... we have Mike Staiger, and Krishna here is from SMCI as well, Supermicro. He's here in the audience. So really excited to have Supermicro here. Thank you, Mike, for coming to our conference here. I have a bunch of prepared questions that I'm gonna go through. I'm gonna leave a few minutes here for the audience, but I do request, if you do have any questions, please raise your hand so we can bring the mic to you. Mike, I'm gonna kick it off. So again, thank you for coming to Citi's Global TMT Conference. You just reported very strong growth. I think it was up, you know, close to 50% for the year. The quarter closed sort of within range.
When you sit here and you think about, you know, how demand has evolved for your company, let's say, a few months ago, or even at the start of the year, there was a lot of crosscurrents, there was DeepSeek, there was tariffs, there's all kinds of stuff. How would you say your business performed relative to your own expectations, you know, let's say, the start of the year?
Asiya, thanks for having us. Before we start, just want to please refer to our cautionary statements with regards to the safe harbor statements with regard to statements on our website. So from that perspective, demand has been great. It's a new technology cycle in general.
Mm-hmm.
We had a fantastic year in many aspects. We had, you know, grew 47%. We grew our scale customers from one in 2024 to four in 2025. We're seeing a continual evolution from a standpoint of our customer base, where the neoclouds have been first and early. They've been growing quite rapidly. Enterprises are following on stream, and we're seeing a lot of sovereign activity as we move forward. You know, we're well positioned in the market, and I think one of the things that just as demand shapes up, as the picture shapes up, you think about what's going on, the hyperscalers have been soaking up quite a bit of capacity.
They've been application optimized for AI, and they've been doing that in various different segments within their service offerings, and they have in-house engineering, they have chip design, they have network design, storage, they have everything, right, so from a standpoint of what Supermicro is doing is we are bringing in the AI element. We are bringing all those aspects, including DCBBS, to our customers that are beyond the hyperscalers. The rest of the market does not have these capabilities, so from a Supermicro perspective, we see demand evolving, and we see the product innovation and expanding within our customer base across those three different categories.
Okay. Within each of those three different categories, like, you know, how is Supermicro positioned? Like, you know, where do you lead? Why do you lead that? Just if you can double-click on the competitive moat across those three segments.
We have no legacy architecture, basically, to support on a go-forward basis. We're all about lowering the cost per watt per compute for our end customers. Charles is completely focused on the customer performance aspect and doing everything to make the systems more performative for the customer. In the current context with AI, performance, power, all these elements are extremely important to the end customer, particularly the early movers, the early adopters, who have limited funds, so to speak, from a build-out perspective, and are being very focused with their spend.
So from a competitive perspective, if we can bring a better, reliable system, one that uses less power, one that puts more footprint on the floor for them, initially, that's been a really big win. And you can see that from a standpoint of where our growth curve has been, relative to our competitors and share numbers, etc. So, that's the focus. And so as you think about what's gonna happen next, we believe there'll be more neoclouds coming into the fold. There'll be. Enterprises are entering into the fold. They're already doing that through some of the larger neoclouds that there are today, and we support most of those neoclouds. But there'll be application-specific and application-optimized solutions for the end market that we will be developing.
If you think about what Supermicro does best, is we offer customers choice. We innovate on almost every platform. We have almost every architecture platform available for end customers.
Mm-hmm.
It becomes a differentiation point at that level because some of our competitors are just focused on building, say, one particular architectural rack, and we're focused on the beyond, including the rack, making that better, but beyond with other proxy, inferencing products, etc. This goes across not just AI, but general purpose compute as well. We have a lot to offer our customers, and they know that, and we can save them money. You know, in the element of the data center building block solutions, we're bringing other components so that we can get the entirety of the data center delivered to the customer, as opposed to customers like: "I just need a rack." It becomes a rack-to-rack price comparison.
But if the customer needs more help, and we see many of them that are coming on stream, like I said, we can, we can bring these, the hyperscaler experience to the rest of the market. So that's a product differentiator for us, and that helps, customer wins, helps repeat customer business. We have a very high, repeat business with our customer base because we keep innovating, and we think that innovation curve, because it's accelerating, is an advantage from our perspective, because we control the design, we control the manufacturing, we control the engineering, so the entirety of the experience, and we understand that with the customer. So as the innovation comes, it, it accelerates on itself and leaves, others who are still struggling to get the, the current generation into the, into the marketplace.
We're thinking several steps forward and what the customer will need eventually downstream.
And how would you characterize, you know, conversations? You talked a little bit more about neoclouds, but we also have the sovereigns, and we have the, you know, enterprises that are also adopting. The conversations, if you can, you know, is there any differences at the speed, at where they are in terms of adoption, of these, you know, not just on the rack stuff, but on the DCBBS as well across these verticals?
So what's interesting is, you know, well, obviously, 47% year-over-year growth is amazing unto itself, right? Especially at scale, from going from 15 to 22. And we see 33 in 2026. But the speed from a customer perspective is obviously fairly quick, from a standpoint of what they want to do. There's many POCs, so that's kind of how you get to market, at least from some of the more scale customers. They'll trial equipment, they'll get specs right. So in this context, you know, we're early days in the adoption of AI in general, right? So you have issues like power at the data center. The data center readiness, you know, concrete's being poured, you hear that all the time.
You know, with respect to sovereigns, there's you know there's licensing elements that's a factor. Like, if they have the licenses, can they ship? Have they spec'd out the data center right? So that kind of goes back to what we're seeing and what we're doing with DCBBS, where we can take, you know, we could retrofit a data center for them, bring the whole experience together. We have the leading edge from a liquid-cooled element. So once again, it comes back to our ability to innovate and kind of adjust to the customer. So we have numerous customer engagements, so trying to balance that out in the context of supply, in the context of, you know, the readiness from the customer perspective.
You know, you can even throw financing for some of the less well-heeled or the startups, where that's somewhat of an element. We don't think that would be an element for some of the enterprises. It won't be an element for the sovereigns, but there's multitudes of factors that impact the uptake.
Mm-hmm.
But again, the speed of the uptake, we think has been very, very good and very supportive of accelerated growth rates for ourselves.
Okay. And then, you know, AI, great tailwind for you guys, some of your peers as well, but it's also extremely lumpy. You know, there's transitions going on at the chip level, and then all the various components that need to go to support, the memory, for example, the cables, all that stuff. So just help us understand, you know, the level of visibility that you have, and considering that there is so many, so much lumpiness, and what are some of the risk mitigation stuff that you do on your end so you don't have, like, you know, terrible margins as a function of the lumpiness?
So from that perspective, the variance quarter to quarter, I mean, it's quite a feat to put a significant amount of rack capacity on the floor installed, et cetera, for any customer, given the constraints of the supply chain-
Right
... where all of different parts, you know, if you don't have one piece together, if the data center's not ready.
Right.
I think we had mentioned that we had one customer who, last quarter, had a design change. You know, something didn't get spec'd out right from or they wanted something changed, and so that kind of impacted what we're doing on a move-forward basis. So these elements are contributing to the variance. And the other thing that I would mention is that, it's very early days in this whole build-out period, right? And so, the adoption of the technology, so there's a lot of folks that have learnings about what they're trying to do before it becomes, you know, a mainstream, you know, product, so to speak. That said, there's multitudes of products and solutions being offered to the end customer in application optimized for vertical AI.
You'll see neoclouds that are focused on one particular vertical as we move forward. So there's so much happening that it's hard to, you know, capture all that at one time against all the other factors that put variation in the results, but I think we've done a really good job of managing through that in real time and again our learnings from standing up some of the largest clusters, one, has driven enterprise inbounds from our perspective, because they know that we have some of the best, most reliable equipment and gear and our capabilities.
So from that perspective, I think it puts us in a really good position to take advantage of the expansion that we will see in this whole marketplace, so we can serve more customers.
Okay. All right, data center building block, DCBBS, if I got the acronym right. You know, just maybe, I know you touched on it a little bit earlier, but just help us understand, you know, why is it unique to Supermicro? If I talk to some of your peers, think I hear some of the similar commentary, but just why is this unique to Supermicro, and how do you think it accelerates your growth? Why does it accelerate your growth? You know, I think you already touched base on maybe the sovereigns are still kind of trying to figure out the licensing issues, et cetera. When do you see that inflection happening outside of neoclouds?
So we are very focused on the customer, and giving the customer the best experience they can get, to give them the best performance, the best reliability, the lowest cost per compute per watt. And the initial customer base, the initial early adopters, the big focus, right?
Right.
We think that'll still continue to be the focus, to be application optimized in the enterprise, to be application optimized as we expand in the neocloud. So what we're trying to do is, like I said before, if you have the hyperscaler, they're doing everything in-house to serve those applications, and we have the ability to do the same thing for the rest of the market, and the rest of the market's really large... and as they adopt that, you know, there's a brownfield site that needs to be upgraded to air cooling, excuse me, liquid cooling, but not quite all the way to full liquid, right? So we can bring some of those componentry into the mix and for the customer.
And we'll be adding more of the pieces of what the entirety of the data center is. So a customer can call us up and say, "Hey, we need a data center in XYZ location." We'll be able to drop in everything that they need so that all the equipment arrives day one, so they can turn on, they're online, they're ready to move and ready to roll. So incrementally, that changes the dynamic with the customer, as opposed to a customer coming to us and say, "Look, I need X amount of racks of whatever architecture, you know, what's your price?" And this changes the conversation. So it includes the service element. It includes, you know, the engineering support because the service is a little bit different. The cabling, which is a pretty, you know... And the connectivity element.
So we'll bring all those pieces, and even, you know, we're building the back-end storage, whether it's gonna be flash or whether it's gonna be hard drives with our ISVs. And if you think about what we do with one of our VDI software partners, we build systems for their software stack, and we're probably one of their better suppliers. So ultimately, over time, as the AI application rolls out, you'll see us marry that, our stack or our data center stack, whether it's a mini data center or a large footprint, we'll be able to bring all those things. And that'll be a differentiator in general, and we'll be able to serve the market. The market's large.
You know, it's obviously competitive, but we've always pushed the curve on technology, and the technology curve continues to accelerate, and we're focused on staying in that, in that zone.
I mean, when I look at your OpEx ratio, it's like great. But just the introduction or the expansion of DCBBS as a growth driver result in increasing that OpEx?
I mean, it's the pace is a little bit ahead of where we would like to be, but generally, with half the growth rate would be the growth in OpEx. So, I think historically, the company's been very, very efficient.
Right.
And, you know, that's the goal, that's the target to continue with that. So, I don't think you'll see any outsized, you know, situations that change that parameter.
Okay, and then the same thing on free cash flow. Does this become a more working capital intense if you're having to, you know, host a lot of the components that go beyond just the rack? I think you talked a little bit about that. Does that become a working capital?
So I think in the early days, historically, there's been a little bit of a land grab, and there's a little bit of a working capital demand from a standpoint of from serving customers. But as you probably well know, recently, we have improved the balance sheet significantly. I think when we reported last, a couple of weeks ago, we had $5.2 billion in cash. We entered into an AR facility that gives us a little bit more liquidity if we need it. We ran the business at up to $22 billion off of $2 billion, I think, I believe, in cash. So we think we're adequately funded. That said, if we... And I mentioned that we scale customers.
We had grew to four in 2025 from one, and we see two to four more coming into 2026. So from a timing perspective, if all four more scale customers and the four scale customers we have all hit at once, you know, it might change the conversation. But we think as we move forward, you see not necessarily neoclouds. I was gonna say sovereigns. As the sovereigns expand, we believe that they'll be less capital intensive because they obviously don't need to raise funds. So longer term, we think that we can get free cash flow close to net income, and that's the goal. But right now, it's about land grab, land and expand, and market share and innovation.
So we're focused on that element, and that's growing the size of the company. As you well know, the scale and scope of the company has changed dramatically over the past couple of years.
Okay. So I got lots of questions on that, you know, your growth and your major customers. So, you know, it's great you brought that up because, you know, just help us understand, what does the concentration look like today? Now you have four large customers. You're talking about two to four more in the current fiscal year. Where are these customers? Maybe vertical, geos, are they neoclouds, are they enterprises?
Yeah. I think it's fair to say it's a NeoCloud with maybe an enterprise in the mix. We think that there'll be an expansion of neoclouds with the existing ones, of course, and the sovereigns. So some context around that would be that, you know, we've heard instances where developers have been able to peg models that are being offered in service providers, where they can run multiple test, you know, programs, and they've saturated systems. Context windows and reasoning models are too small, and if you want to expand those, obviously you'll need more in the language model or the reasoning model, you will need more capacity. And so you see hyperscalers throwing capacity into neoclouds, right?
You see enterprises putting capacity into neoclouds because they haven't yet to build out, so you see this ecosystem where the more you use, you know, AI-enabled tools, the more the cloud-based service providers are going to expand, and if we start moving into, and we are seeing that, the inferencing element, you get this even more users that drive more demand within the large customers, but those ancillary customers continue to expand, so the whole footprint is at the cusp of just, I don't wanna use the word exploding, but I just did, but it's gonna really, it will really move the needle, so we see that ahead, and what we're trying to do is make sure that we have product for optimization for any one of those applications.
So that gives us an innovation element and a solutions-based element that goes beyond, you know, just give me a rack. And it's a different conversation, and that stems some of the neoclouds, some of the enterprises, and certainly sovereigns.
Okay. All right, and again, from a geo perspective, are these- I'm assuming this is not just a U.S. phenomenon, I mean, this-
No, um-
Thinking about these large customers.
These
Yeah.
There will be... I believe, you know, you can obviously look at the filings, but one of them is, you know, they're-
Yeah
... they're gonna expand across the-
Yes
... yeah.
Okay. Just wanted to clarify.
Especially, you know, we, you know, we were talking about, you know, we've made some announcements with respect to some Middle East, you know, Southeast Asia, Europe. I mean, there's activity is strong.
Right
... across the board.
Yeah, but to the extent that they become a major customer this fiscal year, right? Do you feel pretty strong about that two to four extra customers? Yeah.
Yeah, and you know, how that bakes out from a geo, you know, we'll see.
Okay. All right, I'm gonna just see if the audience has any questions here. Please raise your hand. Oh, we have one here.
Hi, thank you for taking the question. Do you have confidence that the inference-based models and some of these new expansion areas will be able to replace the large-scale data centers once those build-outs are completed?
Thank you for the question. I think it's a very good one, but I think the way we see it playing out, like I said, if the context windows are too small, or will expand, then the core language model providers, the reasoning models, they'll just get bigger. And if they get bigger, the number of reasoning, or excuse me, the inference element will expand as well because more people will use it. So think if a user base expands dramatically, and we're already seeing that the, you know, token usage is skyrocketing. So this should gather steam, so it actually should support all sides of the business, and the growth of the business should be pretty excellent.
You won't land on, you know, it's all inferencing versus it's all training, and there'll be new use cases to train, and there'll be specific training model verticals, et cetera, so it's, like I said, it's early days, but that's the direction we see from the customer conversations that we're having across the globe.
One more here.
Could you just, just comment briefly on the kind of competitive situation with Dell? You had the early lead, but they've been ramping fairly fast, on the relative side compared to you.
Um-
Yeah
... so from a competitive perspective, we feel very well positioned. Like I said, we're focused on innovating, focused on making the reference architecture from any one of our partners, better, more performance-oriented, or more power efficient. That's our complete focus. We control our design, engineering, manufacturing, so those elements are very underappreciated, I should say, from a competitive perspective. And so, again, we have limited, we have no legacy thought process in our mind. This is all about what the next generation is going to be, how we can bring more to the customer, and to do it reliably, and have really sound, solid systems. So we think those will support and continue to support our market leading position and will increase, the scope and scale of the company.
Our goal is clearly to be number one, the number one supplier, and we would argue that over time, if you look out two years, we will have the products that you will need downstream because we're focused on bringing those to market. We feel very comfortable that we'll continue to be in a pretty great competitive position.
One more there.
Hi, thanks for taking the question. Can you expand a little bit more on how you see funding in the case where you do need to do a more aggressive cap, I guess, working capital build-out, and how you think about, I guess, issuing more equity, the facility you mentioned, or even raising more convertible debt in the future?
Yes. So again, thanks for the question. I think from a standpoint of where we are today, we feel pretty confident where we are with the capital stack that we have and the access to capital. We've always been, or we've tried to be very focused on not dilute shareholders. So from... I think that if there was anything to be done, we would probably leverage the balance sheet a little bit with maybe a revolver or something along those lines.
That all said, and there was a caveat earlier that if four scale customers, you know, all of a sudden, "Hey, we have new demands," and we had two to four more come into the book, we would have to think about how we would best position the capital, our capital needs, to serve those customers. I'm not saying that's gonna happen, but, you know, that would be the outside band kind of demand element. So I think we've done a fantastic job. The balance sheet's in great shape right now. That should support easily our ambitions for the year, and so we're in a really good spot.
If I can... You know, since we are talking about competition, just on margins, you know, I mean, I think they're hovering around 10% gross margins, so but they are down relative to the last few quarters. As you think about DCBBS, as you think about expanding your customer concentration, to more customers, how could the margins move up towards a target model? Is that the scope for this year? Are we looking beyond this year?
So we definitely have our eye on that, but I think the mix is pretty important. So early days, large customers have a little bit more pricing power. There's a little more eagerness to try to serve a couple of people. Some of the competition, you know, need to get a footprint to say they're in the game. So a very early day, a lot of pressure in that respect.
What we're seeing, as we move forward, is more and more volume of customers, whether it's in the enterprise, or some of the newer neoclouds that need more help, will bring more product or more services into the fold for them, and they won't have necessarily the same dynamics, and they'll be more in the line with where we have normally operated. So that should rise over time. So it'll be a balance. We're not saying that's gonna happen. Obviously, we didn't say that was gonna happen in the next quarter or the next quarter or two, but we see that happening over time, and Charles was pretty adamant.
He said, "Look, if we want to accelerate the growth rate, we could lower the margins pretty easily." But the focus is, no, the focus is on a profitable business, serving a broad base of customers and, you know, becoming the number one provider through innovation, through product, through, you know, a solid customer experience.
Mm-hmm. Okay. And then I know liquid cooling, Charles talks a lot about, you know, Superm icro's moat in liquid cooling. Just help us understand about the adoption there, what's going on, what's your market share as it relates to liquid cooling? Are there certain customers that are perhaps adopting it faster?
Scale customers that have a liquid-cooled data center already obviously are, like, the number one candidates. We've moved beyond the initial liquid cooling to DLC-2, 40% improvement across the board. We're taking all the heat out of the systems. It's always been in our DNA to lower the power requirements and to be thermally efficient, and that, you know, obviously enters into the liquid cooling equation. We've been the leader there. We think that as we move forward, we'll be able to take, you know, maybe 80%-90% of the data centers might be just air-cooled environments, 20% might be liquid-cooled environments. There's no data center that's being developed that won't incorporate liquid cooling.
So, and, you know, obviously, liquid cooling customers have a little more rack concentration, so there's a dynamic there. But we're bringing better thermal dynamics in the HGX series and the air-cooled series, whether it's a 300 or 200 to the market. We're doing that the same thing on the CPU side or the x86 platforms; we can liquid cool those. So as power becomes a constraint, and it already has been a constraint, we're focused on delivering more power and more efficiently.
And if we can do that, and we have been doing that, and continue to do that, then customers will say, "Well, I can save a lot more money by a Superm icro liquid-cooled design than not." And that's a customer win, and it's for us, and it should be margin accretive for us as well.
Okay. All right. Just on the guide, I think you guys talked about fiscal twenty-six, which has already kicked off for you, of more than 50% growth, or roughly the math there. So what gives? You know, what are you seeing now? You know, you've already guided. It's been a month, I guess. But what are you seeing that gives conviction in that? Is that based on backlog? Some of your peers do report that. I know you don't, but what kind of gives you conviction in that 50% year-on-year growth? Admittedly, it's not gonna be a straight line. Like you said, there's gonna be some quarters we'd see a big uptick versus, like, the most immediate quarter that you've already guided to.
Yeah, so from-
Yeah
... from that perspective, yeah, we don't give out backlog numbers. We have excellent visibility from our customer base, and from that perspective, I think we feel really comfortable where we think we'll end up from the year at the year. So, and it's based on our customer conversations, you know, customers' initial order potentials, et cetera. So I think we're shaping up for another really good year.
Okay. Let me ask about, you know, lots of hurdles that are behind you, that plagued you guys last year. I think you're past a lot of those-
Mm-hmm
... and kind of looking forward, I think that's what gives you conviction as well in the 50% growth. Where are we with all those recommendations that perhaps a special committee provided? Are there still things that need to be done? Do you guys feel like, okay, all those things are behind us now, and we're well-set for the growth? Or could there be... I think, as Charles put it, some of those impeded some of your growth last year, last fiscal year.
Yeah, so you think about the 47% growth year-over-year was in the context of a new product launch from some of our partners or new product launches from our partners that were where the end markets were a little, you know, maybe they all went to the hyperscaler, so it was a little bit of a-
Right
... an interesting launch, so to speak. You know, there was the power constraints in some of the data centers, et cetera. So, putting a 47% growth in the context of some of the administrative challenges that we had, that we cleared, the capital challenges that we may have had, that we obviously cleared, now that we have, you know-
Mm-hmm
... significant. So all those things look really good. So 26 , the road ahead looks a lot smoother from our perspective, with respect to all the things that we need to do. So if you think about the size and scope, scale, and the growth rate of and the acceleration, we've done a lot of work on the people and processes element. And we're very conscious of, like, getting those right and taking the time to make sure they are right as we move forward. And certainly, you know, some of the press likes to take, you know, shots at some of the things that happened in the past and relive those things.
But we're focused on moving forward, we're focused on the customer, we're focused on the innovation element, and we're putting the people and processes piece in place, and that takes a little bit of time.
Okay, and then inventory, I know, you know, sometimes you end up with excess inventory, whether it relates to a prior chip generation, et cetera. How well is Superm icro now, with all the processes that you've put in place, to manage that kind of risk?
Yeah, so that's a great question, and it's very forward-looking from that perspective. We are concerned with that, of course. We tend to take very little inventory risk, and be careful about what we're gonna place where. It's probably a little bit more important now that we have multiple different platforms to go to market with, so we're not committing or making a bet on one platform versus the other. And that said, if a customer comes to us and says, "Look, we have a scale order here," and it's, you know, it's a certain platform. When we mentioned the platform decisions, we're impacting some of the things in the prior quarter, where, you know, there could be...
You know, we're gonna do what's right for the customer, right? So from that perspective, we're gonna try to make sure that we don't have an overhang. You know, we had one a quarter or so ago, and it was modest. I mean, we didn't wanna have that, but it was a modest overhang.
Outside of chips, you know, what about some of the other components? I mean, there's always supply, demand, imbalances that are going on there. How does Superm icro manage that? Do you tend to do strategic buys on-
So-
-any of the commodity?
... one of the underappreciated elements of the company is our ability to source and source for a large amount from a goal perspective, and to have choice, and the building block architecture where we can swap, you know, different components within the subsystems themselves. So, we try to be very careful about what we're ordering, when we're ordering. So, and occasionally, we might do a strategic buy, but it shouldn't move the needle, so to speak. And we have a pretty good idea of what we need, when we need it. But if, like, if we have a scale customer, there's a change, sometimes, we could, you know, there could be an issue-
Okay
... but, very, very carefully managed.
Okay. We have a little bit more than a minute to go, Mike, so I wanna thank you, but I wanna talk to you about, you know, what are investors maybe underappreciating, or what would you like the investment community to know more about Superm icro?
I think the biggest underappreciated element is the fact that we're innovating at scale, and we're bringing new. I think I've said this a couple times through this discussion that there's a significant amount of differentiation with respect to the reliability, the cooling and thermal elements, and all the things that we're bringing to our end customers is underappreciated from a street perspective. So I think that the fact that we're controlling the engineering, the design, the manufacturing, and the innovation is very underappreciated. And I think one of the reasons that we're in the position we are today is because we're doing all those things quite well at the product level.
Great. I'd like to thank Mike and Superm icro's management here. Thank you very much.