Promise to deliver better, better factories, better cities, better homes, and we would through the, through the analysis of the data. But what AI is doing is delivering. That's the difference between the term, the bubble and long term. When you think back to some of the long-term plays, like the databases, like before in SAP, they really, you know, they took hold long ago, and they're still with us. And we, you know, we see AI as really the delivery of the promise to big data. Not only understand big data, much, much faster than through that using the insights , which gives you, you know, stronger, better forecasts for weather, gives you reduced times for drug delivery, the ability to offer security in the, in cities and things like that.
So there's a lot of, a lot of use cases that are out there. Really, the next development is really the software, which will help to further utilize the capabilities of these AI machines.
All right, so maybe, David, if you can talk about your target customer segments, just for people in the audience who don't know what your focus areas are.
Sure. We're principally split between, what we, you know, our enterprise and what we call enterprise and channel. So that means we address the enterprise- enterprise through both channel partners and also direct, and our business has tended to go more direct. And then on the other side, we have, large data centers and what we call OEM appliances. OEM appliances are, really, software, packages like that are made by Nutanix, that sit on top of our hardware and, to help them, to help them deliver their solutions.
So, you know, you've had great success with the tier two CSPs, but one investor concern that I've heard is, you know, can we expect demand to sustain at that customer set? Maybe give us your thoughts on that.
Well, you know, the CSPs are one of the best places to deliver AI as a service. The reason for that is they are solely focused on developing the software stack necessary to enable the use cases for speed and for flexibility that are required by you know, AI labs, enterprises, and also large cloud providers. Their focus in that area has enabled them a strong place in the marketplace, and we think that will continue.
Okay. Maybe talk to us about hyperscale. I mean, how do you see that segment as a customer set, and is that a target segment for you? Is any of your guidance based on revenues from hyperscale?
Yeah, well, there is. First of all, you know, a lot of people don't appreciate how much money and effort that the hyperscalers spend, especially money, how much they spend in enabling, you know, their AI solutions. You can think, you can think about the amount of money that Microsoft spent on on AI, on the software side, but also developing chips and all that takes a tremendous amount of investment. So when they are able to develop their own reference designs, you know, then they can- they often will take those, you know, to a CM, you know, to manufacture at the lowest possible cost. So we're not interested in that, you know, in a race to the bottom for reference designs.
However, a lot of times they have need for either, you know, overcapacity or there's some place in the market where we can play. And so in those instances, and that includes, by the way, you know, sometimes they will use capacity through the CSPs. So therefore, there's different places that we can play. We're glad to do that as long as we can deliver the return to our shareholders that we're trying to deliver.
Okay. David, I want to talk about some recent media articles and some chatter about Supermicro losing share to Dell. And really, we hear about CoreWeave, and we hear about Tesla. Can you give us your thoughts on that? Do you agree with that? And what is your strategy for gaining share in the market?
Sure. You know, if you look at our sales performance for the nine months ended March 31st, you know, of fiscal year 2024, versus fiscal year 2023, our revenues are $9.6 billion, which is up about 95% over the prior year. Now, I think that kind of establishes the case that we're obviously gaining share at a rate that's pretty much faster than the industry. And by the way, if you look at our EPS on a non-GAAP basis at $15.77 through the nine months, that's about a 90% increase over the prior year. So, you know, we're doing that profitably. And so, the only thing that has restrained us to date is supply.
And so, you know, if somebody else is able to take a share, it's probably because they have to go to their second or third choice, because we don't, because we can only get so much supply. So supply has definitely restrained our growth, and there's no question that we would be further ahead in the numbers, because that's why what's caused our backlog to grow. We'd be further ahead if we had more supply.
One of the things that we've heard is, you know, you know, Dell has a financing arm. Do you think that gives them a competitive advantage? How do you see that, and how important an issue is funding for Tier two CSPs?
Well, I think if you look at the amount of funding that CSPs, for instance, have been able to obtain, I mean, you're talking about billions of dollars in capital and multi-billions of dollars in debt that's been available to fund GPUs. And so whether you're talking about, you know, Blackstone, BlackRock, Macquarie, you know, Magnetar, there's a lot of people, there's a lot of co-finance companies that are ready to back, you know, you know-- And this, we're just talking about the U.S., by the way, that are ready to back the proliferation of AI as a service and of GPUs. So they're willing to finance that.
So, you know, their money is already—the money is out there, and so that's really not been a case, an issue for us.
David, I want to go back to the target customer segments. You know, at some point, enterprises will start inferencing. Do you think Micro needs to reorganize itself as you target that segment, especially maybe in aftermarket services? You know, typically, you would think of Dell and HPE as companies who are more into enterprise. So what is your strategy for gaining share in that market, and do you think you need to do any reorganization of the company?
Yeah, so first of all, enterprise represents 50% of our business, so the two verticals that I gave you are roughly 50/50. So we're already entrenched into the enterprise, and with a lot of different use cases. So what we need to do is we need to continue to do what we've been doing for the last 31 years, which is bringing the very best AI solutions to market. You know, and by the way, at GTC, Jensen Huang was in our booth, and this is on our YouTube, and he said, "If you want the very best AI solutions, buy Supermicro." And as far as I know, the very best means only one.
Right. Yeah, that makes sense. You know, in terms of spending, do you think AI spending on servers is cannibalizing other areas of IT spend, such as on industry standard servers? How do you see that market evolving?
Yeah, you know, there are industry experts, you know, like IDC, like Accenture, which have told companies, you know, besides the fact that they're putting their money where their mouth is, and they're requisitioning, I don't know, some 80,000 employees I heard, to work on AI. But they've told companies that if you don't figure out your AI strategy, you could be finished, essentially. And so, yes, a lot of companies have decided to take a look. Okay, let's make sure we have our AI strategy in place, you know, and we can do our refreshes later, because you know what?
There's a lot of new processors that are coming out, so we can sort that out later if we, you know, if we want, you know, an Intel or an AMD or an NVIDIA CPU or N GPU. So we can figure out that on, you know, later, but let's right now we'll focus on the AI strategy. So yes, there is some of that that's going on for sure.
David, one thing you mentioned just now is that some of the constraints on your growth is availability of components. Let me just talk a little bit about that. Is your growth constrained by how customers are putting in data centers, availability of power, space, and other components? So just talk about the environment there.
Yeah. So we, you know, we have started to ship liquid cooling at, you know, really at scale, at larger, at larger volumes, you know, in, in this quarter. And so there's no question that, you know, that that industry, you know, is, is coming up to speed with the reality of where we're going, which is, the fact that power is constrained all around the world. And then, and therefore, when you build these large data centers, you're gonna have to, you're gonna have to think twice now about using liquid cooling, because by you using liquid cooling, you can not only... You know, we say that it's, it's free with a bonus.
That's because it's free because you're not only having to put in smaller chillers, you know, you don't have to use air conditioning, you know, if you have really, you know, liquid-cooled racks, but you can put more dense racks and more racks into a data center. So it's more efficient if you're using liquid cooling. And so it's really the cutting-edge companies right now that are putting in liquid cooling, liquid-cooled racks into their data centers. So that it's, you know, the huge use of power right now is going to really drive liquid cooling as much as the fact that all the GPUs and CPUs are running at higher wattage. As they go over 1,000, it's gonna start to become, you know, painfully obvious.
I'll ask one more question on competition. Overall, if I look at this space, the AI server market is becoming really competitive. I mean, there are lots of companies who are making AI servers, and when I look at it, just about everyone has their own flavor of liquid cooling, right? So talk to us about whether you think Supermicro still has a competitive advantage in this space, and what do you think is the competitive advantage for Supermicro, even in this competitive market?
So the competitive advantage, you know, to, to figure out what the competitive advantage is, you just really have to look at our financial statement, you know, the last few years. Because what we said, we were announcing, you know, many quarters ago, I think, you know, I'd have to look back if it was seven quarters approximately, when we said that AI is driving our growth. And we, we then started to give a number. We started to say 20%, and then we said 28%, and then we said 50%. Now we just say over 50%.
So our competitive edge is that we were, you know, our CEO, who is one of the best engineers in the world, first of all, and he developed a supercomputer for Jensen Huang at NVIDIA over 10 years ago to work on AI. So AI is not something that is new to us. We were working on that a long time ago. That's why a couple of years ago, we started to come out with a full suite of product offerings. That's driven our growth. So now, of course, everyone is running, you know, rushing to the party. This is nothing new to us.
You know, it's really a lot of the same, a lot of the same, players out there, because with, with the number of employees that we have, we're half engineers, okay? We're very focused on what we do. We're not trying to be all things to all people. We're trying to build the very best customized servers and, you know, for some of the best companies in the world. And, we, it's taken us, 30 years. A lot of our engineers are well tenured. We probably have 40,000 years of experience that with it, you know, in hardware and software design. So that's all going into bringing the best products to market.
So, you know, you just said, you know, maybe more than 50% of your revenues are tied to AI. You know, market share data in the AI server space is hard to come by. Do you have an estimate for what your market share is in the AI server market?
We don't try to, you know, come out with, you know, market share data. We really just are focused on trying to, you know, do the very, very best that we can to, you know, raise shareholder value, bring good products to market.
Okay. David, maybe I wanna move to the topic of margins, because I think, you know, recently that's come up, some of your competitors have reported. You've talked about gross margins staying in the 14%-17% range. But when we think about new accelerators that are coming out from NVIDIA, AMD, when we think about the GB, you know, the, the new, Blackwell, you know, the GB200-
Right
systems, is it reasonable to think that you can maintain gross margin in the 14%-17% range? And also, when you think about operating margins, you've been able to maintain your overall operating margin in the 10%-11% range. So is this something that you can maintain, or do you think because of competition, we should expect margins to erode over time? So just talk to us about your general thoughts on margins, gross and operating.
Sure. Yeah, so you know, on operating, yeah, the last four quarters, we've been between, you know, 10.8 and 11.3. Our OpEx as a percentage of revenues, you know, if you look at the last fiscal year, you know, the year in 2022, I think it was about around 7%. In 2023, it went down to... Actually, I guess, it's gone down to about 4.5% the nine months. Yeah, but if you take the nine months of fiscal year 2024 versus the nine months of fiscal 2023, it's come down from 7% to 4.5%. So that's really just, you know, leverage from, you know, increasing revenues 95%, you're getting a much better lift on operating.
Now, on the gross margin, again, we've been doing this for 31 years, and we've been competing against a lot of the same companies. We believe that we charge a very fair margin. We're not charging 50%, 60%, 70% margins. We're down there in that 14%-17%, you know, and we're working very hard because we manufacture, you know, right here, you know, in the South Bay area, 70% of our manufacturing right now. We have moved, you know, more over to Taiwan, and we're building a new very large facility in Malaysia in an attempt to lower our manufacturing costs and not stop, you know, doing so much round tripping of components.
So imagine a lot of components, of course, come from Asia, so we're bringing them over to the U.S. We're doing assembly test here. We're shipping back out to Europe and to Asia. Very, you know, inefficient. So, you know, manufacturing more in Asia is gonna help us out.
Just keeping on margins, I think the guidance for fiscal 4Q implies gross margins less than that 14%-17% range. So what's happening in the near term in the June quarter, and how quickly can margins get back into the range?
Yeah. So I don't wanna, I don't wanna, you know, update any guidance that, that I've given, but I'll just say that, that, we've been successful at, you know, keeping our, our margins in 14%-17%. That's our- that is our target. We're gonna, we're gonna do everything that we can to, to, to keep them there. It is, yes, it is, it is very competitive out there. And, you know, you know, there will, there will always be companies that are, that are trying to, you know, you know, buy their way in, but, you know, using margin. However, one thing that you can't, dispute, and that is Supermicro brings, you know, not only the, the best performing products, but also very reliable products and the, the lowest total cost of ownership.
So with, we designed our servers to be very energy efficient, which means they manage heat really well. And, and so that goes into the design of our chassis, which we do, and the design of our motherboards, which we do. And the design of those things greatly affects, you know, the energy consumption. You add on top of that our expertise in liquid cooling and the fact that our company has been devoted, long before it became popular, to being a green computing company and promoting, you know, low power computing. The fact that our CEO has a foundation that's committed to propagation of drought-tolerant trees. So we're, everything that we do is geared toward, you know, energy reduction, right?
So, I guess that was a... Okay. So that's part of our DNA.
David, another investor concern that we've heard is that Supermicro was not one of the reference manufacturing partners for the NVIDIA GB200 NVL72 reference rack. So a couple of questions on that. First, now that NVIDIA has a reference design for racks, do you think that more people will just go to NVIDIA and buy the racks? I mean, what value add can Supermicro provide? Does this... And maybe related to that is, does this talk suggest anything about the evolving relationship with NVIDIA? How strong a relationship do you still have with NVIDIA? And so why don't you answer those two, and I'll ask you some more.
Sure. There's a few questions in there. We'll take them one by one. So first of all, if you talk about the relationship, our CEO, Charles Liang, will be speaking at COMPUTEX at 6:30 P.M. tonight, California time, which is 9:30 A.M. over in Taiwan. So he'll be speaking, and his on-stage guest will be Jensen Huang. So, maybe they'll address some of the things that you might be interested in. But, again, you know, when you talk about that relationship, it goes back 10 years, the fact that we work very close, you know, to their engineers, we, you know, we have no concerns. Some people do like a reference design. Some people like a house that is fully furnished.
They wanna go in, and they don't mind paying more to have all of the furniture in place and the artwork in place and all the lighting fixtures chosen by someone else. But there's a lot of people we've found over the last 30 years that actually don't like that. And they come in, and they say, "You know what? I don't like that painting, and I don't like the carpet." And they want things done their way. And I think if you look at, say, Microsoft, for example, they've announced that they're gonna offer, you know, different, you know, GPU, CPU offerings, because people do like choices. And so... But you know what?
We're glad to build anything that's out there, and we can build it, anything that's out there, faster than anyone else, because we have an engineering-centric company, and that starts from the top. You know, and so, you know, Charles studied engineering down in UT Arlington, and I guess he took his master's degree there, but came over to the U.S. to work on expert systems, oddly enough, which is really, you know, kind of now one of the deliveries of AI.
David, another concern we've heard from clients is, you know, you have a lot of backlog, but do you have the capacity to actually get the revenues, to actually do the manufacturing? Talk to us, how much revenue can your existing footprint support, and how do you see that capacity growing over the next couple of years?
Sure. So if you look at the guidance that we gave for revenue in the June quarter, we gave a guide of, you know, of $5.1 billion-$5.5 billion. Now, that would suggest a run rate, I'm just talking, now we're talking about capacity, would suggest a run rate of $20 billion, right? So we also talked last quarter about a path to $25 billion, and we said, you know, we're finishing out Malaysia, and we're working on other sites as well. So we certainly would aspire to get to $25 billion and, you know, and beyond, but that's, you know, that's all I'll say about that.
Speaking of Malaysia, I mean, is it that, will it be dilutive to margins initially? Because I think, you're building components and not full racks in Malaysia, and how do you see that, that impact evolving over time?
Yeah, actually, we're not building components in Asia. We will be building complete servers, and we will be building, as soon as possible, complete racks and even liquid-cooled racks. So our intentions there are to ramp that side up as quickly as possible, over in Johor, Malaysia, which is about 25 minutes from Singapore. So it's a very good location, you know, in Southeast Asia. And, so we have great hopes for that site. And, but it will, yeah, it will be a full service location.
Great. Maybe, David, I want to ask you about sovereign AI, because, you know, this seems to have the potential for a large, the sovereign entities have a lot of money. So talk to us about how you see Supermicro benefiting from sovereign AI. Are you talking to any sovereign entities? And are there any specific requirements for sovereign AI, which makes Supermicro better suited for servicing that?
Well, I think, besides what we believe is providing the best products, I think that's a good start. But we also think that, you know, we are, you know, we're uniquely focused because of our emphasis on engineering and on being the ability to provide unique solutions. So we not only build servers for, you know, very large enterprises, but we also build very small servers that go into, say, restaurant chains, you know, across the U.S. And in fact, we were awarded a Supplier of the Year a couple of years ago by Yum! Brands. And so we build a lot of different types of servers, you know, across the board. And so, you know, we think that that's a strength.
I want to come back to margins, specifically on liquid cool racks, because I think this is something that most data centers will need over time as the GPUs consume a lot of power and heat. So is the margin range still 14%-17% for liquid cool systems, and why is that? I mean, can you give us some details on what is it that you're making versus what are you going to third-party providers for? And why wouldn't you be able to charge more for a liquid cool rack?
So we absolutely... You know, liquid cooling is definitely an uplift, but it's only gonna be in that 5%-10% range, you know, depending on the configuration of the rack. So it's not a huge uplift in cost. It does cost more, and it takes a little bit longer to assemble and test. And the answer is that we will charge more for that, and we also still expect to get our margins because we put a lot of engineering into the design of that liquid cooling. And we think that, you know, it's a best-in-class solution. So...
And by the way, there's a lot of people that are talking about liquid cooling and having liquid cooling solutions, and then there's people, like I said, are actually shipping it. And so that will soon come out, you know, with who is talking about it and who is actually shipping it, and in what numbers, because we will be ready to talk about how many we were actually able to ship. Of course, as I mentioned, we're still in, you know, we still have supply constraints that we face every day. But we think that, you know, again, that does not change our target margin, you know, at all, the liquid cooling.
David, let's talk about working capital. So to support growth, you need to support a lot of working capital. When do you make the determination, or how do you make the determination that you need to raise more capital? And how should investors think about your, your trade-off between using more debt or using, doing a, another equity raise?
Yeah. So we've had to do a couple of raises, you know, in the past six months because we saw the permanent level, you know, of our business going up higher, so it wasn't temporary. Now, remember, as a manufacturer, if we sell $1 billion, an additional $1 billion, you know, in a quarter, we have to... And remember, when I first started, we were doing about $3.5 billion a year, and now we're doing more than, well more than that per quarter.
When you're increasing, you know, by $1 billion in a quarter, you've got to go out, and you've got to, let's say, even if the margin is 20%, you've got to – you have to buy 80% of materials, and you've got to – you have to carry those, you know, through inventory. You have to carry them through accounts receivable until they convert to cash. And so it becomes an immediate problem. And we've had some very large customers come that I've sat across the table from, and they say, "We have, you know, we have two questions: Do you have the capacity? Do you have the capital to take this project on?" And so we had to go out and get, you know, more permanent capital so we could answer that question always, you know, yes.
We finished out with $2 billion at the end of last quarter. We think that the things that we've done in terms of raising the, you know, the visibility of the company, raising the profits, raising the sales, have been good for our shareholders. We wanna continue to balance that, you know, because we don't like, you know, we don't like dilution. We previously used to repurchase shares. That's still in our tool bag, but right now, you know, it's about being able to deliver against our backlog. Therefore, we will get as much capital as we need to in order to do that.
And so, you know, it's really about whether you see, you know, with short-term debt, we can address, you know, temporary increases. But if we see, you know, sustained orders, such as we have seen, then we're gonna have to do some more permanent debt raises like we've done with the convertible bonds and also with the common stock equity raise.
David, I've got lots of questions and but we're almost out of time. So maybe, maybe talk to us about your guidance philosophy. I mean, it seems to me that the stock can have major wild swings with you just putting out a press release announcing the earnings date. So talk to us about what is your philosophy for setting guidance, and how do you think about pre-announcements, either positive or negative?
Sure. So, you know, obviously, even if you look at NVIDIA's stock, there's swings because the market wants to know, is, you know, is AI here to stay? Right. And so, you know, it, that's that is gradually, we think, you know, being established, but, in the time being, in the near term, you know, we, you know, there's gonna be some fluctuations. But we're not concerned about that. We've been here for 31 years, and by the way, we've been profitable for 31 years. And so, and we have one of the lowest OpExes anywhere at 4% right now. And so, you know, so we run very responsibly. We watch our costs very closely.
So we think that, you know, that we have the right model, and we think that, you know, the time that it takes to, you know, establish AI will quickly be evolving soon. But on forecasts and on guidance, we try to, you know, preannounce if we have, you know, either a downside or an upside, overachievement or underachievement. But if we're within our range, then there's really nothing to report. So that's kind of our overall philosophy.
All right. So I think we covered a lot of different topics. David, thank you so much for joining us today. I really appreciate all the information.
Thank you.