Good afternoon, I'm Samik Chatterjee. I cover the hardware and networking companies at J.P. Morgan. For the next fireside chat, I have the pleasure of hosting Supermicro. With me is David Weigand, who's the Chief Financial Officer. David, thanks for taking the time to come to the conference, and thank you to the audience as well. I'm starting off all my companies with a macro question, or asking you to predict what the future looks like. The question I have for you is, how are you thinking about what the end markets look like in 12 months? We all know where we stand today, and maybe more specifically for you, it's about forecasting what AI-led demand or investment cycle looks like. Share your thoughts.
Thanks, Samik. First, thank you for having us here. And I wanted to remind everyone to check our cautionary safe harbor statement regarding risk factors and forward-looking comments on our Supermicro IR website. So, for the first question, what I would say is that we're still early stage, we believe, on the investment cycle. And, you know, when we see some of the recent capital raises domestically, both on equity and on available debt lines, there's still some very large numbers being put up. We also have knowledge that both sovereigns as well as other enterprises in EMEA, as well as Europe, are really starting to plan a lot of investment in AI. So we believe that in 12 months, things will still be going at a very strong track.
Okay. On the flip side, and just to sort of take the other side of this argument, a lot of investors are concerned that one day we wake up, and everyone realizes that the returns on these investments are not material enough to continue to invest towards AI. How do you think about those risks? And I mean, to some extent, investors talk about it being similar to the dot-com bubble in that sense. How do you differentiate versus that and say, "Okay, no, this is definitely different because we have line of sight to XYZ?
Yeah, you know, what accelerated computing has done is it has sped up the time, the times that normally took applications to run. So in other words, if you look at weather forecasting, for instance, I mean, what accelerated computing does is it speeds up the time in which you could make more accurate and timely forecasts. So I think, you know, that is a... That's a real business return, so is the reduced time for drug discovery and things like that. So there's so many applications, this is not, you know, this is really a game changer, you know, it's a, it's a, it's a paradigm shift. And so, you know, we don't see that as a high risk.
Okay. Okay. Before we go forward with the AI demand questions, and we obviously have a lot of them, but more internally, how do you think about using AI or leveraging AI internally in the organization? Do you see enough tangible improvements in-house as well when you are leveraging AI? Just share your thoughts and sort of what you're seeing, because I think that in itself is the confidence level in terms of the sustainability of that demand, right?
Yes, yes. We're using AI internally in a number of areas in the business, both for in R&D and in testing, as well as in you know quality control, marketing, and even sales. And so, there's a lot of companies doing that right now. They're exploring you know ways that they can use AI in their business. Some of them are doing that through hyperscalers or CSPs. Others are you know going ahead and taking it into the enterprise and developing those applications-
Okay
... for customer service models or otherwise.
I mean, from a timeline perspective, do you expect those improvements to be tangible and something investors can track in the near term? Or is it much more of a longer timeline in terms of when you see those productivity improvements?
I think that there's gonna be both, near-term and long-term, you know, evidence of returns. And so I think, you know, we're just starting to get metrics on those now.
Okay. One of Supermicro's key differentiation relative to the rest of the industry has been the speed with which you get products to the market. That's raised a lot of questions around how sustainable that differentiation is, when supply lead times start to moderate. Share your sort of perspective on that when you think about the current supply environment, and if availability of products like GPUs improve, how sustainable is that differentiation that you have?
You know, actually, it improves, because I would say that if we had unlimited supply chain, our edge in being first to market could have been exploited a lot more. So therefore, as supply chains smooth out, as silicon is more available, as your lead times go down from 52 weeks to 36 weeks to, you know, maybe, you know, 4 weeks, 6 weeks, our ability to take products to market gets better, and it's because of our engineering prowess that we're able to take products to market quickly.
Okay, interesting. Talk about the modular approach for your systems. How do you think of that as a differentiation, relative to some of your competitors? Have you seen any competitors take a similar approach, or how many years of investment and R&D would it require for your competitors to take the same modular approach that you have with your systems?
... Yeah, you know, it really, it took us almost 31 years to get here. And we have a very engineering focused company, so half of our employees are engineers. We also have a CEO who is, you know, himself one of the best engineers, you know, around. And he actually, you know, drives our culture, and he drives the engineers to build some of the very best products, you know, in our industry. And that's really our edge, is that we have a focus on, you know, high-quality products.
We have a very low, you know, OpEx. We have about 4%, and that's what allows us to do what we do.
Okay. You have greater than 10% customers in most quarters. Help us think about how concentrated or diversified the demand you're seeing from the infrastructure side is right now, and how are you thinking about broadening out the customer base?
You know, there's always... You know, with the developments in AI, there are those companies that are gonna be disruptors, and then there are those companies that are gonna be disrupted. So some of the early disruptors are really pursuing this fast. And so they, they've emerged as kind of our high 10%+ customers, and we're glad. We've had four of them, you know, over the last four to six quarters, and we expect to have more. But we believe that as AI moves more broadly into the enterprise, it will continue to filter down to the rest of all midsize and larger enterprises.
We think that will, that will diversify.
Okay. Okay.
But right now, people are using, you know, CSPs, for instance, to walk into the AI marketplace.
Okay. So one of the, I think one of the comments you've made historically, if I get it right, is that you're under-indexed with hyperscalers, but have done really well with Tier 2 cloud and companies, which in some cases are selling the capacity back to hyperscalers. What has driven that customer dynamic? I mean, which companies do you see as primary competitors when you go up against in terms of bidding for contracts with the hyperscalers versus on the Tier 2 side? Is it a different competitive landscape as well? What has driven this market, your sort of indexing being higher to the Tier 2s than the hyperscalers?
You know, the market has grown so fast. I mean, first of all, if you go back and look at what we've done with AI, we've actually been growing with AI over the last two years. You know, we didn't just recently discover AI. In fact, you know, our CEO, Charles, he built a supercomputer, you know, for NVIDIA to develop AI, 10 years ago, back when you know, when Jensen was talking about the GPU becoming a prominent player, having a prominent position in the server. So we've been working on AI for a long time, and it's driven our revenues the past two years.
And now with large language models and ChatGPT, it's, you know, it's kind of its growth has obviously expanded, you know, exponentially. And so, so we think that, you know, that will continue and, you know, and go on.
Is it a different competitive set when you go up against for the wins in the hyperscalers versus the Tier 2s, or is it pretty much the same competitors you run up against?
Well, you know, in this market, a lot of the hyperscalers are now getting capacity through CSPs, and so, and we sell to CSPs. So, you know, so indirectly, and in some cases directly, we're supporting, we're selling CSPs to hyperscalers.
Okay. Let's move to talking about NVIDIA, which is a large supplier. As a company, you have a great relationship with them as well. How does the transition of product generation in relation to NVIDIA's transition here impact demand for Supermicro products?
Yeah, so we believe that, with respect to GPUs, we have, you know, we have a number of great partners. You know, we have NVIDIA, we have Intel, and we have AMD. And so, we always wanna make sure that we're bringing the very best products to market. And by very best, I mean, you know, the highest performance, you know, per dollar or per watt, the highest reliability, and also, you know, the great things that, the great support that comes with selling those products. So that's really our goal, and we want to partner with all those, all of those companies, to bring the very best solutions to market.
And we have always found a way to innovate, and we have to continue to do that, you know, in the future as well. But I think with our engineering, you know, based culture, we think that, you know, we're very optimistic going forward.
... when you, maybe going back to that question, when a product transition happens for one of your large suppliers, be it Intel, which will obviously have product transitions once they are more, relevant in this market, do you necessarily see it just as a pricing opportunity, or do you see it as a volume opportunity as well with your customers?
Yeah, it's both. I mean, certainly the ASPs have gone up, you know, with-- Really, anytime you have a technology which, you know, which returns more to the customer, then you're gonna have an increase in pricing. You know, it would be like going to a, you know, a 1500 horsepower motor, you know, from... And you're gonna be able to charge a higher price. But we think also volume, because everyone sees the potential for, you know, for AI in, you know, in their applications, and we think that'll drive higher volume.
Okay. What about this concern that the market has, that there will also be pauses from the customer base in terms of buying before the product transition? What do you see in that regard? Are you seeing customers act that way? What have you seen?
Yeah, there is always gonna be some customers. And this used to happen when, you know, back when there was, you know, tick-tock changes in, you know, in Intel CPUs, for instance. So some customers will choose to wait for the next technology, you know, that comes out. But again, you have to go back to the disruptor or disrupted model, because, you know, some companies don't. They fear that if they wait, there's gonna be a delay on general availability, there'll be problems with the general supply chain, or, you know, that they will just have their competitors overtake them, and so they don't wanna wait.
So there's a balance of the desire to wait for the next thing that's coming out, versus the fear that you're gonna lose, you know, traction in your market because you waited. So those things help to balance, I think, the sales.
Yeah, got it. NVIDIA also has expressed interest in doing full system solutions-
Mm.
rather than just the components alone. How do you think about that impacting your TAM and the addressable market you can think of with some of your customers?
Yeah. I think that we always manage to innovate, you know, and we've done that with you know the H100. We've done that with you know the MI250, and we've done it with Gaudi 2. Those are all products that we're shipping. So we always find a way to make our solutions the very best, based on the technologies that are out there. So, you know, some... I think that's really the—Some customers want a turnkey solution, and they're willing to, you know, to pay for that.
You know, really, a couple of years ago, people used to ask us: "Well, you know, isn't everything going to go to the cloud?" And we would say, "You know, we think that that's probably gonna play a part, but, you know, there's still gonna be a place for us." And what we found is that indeed, customers want choices, and they don't wanna be, they don't wanna be locked into just one solution. So we think that the market is very large, and we specialize in customized solutions. So we think that customization is definitely appropriate in a disrupted market such as we have now. But what I mean by disrupted is that there's a lot of really good technologies that are coming out, and that it's a positive disruption.
Okay, great. So this goes to your point about innovation, but there's also been the discussion about NVIDIA changing partners in relation to reference designs from the prior generation to the next generation. How does that impact your position in the market, given that you were one of the reference designs last in the prior generation? How do you innovate, and what do you provide as incremental sort of features to the customer with your customization related to a reference design?
Yeah, yeah. Again, we always, we always find a way to design in the very best features of, you know, of every new, every, every new technology offering. And so we'll, we'll do the same, you know, and we'll do the same with all of the new releases that are coming out from AMD, from Intel, and from NVIDIA. So, we, we welcome those, we welcome those new technologies, and we believe that they're gonna ultimately benefit our customers, and, and so we, you know, we look forward to, to coming out with full releases.
Okay. Moving to then the alternate suppliers, Intel, AMD, what are you seeing in relation to customer interest, as well as any sort of things you can share in relation to pipeline of opportunity relative to those suppliers when we start to compare to NVIDIA as the incumbent here?
Yeah, I mean, you probably saw the announcement by Microsoft. I mean, they have an interest, I think, in alternative technology. And that's just normal, there's normal technology choices that every company makes. And then, you know, we actually have, we've prepared. We'll prepare whatever CPU, GPU, or combination thereof that our customers want. And by the way, not every customer needs a GPU, depends on their workload. So, whatever optimizes the customer's, you know, particular need, that's what we'll build for. And so we're, you know, we're glad about all of the technologies that are coming out. But we have a very close relationship with all of our technology partners.
We work, you know, in Silicon Valley with our engineers, and, you know, we always are looking forward to the next things that are coming out. And you asked me about the pipeline, so we still see a lot of interest. We have a lot of interest in, you know, the MI250, MI300, you know, Gaudi, Gaudi 2, and Gaudi 3. So we're looking forward to those, you know, to those orders.
Okay. Before I move on, let me take a pause and see if anyone in the audience has a question they want to ask. Any questions? Okay, so let me continue, and we can check in again in a few minutes. Revenue drivers for Supermicro, how do you think about the next—when you think about the next few years, how do you think about the drivers between pricing and volume for the company?
Yeah, we think it's gonna be both, you know, higher volume and higher pricing as well. You know, because there is, you know, there is no doubt about the fact that accelerated computing is here to stay. I mean, you know, it's not even a question that a server runs better in many applications, not, again, not every application, but in many applications with both a CPU and a GPU, it just increases the throughput. So, you know, the battle is over for a large part of the market. And so it's. You know, we think that that's gonna help to continue to help us.
Okay. So on that pricing discussion then, you've acknowledged your willingness to be aggressive on pricing to solidify market share-
Mm-hmm.
At this point. Then how do you envision pulling it back at some point in the future? Once, let's say, the supply chain constraints are done, GPU availability is pretty normal. If you are aggressive at this point, how do you, at a later point, pull that back?
Well, you know, we, the reason that we've been able to grow, you know, through the nine months ended March thirty-first, we grew, you know, 95% over the prior year's nine months. We, our CAGR over the last two fiscal years has been, you know, approximately 42%. So, you know, AI has meant a lot of, a lot of growth for us, but more importantly, we have, we've, our, you know, our position in the market has risen up because, you know, we're a company that people don't look past anymore.
So therefore, you know, going forward, we think that, you know, our improved brand is assisted by being a little bit competitive on pricing and getting a beachhead, getting some good installations. So we've done that with liquid cooling more recently, and we think that we're one of the early companies to ship, you know, higher volumes of liquid-cooled racks, you know, at scale.
Okay. I'll come back to your point on liquid cooling in a bit, but interesting you mentioned the improved brand. How do you expect that to play out when it comes to enterprise demand on the AI side? I mean, traditionally, the enterprise market has been branded the Dells, HPs of the world. When you now consider Supermicro going into that market, the enterprise demand relative to AI picking up, how do you think about what your share, fair share in that market should be?
Yeah, you know, we're focused just on growing faster than the market. You know, we want to, we wanna... Our goal is to always grow faster than the market. I think that's our target. So, we think that, if we continue to focus on innovation and delivering, you know, quality solutions at a fair price, we'll do that.
Okay. Maybe just to rephrase that question, the branded companies have done well with enterprise, but they've had to invest a ton in the channel.
Mm-hmm.
Where do you see Supermicro today positioned relative to that? What's the path forward in terms of investment to then drive that share with the enterprise?
Yeah, so currently, by the way, our enterprise channel represents 50% of our business, and then we actually have, you know, we have 10% customers, you know, in that have appeared in that vertical. So, we think that we've already been recognized as a player in that space, in the enterprise space, and we think it'll grow, it'll continue to grow.
Okay. Let me just, again, pause and do a quick check if there are any questions in the audience. Okay, so let me continue on the enterprise part that you mentioned. There were a couple of reports last week of share loss to one of the incumbent branded players relative to one of the enterprise customers that's been your 10% customer in the past.... How should we think about the likelihood of share loss? And again, the compare is, how much of a read-through is that into competitiveness of the market in relation to pricing?
Yeah, these, you know, the good news is, Samik, that we are-- we're actually gaining a lot of new customers right now. And, you know, we're quite happy that some customers are or some suppliers or competitors are have figured out that AI is here to stay. And, you know, we expect that there will be some additional customers that we'll gain also from this. And I think that we're seeing a lot of development over in Europe and the Middle East, as well as Asia. And, we... In fact, our Asia sales grew most recently. So we think that we're not concerned about, you know, losses at any particular customers. We have very good relationships with our customers.
Mm.
In fact, if you look at our charts, some of our 10% customers are repeat 10% customers.
Yeah.
We're not, we're not concerned about that.
Okay, great. So one more thing from the earnings call, and then we'll take this question that come in online. You hinted at margin dilution in your fiscal fourth quarter, ending in June. Can you clarify if you imply, you're implying that the fourth quarter margin could be below the long-term guidance for margins that you've had, which is 14%-17%? And why, if that's the case, should we think it's a temporary headwind and not really representative of a longer-term trend for the company?
Sure. Yeah, so our guidance is, you know. Our target is 14%-17%. So in any, you know, in any one quarter, that can fluctuate, you know, up or down, and so, we're not gonna update our guidance. We still are targeting that 14%-17%, and we still think that's a good long-term target. Our operating margin has been hovering over the last year at, say, 11%-11.3%. With a 4% OpEx, we believe that as we continue to grow, we will, you know, continue to get operating leverage. And so, we're, you know, quite happy to continue to deliver that.
Okay, and why, when you're in the scenario where you dip below 14, should investors think that's a temporary headwind and not a longer-term trend?
Well, you know, we've liquid cooling was an opportunity for us to, you know, to take an early lead, and so we mentioned that we were gonna be shipping 1,000 racks of liquid cooling, liquid-cooled solutions. So we think that we wanted to establish a beachhead, and then build in and out, you know, at, you know, at our customer's site. So we think that we will opportunistically take lower margins, but it's only so that we can, you know, attract more customers and more and more design wins.
Okay. The question here, on the portal is, it reads: Can you talk about your recently announced as a service product called Zero Gap with Vapor IO?
Zero Gap with Vapor IO. So I'm not, I'm not familiar with that recent announcement. I'm sorry.
No, we'll skip that one. All right. Okay, so let's maybe talk about the competitive landscape here a bit. When you compare Supermicro technical capabilities to the ODM companies in Asia and EMS and ODM companies like Celestica, in the Western Hemisphere, how, how would you compare what you offer relative to some of those ODM companies, to the sort of the hyperscaler customers or the tier twos? How, where is this differentiation?
Yeah. So, we have, you know, half of our workforce is engineers, and I think, whether whoever you compare us to, I don't think you'll find another company that has, you know, that ratio of engineers as a part of their workforce. So that's really Supermicro strength, is its engineering prowess, and that's a big differentiator between us and anyone else, you know, in the market.
Okay. Do the same when we talk about branded companies, Dell, HPE, how do you see the differentiation? Is it just the fast time to market, or do you see other layers of differentiation?
You know, I think it's going back to what I talked about earlier, the kind of our engineering culture. We, we are able to respond really quickly, not just because of, of that engineering prowess, but also our basic approach to building servers. So we don't try to build servers that we can, you know, kind of a one-size-fits-all, or we don't try to just sell a limited number of SKUs. We look at, at building servers as a, as a custom fit for, you know, a particular, customer. And, so we, we'll often build multiple servers even for one customer, depending on the workload that they're using those for.
Okay. We'll talk about capacity in a bit, but first, capital raises, and you've done a couple here to buy inventory of critical components.
Mm-hmm.
How should investors think about the need to pursue similar pre-buy of inventory if GPU lead times are now subsiding? What is that sort of level of growth that Supermicro can sustain without really needing to come back to the markets for a capital raise?
... Yeah, so we have, you know, we've grown very fast in the recent last couple of years. So, you know, profitability alone can't provide enough working capital to fund, you know, really high growth. So we've had to go to the market to raise additional capital. The good news is that having become an S&P 500 company and having, you know, attained the financial metrics that we have, we believe that we are, you know, an IG profile company. So that means that we probably will move to a more unsecured debt. That will help us, you know, to make some opportune purchases for sure of inventory and things like that, without having to go to the capital markets every time.
But we'll be, we'll be selective about that.
Got it. Got it. Capacity, what's the level of capacity you currently have in terms of overall racks, as well as liquid cooling capacity that you have? And, how do you, sort of, how should we think about that growth and capacity from here on?
Yeah, so we estimate by the end of June thirtieth, which is our fiscal year end, we will have a rack capacity of 5,000 racks per month. And we will have a liquid-cooled rack capacity of 2,000 racks per month. And so-
How, how should we think about investments to increase that capacity?
Uh, investments-
To increase capacity on an ongoing basis.
Yeah, we have Malaysia, we have a site that we're building in Malaysia that will come online in July. We'll start production and we will really be able to double our worldwide capacity eventually at that site. So, we have the capacity to grow quite a bit more.
Malaysia will do both general racks as well as liquid-cooled racks?
They will have the ability to do either. Yes, that's right.
Okay. Got it. Got it. Last couple of questions. Actually, let me just do a quick check again, if anyone has a question. I see one back there, so maybe you can take that one, if there's a mic.
Thanks for letting me ask a question. I think in your guidance, you mentioned that there's about $1.6 billion of projects that will be satisfied in the June quarter. Can you talk about how much inventory you will have going into the back half of the year? If I look at the growth in your balancing inventory, it's over the last, call it three quarters, you will consume materially all of that growth for your projects in two Qs. So can you talk about both your supply side into the back half of the year, as well as the business that you have? Is this $1.6 sort of a one-off that we should think that needs to get backfilled in the back half, or is this sustainable business for...?
Just how should we think about that dynamic part of your guidance into the back half of the year?
Okay. So we actually have a very large backlog. We haven't talked too much about that, other than to say that it's at a record level. We think that, you know, we think that we've given pretty good guidance on our June quarter, the current quarter that we're in. And we've also given some guidance even as to, you know, the quarter after that in September. So we will have to, you know, maintain a high inventory level. And we actually hope to—we hope that at that high inventory level continues with... You know, we can't make any forecasts right now, but we're trusting that it'll continue to need to stay where it is.
Great. I'll wrap it up there. Thank you, David, for to make it to the conference. Thank you to the audience as well. Thank you.
Thank you. Thank you, Samik.