Super Micro Computer, Inc. (SMCI) Earnings Call Transcript & Summary

All right. Welcome, everyone. Day 1 of Citi's Technology Conference. My name is Asiya Merchant. Afternoon here of day 1. Glad to see everybody. I hear -- I lead the technology hardware and tech supply chain here at Citi Research. Really happy to have Super Micro's management here with us. Mike Staiger here with me, and we have other folks from Super Micro as well in the audience. This presentation here is fireside. So I do have a bunch of questions. I'm going to allow for a few minutes for investors to ask questions as well. And please do raise your hand so we can bring the mic to you. All right, Mike. Pretty good. How Super Micro?

Super Micro is doing great.

All right. Well, you guys had a very, very strong fiscal '26 growth rate, right? I mean you're talking about -- I think you talked about 78% that you reported, you have Q1 revenue that's very, very strong coming in. Talk to me about just your guidance, how are you thinking about other constraints that are kind of factoring into your guidance here? And just how are you thinking about the picture that you're sharing for Q1 relative to some of the other constraints that are out there?

Well, interesting, it's good to see you again. I think it was like a couple of years ago, we were on stage, and we had just done a $14.9 billion a year. And now we're talking about doing that at the midpoint of $15 billion quarter. So how things have changed or progressed over time. So anyway, obviously, demand has been strong in the backdrop, and we've reflected that in our guidance. And we talk about customer diversification and expansion of opportunity. There's an expansion of product platforms across the board. So from our perspective, we're trying to fulfill that demand. And part of the cadence on the order book is having a pretty good idea of where the customers are going to be at what time, what they need, where they need it. And I think that's pretty helpful from the standpoint of backstopping the order book and the guidance that we have on a go-forward basis.

And there's a lot of discussion around constraints, right? I mean everybody is reading the papers. Everybody knows there's constraints, whether it's memory, it's CPU, it's power, how is that kind of -- why do you feel like this demand is durable, and it's not just they're coming to Super Micro, they're coming maybe to your peers and saying, give us this, and there's a little bit of double ordering perhaps because everybody is worried that they're not going to get it.

Yes. I think the durability -- so one, the guidance across the board -- the outlook across the board for the industry has been robust. There's been outsized performance from a lot of players. And the backdrop of this has been the applications. And you see application development is off the charts for new AI workloads -- were application optimized. You see that the solution sets that are being offered to customers from our partners, whether it's NVIDIA, AMD, Intel, ARM and others. There's multiple different solution sets that they're targeting different verticals. So this expansion of opportunity set at the end user base is significant. Last week, you saw VMware enter the fray on the BCF for the AI factory stack. You saw our announcement and engagement with Cisco. So those are enterprise-focused elements where I think the market a year ago was as the enterprise is going to adopt AI. And there's clear proof points that there's an expansion going across the board. When you look at our -- again, our diversification across the customer base is really supportive of this. And the solution element is where we're really focused on. We mentioned that to you in the past, I mean, months or quarters ago that DCBBS strategy will help the solution to the customer, and we're focused on delivering an application optimized AI solution or a factory, so to speak, and there will be a mixture of components that are in there, and we're going to mix those up and sort of supportive of margins and the application stack is expanding. So we're pretty excited, and we're racing ahead to build what customers need.

Okay. I think customer diversification has always been one that focus -- that's always been our focus for Super Micro, right? I mean you started off with these large systems. And by virtue of selling these large systems, they were typically one or two customers that were key drivers of demand. But now you have more than that. I think you have more than 9 customers that are generating $1 billion in revenue. So the diversification story has played out well. for you. And so any one quarter, that could be maybe more customer concentration. How are you thinking about customer concentration today? I mean is this a path that is still sort of for you guys to still to diversify even more? Or are you thinking about some big customers that are going to be primarily driving a lot of that demand.

So we were pretty clear about the fact that we were focused on expanding our footprint with respect to the enterprise. Sure, there was a few large customers in the past and still will be, as we move forward, as the initial build-out was occurring. You saw a few large customers where everyone wanted to participate in -- with those customers. So there was a little bit of pricing pressure, et cetera, if you want to call it that. But those customers are expanding their use cases and their needs and expanding upon what services they can offer, like AICP systems are going to be offered by some of the larger customers, an enhancement of what they're doing at the large customer site. But under the belly of all this, again, the enterprise customers, neoclouds and even the sovereigns are looking to stand up infrastructure to take advantage of the applications and we're there to support them. And we have the -- basically almost any diversification and breadth of platform for them. And so this will be very helpful from a standpoint of broadening out that customer base. We're already seeing it, so that's great. But we'll always have -- and we always have had some larger customers in the mix. And the argument has been we're reliant on one large customer or two large customers and they'll be displaced by someone else. And we've had a very nice history of gradually expanding with our customers and expanding with our capabilities for them and serving them. So customer focus is a huge thing for us and we do -- we're doing -- we're offering them a whole host of different solutions to accommodate their needs most effectively, well.

And then just a little bit on lumpiness, right? I mean you could see that in the last quarter where the revenues came in a little bit towards the low end of the guide. There's always lumpiness in this segment, right? There's always power constraints or something. So how are you guys managing that. Maybe you could talk a little bit about your visibility, how you think about lumpiness, how it flows through like on your working capital side as well?

Yes. So from a lumpiness perspective, if you want to call it that, we've had some customer -- there's been some challenges at the customer sites or supplier sites or whatever it may be. And those have always -- we've always captured those revenues downstream. It just might not be in the same time frame as what we had anticipated due to things beyond our control. So one of the good things about this situation is as we move forward and diversification occurs, there's a better visibility of what, where and how. Again, and the other thing is that the industry has matured to a certain extent, where there's more folks who understand the dynamics of standing up new services or a new data center and have been preparing for quite some time. So we're kind of meeting that in a better fashion where we anticipate or the customers anticipating their needs are met. And this also kind of backs into the DCBBS strategy because we saw early on customers could order a large amount of racks, and they might not have all the componentry that they need to stand those racks up and get their systems working. So if they have made a of significant investment, let's say, $0.5 billion or $1 billion, and they're sitting there with idle systems because they can't integrate it. They can't get a networking stack to work. They can't get it integrated with their storage stack. The power cooling is not in the right area. But we're productizing those things so that we can drop in factories, essentially AI factories for the customers that when they make the investment, it hits the floor, they turn it on as prevalidated, as this goes to the L12-validation services. They're paying for the use of the systems. They're not worried about trying to stand these systems up. And the hyperscalers can do that because they have all the engineering support in the world. But enterprises and many neoclouds do not have this capability, sovereign don't have those capabilities. But we're doing it for them, and we're pricing for value for those customers. When they get these systems, the more important part is the reliability of those systems is significant. So the uptime that's available to that customer to run their workload, they have more of the available system. So that saves them significant amount of dollars. I do believe one of -- there are some figures out there like 8 racks of NBL -- 72 being down for 30 days or not being available for 30 days is 3 million. So you can see customers that have 100 or some customers who might have 1,000, that's -- or more. That's a huge amount of dollars for one month idle kind of situation, right? So customers are now realizing that where they might not have realized it before and the race to a low price was the element, not the race to have a system that works as services are available as more of a thrust delivering the value is what we've been doing, and we've articulated that even a year ago with the DCBBS strategy where people are trying to pick apart what does that mean for margins for you, but what it means for the customers where we're going, we're enabling the customer get some work gets their systems up, working faster. And it's better for our partners because everything is more synchronous and that is -- lowers the capital burden too as well because the timing of putting all this together delivery payment, contractual terms are far more favorable for us.

Okay. And then just within that mix, you have enterprise, you have sovereigns, your neoclouds. Any further incremental like breakdown, you could say like what percentage the revenue or at least of the forecasted revenue you're thinking those three buckets?

We haven't given that kind of level of detail because the irony here is we're not even talking about hyperscalers, but hyperscalers are putting workloads in to our customer base. That customer base is expanding, and you're seeing some newer contracts that have come up, you're like, holy cow, where is -- I think a year ago, so where is the revenue going to come from? And there -- we're seeing it -- so you have improved credit quality across the board and enterprise is participating. And so I think the rising tide, so to speak, is lifting all boats. -- and you see enterprise activities accelerating. Again, we talked about or I discussed some very large enterprise partners, Cisco or VMware Like, they're that cadence of adoption is we're in early stages, but it's spreading out. And it's proof points in our order book and probably others in the industry.

Okay. And then you talk a lot about DCBBS. I know that's a topic that has served you well, and we just talked about it in terms of being able to speak to market the reliability of these systems. Any color you can provide? What percentage of your revenues is it? And what opportunities lie ahead? And would it be across -- like are you seeing adoption at an equal pace between the neoclouds, the sovereigns, the enterprises -- or is 1 of them further along?

Yes. I don't know if I can -- I would assume neoclouds is probably further along, but the real thrust of this is not just that we're supplying all the component treated they can do, but the integration of the system so that they can actually use the system without any -- with an increased enhanced reliability is paramount, right? So all the different parts, we need to integrate them to actually make them work, firmware, et cetera, software and our software stack that sits on top of it. where ultimately that software stack will be able to run across different AI factories and integrate those as well. So we're trying to really empower the customer to leverage the infrastructure that we're providing them. Again, this is solution value. And so it's something that we think that we're well ahead of our peers in providing and we'll continue to forge ahead with new capabilities as those capabilities become available or known to the market.

Okay. All right. Let me just ask the audience if there's any questions, please do raise your hand. Okay. Let's talk a little bit about gross margins. You guys had like stellar margins in Q4. You talked a little bit about favorable mix here. I think then you guided for a little bit of margin volatility in 1Q. Aside from the quarterly variance, like how should we think about the overall margins of this -- of the business? Like where do you think the target could be as we're growing your revenues, yes?

Pretend on the gross margin side, we're giving one quarter at a time because of the variability in what we're actually doing. But the longer-term goal from an internal perspective is to double-digit and move that bar higher over time. I think there were some older targets that were flowing around there were 14, 17. We just showed that we could do a 17 and change quarter. There was some favorable things in that mix. And as we move forward, it's hard is to solution the customer and provide more value to them so that we can capture a better margin number. So the long-term goal is clearly higher. We called that a while back where profitability was important, and we've been working on the profitability element. And so I think as we encapsulate more value and more componentry in the systems, I think the market looks at AI systems as like a low margin build and they're going to -- in a lot of our success to that. But what we're doing for those customers, there's multiple different variations and customers are kind of aligning them to the application. And so there's oftentimes where we might be the only vendor supplying that. And if we put some more component country and whether it's power cooling into the mix, that gives us a better value to the customer, and we're getting paid for that whether that's 10%, 12% or whatever the number is, we're getting paid for that because we're bringing the value to the customer, and we're saving a significant amount of dollars in the back end of having the system available and having a more reliable system. So we're moving into -- more into that model. We've called that out prior, and now we're actually delivering that. And so we're on a pretty good path here.

And then what about on the -- as you're trying to scale these DCBBS, there is a lot of services, like you said, attached with it to make sure that these systems are reliable and enterprise or wherever the end customers can turn it on, it works, -- what about on the OpEx side of things? As these are scaling and you're growing them as a mix of your business, do you also have to invest as much in OpEx on the service side?

So we're -- currently, we're investing in go-to-market and services. We recognize the fact that we have much to do to expand those capabilities because the customers are demanding it, and we want to support and service our gear. And of course, it's a little bit -- it's enhancements to the business model. The L-12 element of validated services is super important. And so we want to support that as well. It's built into the guide in the model, and we've been very efficient from an operating expense perspective. So we're not going to break any eggs in this situation. So could we spend a little bit more if we need to? Probably wouldn't -- you probably wouldn't see that in the mix, but we're investing in that. We've been investing in capacity. We've been exiting a lot of different areas and building the business as quickly as possible. So it's a real-time run-time environment, large start-up, large public start-up.

Similarly, like people just -- I mean, I know you just talked about it, the go-to-market, but the enterprise business in the channel, like it's pretty diverse, especially as you think about the international as well. You have to support these customers. Just help us understand, like given the model that you guys have typically run where OpEx is still a very, very small percentage of your revenues. Like how are you thinking about penetrating this enterprise base,and what additional investments that you talked about, are you doing that's going to support the go-to-market and the sales and operational side of...

I think it's important to understand that the basis of the organization is enterprise-focused and always has been. And what we're bringing to the enterprise is something unique and different, at least we feel, and that is AI enablement, and the infrastructure looks a little bit different than the traditional what you've been covering the COTS, server networking, body off-the-shelf server. And so from that perspective, we're focused on delivering these solutions. So again, we have multiple different partners, multiple different platforms within those partners that customers are now understanding that they need. So we are front face with the enterprise. It just happens to be the enterprise -- that segment of the market that we're engaged with are ones that are very technology foot forward. And the proof points are -- have been our delivery to them some of the large names that you know, the proof points are to some of the neoclouds that are serving the enterprise. It's arguable that neoCloudis our enterprise businesses. We have some really strong relationships on the OEM front where our systems are literally installed everywhere. And so our motion on enterprise is to expand it more to serve more of the customers because we see the need for AI applications expanding and that we can take meaningful share, and we were going to support those customers on the services side. And if there's markets that were not necessarily present, we'll partner, and we're 100% committed to it, and that's been the engine of growth over the past let's just say, 10 years for the organization. So putting more performance, more power and a smaller envelope, all the things that a customer needs, and we've argued that enterprises are aging equipment. And if they want to put AI factories in, there's a likelihood of a significant refresh of the existing CPU-type business, consolidate their more density more power and then have some room for or an AI system or an AI factory to be installed next to it. So that looks to be the direction of the market, and we're going to supply that. We already are supplying that.

Okay. And then just a little bit about liquid cooling. I know that was something that would get a lot of attention. It still does. As you're -- as that offering expands just from high-end GPU deployments, maybe to CPU deployments. How should we think about liquid cooling and how Super Micro is differentiated in that space? And what it could do to your overall product margins as well?

So on that front, we're providing customers the liquid cooling, the power if and where they need liquid-cooling GPU systems, liquid-cooling CPU systems. And so our goal is to optimize the liquid cooling element to get the best thermal dynamics for the customer, which we have always led in that category. And so I think that we're going to continue to push forward. I mean, there's lot of different ways to do liquid cooling, including immersion, and we've done that in the past, and there's a call for that in some particular cases in the future. So the power envelope and doing our power supplies and our banks, et cetera, we're 100% aligned to optimize those for the system for the factory, so to speak, that we're going to deliver to the customers. So it's part of the dynamic. It's part of the offering, and we'll just continue to improve upon it as we move forward.

Any updated adoption here?

I don't think we do necessarily outside of uptake of the systems that are naturally known to be liquid cooled. The shipments of -- those shipments are...

Okay. For the large labs, yes.

And then there's customers that are retrofitting air cooled systems with some cooling element to get more performance out of them to cool them to get more power efficient. So we're able to supply that in many cases.

Inventory. I know there's a lot of changes that are constantly happening on the architecture side and GPU transition. So your inventory did bump up quite a bit. I think you talked about inventory days roughly around 19. So how are you managing this risk? I mean, given that customers are changing architectures are changing, customers are adopting, deciding what works, what doesn't work for them. And then relative to the inventory that you have, like how are you managing that risk of making sure that there isn't any obsolescence risk there?

Yes. SP-7 There's always been a fear of obsolescence risk, but I don't -- I think it's overstated. We've been able to manage that quite effectively. We have had some -- a few minor inventory charges in the past than a reversal recently. Customers are looking to use whatever they can get their hands on. So there's been a an element of every tier within the market is finding homes for systems and the durability of these systems is pretty extensive. The $1.9 billion in inventory. I think you can kind of kind of match that up to the forward need of the company or the guide. So we've been pretty good at managing that quite effectively to minimize risk. So I think there's been no real change there.