it out, but it's kind of in the zip code of twice of what HP, Dell, and Supermicro is. It's 'cause we're a solutions company, we're not that. So the outlook for us is generally very positive. Now, we're smaller. I mean, you know that.
And we grew at 30% a year for three years in a row. Well, you know, pretty much 30% a year for three years in a row. People always ask me, they goes: "You know, what are you thinking about? W hat was the hardest part about it? What do you know-- w hat are the things you'd, you'd like to go back and think about, and how would you do it differently?" And I would just say that when you grow 30% a year, like when we walked in, this was a declining asset, it's SGH. It was down. When-- They bought it at $200 million, it was down about $177 million or something like that.
They just weren't doing anything with it, and the team wanted to be invested in. It just wasn't happening for other reasons. So, when you grow that fast, things, there's trade-offs. And quite honestly, when we got a big order, like 16,000 GPUs, you know, we were using bubblegum and Scotch tape just to make, just do a good job on it. I mean, it was just because it was a massive installation for a small company. And so, you know, we've, we've grown a lot in terms of the sophistication of the company, but we're still in process in that part of our business. SMART Modular, super mature. Cree, super mature, but the Penguin business is kind of a entrepreneurial spirit, and so the go-to-market thing for us was more of a 2023 focus.
And so we've hired new field teams to go to market and talk to enterprise customers. We've got a new head of sales that just joined us in the last 30 days. And so the engagements we're having are fantastic because, as I said, 2024 and 2025 companies are now figuring out, we don't have the capabilities in-house. Now, some people do. Yeah, Microsoft's got, t hey can figure it out themselves. They're in this business. They're in data center business, cloud, you know. But even, by the way, Meta, you know, they, they're in the data center business, and they couldn't get it done. So when you look at more commercial enterprises in the verticals, the financial institutions here in New York, or oil and gas, or the healthcare, or education, these folks don't have the depth.
I mean, it's not. We're not talking about buying a computer and plugging it in. We're talking about from the grid to a building, inside the building, design architecture for the most efficient layout, cooling systems that keep the computers functional and up, networking, software, which is a big key for us. We have a platform called Scyld, and Scyld Cluster Management really is the workload optimization and provisioning part of all these jobs that go on in this big architecture and how it runs most efficiently and, reliably. And then the managed services, of course, is which, which is what we are on site to do, which is really keep the site up and running and, then high performance. So the market tailwinds are behind us are good. They're very favorable. We're growing. We're adding resources.
It's a fun time to be in the business, but like HPC, this is a choppy business. We get big orders, and then we go quiet a little bit, which is nice, as we have these other businesses like Stratus and like Smart Modular, that are kind of like levering, you know, smoothing out a little bit, teasing out the volatility. But we'll have volatility. We'll get big projects, and then we'll, you know, have hardware deals and then software deals. It just. That's the way the business is, but in general, we are in a good space. Last quarter, we just announced, and remember this now, in the world, in the bottom of memory market, you saw the results from Micron and the likes.
In the bottom of memory market and a very similarly, you know, tough LED environment, we just have set record gross margins, and we have the most cash we've ever had in the company on the balance sheet, and just great customer engagement. So I'm very bullish on the business. And the other thing is, I would just say, Ken joined me, early on, like six months into my being at the company, Ken joined me. And I would just ask you from a, just, if you look at leadership and teams and whatever, go back and look at our time at Micron. We had I was there for 10 years, 40 earnings calls, plus or minus. We're gonna tell you what is going on. We're gonna be honest, like we were in the first quarter.
We told you what's gonna happen, and we're gonna be very honest about the business, 'cause we're building something much bigger than one quarter at a time. Now, unfortunately, in public companies, capital markets, that's what we've got to do. But we're building something much bigger, and we're excited about it. You know, our last call was generally very favorable, and we think there's, you know, kind of a good path going forward. So we're excited.
Yeah, we came away from earnings last week certainly thinking that the revenue trends had mostly bottomed up. It guided IPS up 15% sequentially. I know there's a little bit of seasonality in the LED business.
Yeah, right.
But generally, it kind of feels like, you know, momentum coming back into the IPS business. You know, as you look to the second quarter of the 15% growth, but then also into the second half, what are some of the growth drivers in the IPS business? I think near term, it's coming mostly from existing customers, maybe a few proof of concepts. But, you know, as you look into the second half, you know, what are the opportunities?
Yeah, I mean, it's really a couple things. For those of you who follow the company carefully, I made an organization change in the middle of last year, July first, and I had to do it. I made a mistake, and so I had to fix it. So now, rather than bring in someone right now, I wanted to get much closer to the business, 'cause I was brought in on this holding company model, and I kind of let it run like that for a while, and it didn't sit well with me. 'Cause anyone who runs a business, you've got to know the business. And having it to go through a leader, and the leader's not performing, it kind of just. So we changed it.
We're changing the structure of the company to be a more traditional company, like a traditional structure. I didn't even hire a replacement 'cause I wanted to be this person and go make sure I understood the business. We've gone out, and I've talked to, I don't know, indirectly or directly, meaning email or virtual or whatever, but more like I've had 30 or 40 meetings in person in the last 90 days. And I've also kind of done virtual meetings on top of that, and the number is 71, 'cause I've tracked them. I've got 71 meetings with CEOs or comparable level IT executives at companies who are running AI. We're not gonna win them all, obviously. I'm not saying that, but there are a subset of these customers that need this trusted advisor because they don't have the capabilities in-house.
Just that simple. And for me, it's what's nice is we're technology agnostic. I don't have to sell the hardware. A lot of people want that 'cause they want one throat to choke, and they don't wanna have... But I actually got called into a hyperscale company, a Tier Two hyperscale company that's well-funded, a billion-dollar fund, you know, capital raise, and they fired HP and Dell. And he calls me, and the CEO called me 'cause he got a reference from Meta, and he said: "Hey, I want you to look, I need to meet with you." And I was like: "Well, great. Like, we're coming up the holidays. I'll meet you in January anytime." He goes: "No, no, no, you don't understand. We're in the middle of this, and it's a mess.
It's not working." There's that type of environment out there for what we do well, and as I said, you can't bottle 25 years of deployment expertise and failure, stuff we learned. And so that's a real differentiator for us. And the other thing is we won't sell hardware for hardware's sake. We will not do that. Right now, we have a control in the company. It's relatively new in the last 6 months, but you cannot bid and deliver a proposal as a sales team unless you have software and services, and it just can't happen. So that's kind of the mindset we have, and it's kind of a, it works.
The nice part about it is, it's not so daunting for big enterprises to trust us in that role versus, you know, that we're gonna do $5 billion of hardware when we're $1 billion in revenue. That's a little bit tougher, but this is a, this is a process and a strategy that's aligned pretty well.
Yeah, and Quinn, I mean, just if you look at the to add on to what Mark just said, if you look at the business, right, we, we do sell hardware, that's part of it, but the services piece is a large component. We did $68 million in services last quarter. We did about $250 million in services last year, and so that is part of the strategy, is to not only deliver on an overall hardware solution but provide the services and the software and the solutions to make that run.
The perspective that Ken just shared with you, now think about it, because those services, you know, vast, vast majority are in the IPS business. Services are running about, I don't know, 50%, 45%?
55%
55% of the overall, the margin, but I was thinking about the mix of the revenue. It's like almost half of the revenue. Almost?
Yeah, that's right, for IPS.
You can do the math. It's like $250 million on, you know, so it's a little bit less than half, half of the revenue. That's not true anywhere. I mean, it's a good business. It's a good business model, and we're gonna stay disciplined.
Just kind of, I guess, maybe summarizing it, 2023 was sort of the, the year of hyperscaler AI deployments-
Yeah. Yeah.
So, like, 2024 and 2025 are enterprise with Tier 2 CSP. That sounds like that's probably a nice tailwind for the IPS business.
No doubt. Now, the interesting part in the market, the broader market is, there is a staggering, staggering shortage right now of energized, power-driven data centers that can handle the requirements for artificial intelligence deployments. Staggering shortage, and if you don't, you know, if you don't know that and you want to test it, go talk to any of the big enterprises you're connected to. You can't find them. So from our perspective, the value add for us is how do we help customers bridge that gap and build a platform for the future in a more total solution environment? And I think we're in a really good position that way to manage it for our customers.
I know the focus, you know, does tend to be on some of the smaller, or not smaller, but enterprises, Tier 2 service providers, but you've mentioned the Meta 1,000 GPU deployment. That was obviously a big, big driver of revenue in, I guess, late 2022, or calendar 2022. What's your outlook, you know, for that engagement going forward? Do you see further deployments? What can you tell us about that?
So we don't necessarily forecast. I'm gonna answer your question. We don't necessarily forecast any of our customer stuff. We don't speak for our customers. It's just something I just won't do. Having said that, they're gonna be, you know, our biggest customer in this business again in 2024, likely , 'cause I have my CFO here, I can't say definitely. But what I would tell you is they're gonna be a great customer for us. We've got a great relationship. There will be things they ask us to do that are difficult, and they're gonna do as much as they can in-house 'cause that's the right thing to do for their business.
But we're still, you know, a lot of these managed services are things we're still running data centers for them, and that looks like it will continue for a while, and that's just the, you know, the position we're in is really good. The good part for us, I would say, is if you take, if you look at that trend line of that customer revenue, it's great, profitable, and very big, but it won't be as big as in 2024, 2023, likely. But the rest of the business is growing, and that's a good sign. If you break it apart, you say, "Oh, this is actually working." We're having other customers make up for some of that. Not all of it, but some of the loss.
And so what I'm speaking to is the Q1 guide was down because we didn't have two major deployments like we did in 2023, and so we're filling that up. If you break it apart, the rest of, you know, the, like, the non-Meta part of our business is actually growing pretty nicely, and the Meta piece is actually flattening out from a revenue standpoint. Still very profitable because it's services and software. So what I would say is, they're gonna be a good customer for a while.
Yeah, that sounds like a high-margin customer.
Yeah, it's a good partner, and by the way, they're a leading edge. I mean, they may have the best use case model for AI of anyone I've seen. We've got an oil and gas customer that's pretty interesting, too, around AI, but this Meta on the ad revenue side for Instagram and Facebook, pretty good. I mean, they, they're driving, like, billions of revenue. It's not trivial, and that's why they're investing so much in it, 'cause that's what they can generate. So it's a good use case model of AI, and they've pushed the envelope, and we've been there with them, pushing the envelope, and it's gonna be like that on a go-forward basis.
You'd mentioned that non-Meta portion of IPS. What kind of growth rate do you think that might achieve over the next three to five years? Or won't necessarily hold you to it, but it sounds like-
Yeah, no, I think, I think it's, it's, it's... Well, certainly, I, w hat's the market, like, 15%?
Yeah, market's right around 15% .
And we, we've grown 30% of it for the last three years. I don't know that we'll hit 30%, but we're, you know, we're, we would like to outgrow the market over the next three to five years, and as I said, it gets lumpy. We say that every call because it is, but I think in our positioning, I think we're in a really good place to add that value as it converts away from the ultra-scale, hyperscale thing to more of the traditional enterprises. Those are the people who don't really have the infrastructure to do what we do, where, again, a Microsoft or a Google or an Amazon, they probably have the capabilities, more or less, to do what we're doing, and even then, they don't, you know, not fully.
So it's kinda the tailwinds are lining up well for us.
Great. Last question for me on IPS. We've gotten a couple of questions from investors about the press release you put out in September about your DGX- Ready Managed Services partner with NVIDIA.
Yep.
How, how important is that? Maybe describe that program.
Well, it's a validation of everything I said this morning about our differentiation. So when you think about managed services, that's not hardware sales, that's not computer system sales or storage sales. That is, "Hey, we know this company is capable of deploying our technology in complex environments," and you get certified for that. And also, I don't know, as of today in 2024, but up and through most of 2023, we had deployed the most GPUs of NVIDIA of anyone in the world. I think it probably I d on't know, because of the sales to the channel people, what have you, I haven't tracked it in the last 90 days, but we're a big partner in that way. But what they love about us is we're not just kind of selling hardware, hardware sideways.
We actually make the stuff work, which actually is an enhancement model for them. You know, their CEO. By the way, it's an amazing story. I actually knew NVIDIA when I started 25 years ago. I sold, I think I sold one of their first graphics cards when I was at Creative Labs when... And so there, these are graphics chips that are AI chips. Somehow, that evolution, the nature of processing architecture of GPUs is just fascinating for AI. But this guy's. It's a great story, and Jensen's an amazing leader, what have you, and he's got wild ambitions to be kinda vertically integrated, but he's not gonna get 100% of the market, and that's where we're a great partner for them. Today, they're not really ready for that.
He's got a vision, but they're not ready to be a deployment company, whatever. That's why they have this program, and our success in this program is because they know we're good, and they know what we do.
Yeah, and Quinn, the other thing is, if you look at that IPS business, and you look at part of our strategy, it is to be neutral or Switzerland to some extent, because what we're trying to do for our customers is develop the best of breed solutions. So that means the best compute, the best storage, the best networking to create that overall architecture. And so, you know, part of that, most recently, we talked about immersion cooling, because as you think about the AI factories of the future, power and dissipation of heat is gonna be a critical factor in terms of these data centers. So investing in whether it's liquid cooling solutions and investing and partnering with folks on the immersion cooling side, is also an important component going forward.
Partnering with the best compute, best storage, the best networking providers, and then developing that overall architecture is gonna be critical to our success, so-
A real live example, one of our customers is a massive conglomerate in energy, oil, and gas... and for a variety of reasons, some of it was social ESG-type matter. They wanted to use recycled oil to do oil-based immersion cooling. I mean, when they came to us, we were like, "Huh!" And we actually deployed that last year. It's in production today in their data centers, oil-based immersion cooling. So it's cold, recycled oil, and you drop the computers into this, and that keeps the temperature of the computers functional for maximum uptime. This is kind of the stuff we do. You know, like, you wouldn't get that done for 15 different reasons very quickly at some of the bigger companies. But they know us, we're a big partner of theirs in AI.
We talked about it, we figured it out, we tested it in our labs in California. They liked it, and we rolled it out. And that's the kind of intimacy we have and relationship we have with our customers.
Now, I want to pivot for a few minutes to the memory business. I know that business has been, you know, sort of caught, like many semiconductor companies-
Sure.
In a pretty big inventory digestion period. You guided that business flat for next quarter, I think, as a result of some of that inventory correction. But, you know, just where do you think we are in the inventory correction process? How long does it take?
Yeah, sure. So again, background, I spent 10 years at Micron, so by de facto standards, I've been impacted psychologically by being in the memory business, and I highly say, don't do it. But I would say one thing: cycles are predictable in terms of the steps, it's the timing that changes. And when you go into a supply and demand imbalance in memory, it normally starts in the consumer side, because that's where all the consumption of memory is. All the good stuff is about 30% of memory. The 70% is phones, cars, laptops, the likes. And then, when you talk about specialty memory, which is what we're in, we're in the higher value segment of memory. When you talk about that segment of memory, it's a little less elastic and bumpy, but it also kind of trails the market on both ends.
So what happens in a memory cycle is demand gets shocked. Okay, so what happened is, in this cycle was, everyone bought five computers for themselves during COVID, and then they stopped buying after COVID. So the post-COVID dynamic in memory was severe. This is the worst cycle in memory, dollar-wise, ever, and I've seen them. 2008 was bad. This was pretty bad. So what the first step one is, manufacturers, and I was I did this multiple times, you just cut CapEx, so future supply starts to normalize, and then you hope to cause some... Okay. If it's a medium to bad cycle, you that's not gonna be enough, and then what happens is, you take production offline. Because you-
Well, the game in the memory space is, do you, you wanna keep revenue ASP above your marginal cost because you've already got the fixed cost of running a factory. So you do first, reduce CapEx, then you take production off. And that cycle goes on for a while. The third step in a cycle is the memory guys are bleeding. Micron was losing $2 billion a quarter for a while. Great company, but just that was the dynamics of the market. The third step in the cycle is, we've done those things, we can't do anymore, we're gonna raise pricing, and it's kind of like a poker game. Well, what happens is, the buyers are smart. They always were.
The enterprises and the wholesale scale, they all know when it's when these people are bleeding $2 billion a quarter, you know, and 40% negative gross margins, they know it's getting there. So what they do is they start to hoard inventory as much as they can, within reason, on their balance sheet. They put that. And then that becomes a poker game for six to 12 months, and we're in that game right now, which is like: "Hey, we're raising pricing." "Okay, I got some stuff. I'm good." And that goes on back and forth. We're in that phase right now. I believe, and I'm not you know, I wouldn't bet my life on it because this is a timing issue.
I believe right now in memory, we're probably springtime to early summer, that we'll start to notice that the financials are better in the broad memory market. Now, for us, let me just clarify something. We're not exposed, we're exposed in the top line to revenue, but we have a fixed value add component of our business. Everyone knows the memory pricings, and that we virtually wide, you know, open book it, but we sell a subsystem with our value add on it. So ironically, when this goes down in price because memory's down in pricing, our gross margins go up, 'cause our value add's a higher percentage of the bond. So yes, we get hit on the top line because the demand equation piece, but in Q4, for example, I'll give you a Q3...
Q1 number in a second, but Q4 2023, in the midst of this abysmal memory market, 14% operating income for specialty memory. Staggering. I mean, everyone's losing money, like bleeding. 14% operating income because of our value add. And so that's the way we play the market. I mean, that's the way our business operates in the market, which makes it a pretty good, you know, stable business, offsets the things like high-performance compute lumpiness and the like. So I think we're getting to the end of the cycle. I think three to six months is fair. I think it's where I would pick it right now. We're starting to see signs. Yes, pricing is going up, but remember, that's kind of data out in the market, but you got this inventory dynamic as the last stage.
So I think we're probably a little bit away from getting back where... And then what happens is, it goes up, it goes up, and then it goes, "Psst!" 'Cause what's gonna happen, my prediction is, the OEMs are gonna not have inventory for next fall, and then they start to game that, and then the pricing goes up because they know they're trying to get as much as they can early, and that's what triggers the pricing, then it starts. So that's why I think it's gonna be earliest, maybe late spring, early summer, that we'll start to see some benefits, and then I think we'll be back into it. And that's assuming the world stays rational in the macro.
Yeah, but, you know, for your business, you'll benefit from the higher memory pricing on the top line, but margins, you kind of think of it more of as like a fixed cost plus model.
Right. So the margins, but our margins, we're only talking about one or two point swings. What I'm saying is, the benefit on the recovery piece is units go up. So gross margin dollars, nominal dollars will go up. Our gross margin percentage will probably go down a point or two.
Yeah, I mean, the margins, the way we think about it is that operating income percent, right? Last quarter, Q1, we did close to 9%, but this business was running kind of low-to-mid-teens here if you look back over the last couple of years. The other thing, when we look at this business, it peaked at about $152 million-$153 million in terms of revenue. You know, we guided or we did $86 million in Q1, so and we have room to get back up to those levels. It will be a combination, as Mark highlighted, ASPs, but also just unit demand getting back to a more normal trend over the next 12-18 months.
Beyond just the recovery part of our comeback is specialty memory broadly is going to do quite well in the next three to five years. Why is that? Well, if anyone's, like, you know, close to the engineering side of compute, and I'm not, but I know this just because I'm around it, the biggest challenge right now for GPUs and CPUs and AI in general is, it's not the compute, it's the memory bandwidth. There's not enough of a pipe to allow memory to interact with the GPUs and CPUs at the right speed, so memory is the lowest common denominator. So we've talked about on calls a technology called CXL, which is basically a protocol standard that's going to allow you to do more across the pipe.
And then actually in version 3.0 of CXL, they're going to allow for memory pooling, so you can have multiple hosts into a memory data bank, which today it's kind of just think of it like peer-to-peer, one-to-one. In the future, you'll have this capability to have memory sharing, which will dramatically help on the economics and also the throughput. We've invested in some new technologies that, with CXL on top of it, optics and fiber between the CPU and GPU and the memory. We're going to be out investing and growing this business because it's right, it's what we do, is like we're a specialty memory company. So beyond the pricing and the demand and the unit stuff, I think the category is going to grow, the one we're in, because of this dynamic of really driving computer performance and optimization.
Is that CXL story more of a 2025-2026 story?
Yeah, yeah, that's right. I mean, just we're in pilots right now with major customers, but I think from a volume hitting my P&L, I think it's probably more fiscal 2025, 2026.
Yeah, we need the, the processors, right, the Emerald Rapids CXL.
Correct.
Correct, yeah. And DDR5 is required, and DDR5 has just been slowed because of how bad the memory cycle is. So yeah, I think we've said that before. Like we're in pilots and tests, but in order to get meaningful revenue and the like, so we're still, you know, I think it's kind of mid- to late-2025, early-2026.
Yeah. Just, you know, maybe a minute or two on the LED business. You know, I think you've done a good job of reducing channel inventory in that business, but it's still, you know, a little, well, seasonal, I guess, this quarter.
Yeah.
What are the biggest drivers of that LED business? Is it mostly just macro-driven, or are there opportunities that, you know, drive it better than sort of macro growth rate in that business?
Yeah, I think it's important for everyone to know we're not in the broad LED, you know, what they call Christmas lights and flashlight business per se. We're in specialty applications like emergency vehicles, like scoreboards, like architectural lighting, like horticulture, stadium lighting, and the likes, and so it's very custom, and there's technology that's not very creative in name, but high power, mid power, and low power. We only play in high power because we have IP that protects that, and Cree's, the IP leader, 1600 patents and cross-licensed with other people. But our model is very good. What caused the challenges in this market wasn't consumer as much as for us, is the lack of growth in China was kind of a big hit to the business.
They literally shut off production and investments into new builds and what have you, and so that business has been challenged. Now, what's nice is, in an industry structure that's pretty fragmented, we have the best model and the best margins, and so we're expecting to, I expect us to take share over the next three to five years in this business. It's a great business. Like I said, the best gross margins in the business. We have a factory light model, where we're optimizing, like think about it as a foundry model. We design everything, but we use the excess capacity, of which there is massive excess capacity due to China overinvesting, that we have access to capacity for virtually great pricing, better than anyone who owns a fab, and I think we'll get some share.
And I also think some of the markets we're in are just nascent, starting to get bigger, like LEDs in horticulture is a brand-new model. We're a market leader in horticulture. So if you think about some of these things, I think it's new builds, construction coming back on more broadly in the global economy, as well as some of these verticals that'll help drive some growth in the future. We've had, by the way, you're right, we've guided down slightly this quarter because of seasonality, because there's three holidays in our LED business that kind of impact the demand there. But prior to that, we've had three successive quarters of growth in the bottom, incremental growth. So we're feeling pretty good about the business in the general direction.
It sounds like, China's an important region for the, for that business. I know China, many companies have said, demand is still pretty slow. Have you seen any green shoots in, just the broader-
I think we're starting to see, for the LED business, some sampling activity over, w e've talked about this in the last quarter and even into this quarter. Typically, that is a precursor to revenue starting to grow. It usually takes about six months or so for us to translate the sampling into order activity. This business goes to a broad set of customers. If you look from a customer exposure standpoint, you know, we have a few customers that may be $10s of millions, but no significant customer like our other businesses.
I think, yeah.
That's broadly through the distribution channel and through thousands and thousands of customers.
I think one thing, other thing to note really quickly here is the industry structure in LED is a disaster. Our biggest competitor is Osram, and AMS bought Osram, I think, four years ago for $3.5 billion. You know what AMS' market cap is today? $1 billion. Lumileds is in Chapter 11. They have 40% utilization in their factory. Nichia is a really good competitor, but they're culturally not gonna do much to force consolidation. So Cree is an interesting asset 'cause we have a bunch of excess capacity that can travel. We have a bunch of IP, and we've got the best gross margin in the industry. So it's an interesting business for us to be able to capture more share and take advantage of the kind of weakening of our competitors.
And we'll think about how to play that, and we'll think about how to, you know, generate the right returns for our shareholders in that business.
Great. My last question, just you've been an acquisitive company. You've got $530 million in cash on the balance sheet. What should we be thinking about on the M&A front? Is it sort of a smaller tuck-in around the IPS business, most likely? Would you consider other types of transactions?
I always have a tough time answering that question in the way you framed it, which is we look at things that are good for our company. It could be massive, it could be a tuck-in. We don't think in terms of it has to be something. We look at deals on the merit of the deals and what they can do for our shareholders and our overall growth of our business. We're not gonna do anything on with the cash part of our business, necessarily, that that puts the company in jeopardy. We don't feel like we have to. We have. We feel like we have good markets. We have been an acquisitive company, and we've done really well with our acquisitions.
I'm gonna let Ken talk to the capital allocation priority beyond investing in the business, because we've got some things we wanna do.
Yeah, and we outlined this on the earnings call as well. We have four priorities: One, invest in the business, and a lot of that is on that go-to-market engine within IPS and some of the R&D efforts and activities to support that growth profile, which is significant if you look at the overall SAM. We talked about that growing at a 15%+ type of CAGR. So there's some great opportunities for us to invest in that business. Two, we will look at inorganic opportunities, we've done that over time. Stratus, the most recent acquisition here, about a year or so ago, and then three and four, capital return and making sure the balance sheet is strong.
So in the last two years, we've repurchased about $71 million of stock, and we announced a authorization that expands from, by another $75 million. But we're opportunistic in terms of the share repurchases based on price, and if there's dislocations, we, we try to buy. And then finally, making sure the balance sheet is strong. So, net leverage is below two turns, as of Q1. Gross leverage is a little high, so we'll take that into consideration as well.
Perfect. Well, I think we're at the end of our session, so, Mark, Ken, thank you very much for joining us at the Needham Growth Conference. We really appreciate your participation.
Yeah, thank you for having us.