Super Micro Computer, Inc. ($SMCI)

Good afternoon, everyone. Welcome. I'm Samik Chatterjee. I cover the hardware and networking companies at JPMorgan. For the next fireside chat, I have the pleasure of hosting Super Micro and from Super Micro, Michael Staiger, who is the Senior Vice President of Corporate Development. Michael, thanks for being here at the conference. Thank you for your time.

Maybe let's start with the broader question. Super Micro initially to investors was known as a server vendor with the advent of AI, the company moved into being a rack vendor, you now do fully liquid cooling integration. You've now rolled out plans to do DCBBS, which is your data center building block solutions, which includes compute rack, switching, et cetera. So a broad range of capabilities, and that's expanded over time. Just highlight to us or maybe talk -- flesh out a bit more how you're thinking about the addressable market and expanding those? What's the strategy going forward on that front?

Yes. Great. Thanks for having me. So it's interesting that the evolution of the company has been growth and empowering customers and bringing solutions to the market. And if you were to take the view of our platform providers that the market is a $2 trillion to $4 trillion in the short term. And right now, we're 8%, 9% in supplying our customers of that market, and you apply that math to the out years, it implies a $200 billion revenue opportunity for the company. Now if you go a little backwards into the -- into how we get there, and I think we characterized on our recent earnings call total solutions provider, data center provider. So from parts to pieces to servers to racks to more integration. And I think you've heard some others at the conference talk about the speed at which customers are moving and the needs that they have. And we would argue that in between the software and the silicon layer, there's a significant amount of value add to bring to our customers. So -- and we've always talked about application optimization and architecting those systems for those customers to capture that. So it's difficult for me to tell you what the exact TAM is outside of the fact that we'll have a growing piece of it. And from that perspective, we're very well positioned.

Okay. So maybe just keep going on that thread. How do you think about the margin opportunity that, that provides you? As you go more into either DCBBS or more racks than doing individual servers, what is the margin opportunity?

Yes. So we've had an interesting margin progression over the years. Again, I think the -- if you look backwards, there was a significant amount of thrust on getting some customers to market and standing up some large footprints and you can see the power in which they're moving and the speed they're moving. So from that perspective, as our customers now on a continual path of growing and expanding the opportunity set for different platforms and bringing more of the systems together. You recall that there were some pushouts or delays, right? So one of the elements of the pushouts and delays kind of wraps around what we're doing with DCBBS basically total solutioning to the customer, right? So the customer is not waiting. So time to online becomes paramount. So they can get their services in front of their customers so they can drive revenue with the material sums of dollars that they're spending on the equipment themselves. So it kind of is a full-scale wraparound of the strategy that we're deploying for the customers. And so in that speed and that cadence is faster we can get these products to market, the faster we can architect these things at DCBBS, we're talking about 20% baseline gross margins on top of the rack build type margins. And over time, we expect that the solutioning will become a material part of the -- a large part of the business and it will change the margin structure as we go forward. Quite frankly, at the same time, it will be pretty supportive of turns and working capital needs.

Got it. And maybe lastly on that front. Talk about the competitive set of companies that you go up against? How does it change when you look at these different parts that you're more vertically integrating into how does the competitive set change for the company?

So we kind of take a view of a customer approach first, right? So -- and I know everyone out there has a view of, well, there's a couple of OEM competitors that we have. There's ODM competitors. We'll leave names off the table. But when the customers come to us, they're generally looking at how they can get their applications through that system faster. And when I look at what we're doing and the platforming that we're doing for our customers, generally, we're first to market because we can actually turn the systems, the silicon makers, the platform providers of the world, we can get their systems into customers' hands -- and right now, that speed is starting to accelerate. We're talking about inferencing. We're talking about agentic AI, and there's specialty or systems that are developed around those particular platforms that we feel that will lead in the marketplace. And the core customer set will expand and they'll start using some of these things agents will be -- basically could be CPU-based or it could be GPU-based. But whatever the customer needs, we will develop that for that customer. And that -- when you do those things, we don't have a commodity-like build, so to speak, which is what kind of the prior server world, if you want to call it, was thought to be. It's all commoditized. But we're taking these pieces and these parts and components and putting them together, the memory, the drives, et cetera, and making something special for customers that is -- should be margin accretive as we go forward and will wrap DCBBS around it. yes, NVIDIA platforms, we support a significant -- every form factor that they have. We do the same thing with AMD, Intel, Arm, any system, any silicon, any rack, any data center.

Okay. Great. So DCBBS, let's talk about the portfolio there. I mean the offering includes a broad range of solutions, like that solution has a broad range of products you can integrate with it. How are you making the decisions in terms of make versus buy? Do you eventually get to a point where you're developing proprietary IP for most of those components? Or are you looking to buy third party and then integrate it?

So we like to maintain the flexibility, and we've always had the ethos of design and develop it internally ourselves. And so in general, as we move forward, then we'll take other -- obviously, take other components and incorporate those. Right now, some of the activity in DCBBS is around the power, the cooling, actually, we lead on the cooling front, we have some innovations there. There will be memory side cars, the CXL, CMXL kind of opportunity sets, and we'll be able to wrap them around the systems that were delivered to the customers in the form of DCBBS. So I don't think any -- every customer is going to take all the particular components. It depends on what their storage and networking strategy is. But over time, we'll -- and you probably know and have seen, we have relationships with many of the providers and the storage vendors, and we'll build the back end of it and we'll optimize on that front. And those are higher margins in general -- and we will be doing more and more of that and typically shy away from purchasing anything. So we're also not and want to be flexible because some of these architectures will change and some that are wanted is going to be the next thing might actually be the next big thing. So we don't want to make a material bet in any one particular category, but we'll develop and optimize them over time. And let's touch on the software stack and the services. Those things too will be are a part. I think last quarter, we did $140 million in software and services, and we think that we can exit the year -- the calendar year at $1 billion.

Okay. Okay. Maybe let's talk about your competitors a bit more. You talked about the ODMs and how you differentiate them. But how do you differentiate relative to companies like Dell or HPE? And what do you think is the strategic moat that doesn't enable them to do the customization that you're doing for your customers?

Well, we're Silicon Valley based. We're -- I think we've publicly announced that we've expanded our capabilities there. We have significant footprints. We're next to all of our platform providers. We're working with them in design and integration in real time. We haven't outsourced or design portion, the engineering portion or the manufacturing portion. So we generally can control everything that goes into the system element so that we can turn the next generation or the coming generation or quite frankly, make the current generation better with enhancements. So that control plane is very important to a lot of our customers. And so it's a huge differentiator. Another differentiator is scale sourcing, scale manufacturing, scale engineering. And so those things are pretty important. And that's -- I think that's why we're almost 10% of the market.

Okay. Similarly, NVIDIA has expressed interest in more vertical integration over time as well. Do you -- how do you envision that you convince your customers that a customized solution is the better approach than what they would call like a reference architecture, for example.

Well, I think all of our vendors, our platform providers have some form of a reference architecture, right? So the end customer knows what the baseline is. And so it's been broadly discussed at NVIDIA's "taking more control over the design." There's a lot of elements in the design are far more advanced than the prior one. So each step of the way in the evolution of, let's say, server to rack data center has become more complex and more integrated and more difficult. So the co-design, so to speak, is an important piece of this. And so we're keeping pace at that, if not moving it forward. So we'll help them for sure. But at the same time, if you look at what's happening in the end markets, there's a lot of other platforms that are in demand or use cases that are emerging that expand beyond just being tied to one platform, which is kind of how we got to this point. So the differentiation is there, is we're optimizing for all those different platforms, and we'll be well ahead of our competitors in delivering those.

So maybe continuing on NVIDIA. We have this upcoming Vera Rubin platform or right now where they're shipping out the Vera Rubin platform relative to Blackwell or Hopper in the past. How are you thinking about the volume opportunity as well as the pricing opportunity for Super Micro within those deployments?

Well, for starters, I think it's the back half of the year is I think what we've been telling folks. From a standpoint of volume, their volumes could be pretty large. But at the same time, there's a significant amount of other products in the stack that customers need. Not every customer is going to be a Vera Rubin customer. There's other customers that are looking at scale platforms to take advantage of Vera Rubin in the form of agentic CPU, AI CPU kind of -- so we'll integrate all those for our customers. And not only will they integrate, but we'll see if we can put together a solution set for them to run that. So when we get to the Vera Rubin element, we won't be just assembling sheet metal. We'll be delivering solutions alongside with -- to leverage the platform and/or what are the other platforms that customers are going to wind up using to leverage their workloads.

Okay. But how are you thinking about pricing in terms of -- is the step-up similar to when you went from Hopper to Blackwell? Or maybe help us just framework that in terms of...

I don't think there's some generalized pricing that's been tossed around and it is higher. So we'll just work through the transition. We don't have details on numbers at this point in time or guidance for that matter from that product set.

Okay. But it should be good. what are you seeing in terms of customer demand for alternative GPUs like AMD and others? Like is that pipeline? How big is that pipeline? Is it building well?

The pipeline, I would say, is building, is big. I mean it's evident in the marketplace. It's evident in the prices of the platform providers. So all of a sudden, folks are waking up and realizing that it's a much, much larger market in parallel with NVIDIA. We'll see what they say out Wednesday. But the opportunity set there is pretty expansive. We have some recent partnerships that have been announced, the particular Arm stands out. But everyone knows that we're in line with AMD. We have plenty of products with Intel and Arm and there's likely others in the pipeline from a development perspective. I would have to add that almost any silicon provider or a platform provider is coming to us to ask for assistance providing some form factor that it would be application optimized for that particular set. So the demand is just -- it's not a demand issue at all from our perspective.

Okay. Maybe let's talk about the industry discussion around CPUs and CPU-based servers seeing an inflection with agentic AI. I mean when I look at your revenue mix, 80% of the revenues you reported, I think, last quarter were driven by GPU platforms. How should we think about Super Micro's leverage to eventually, if we were to see an inflection in CPU or CPU compute TAM, how would Super Micro be positioned for that?

Well, I think, number one, we ground our position in the marketplace on a CPU-based architecture for quite some time. We were able to put more compute in the footprint at a lower power envelope, a better thermal envelope than anyone else at that time, as you know, enterprises didn't care. Operators don't care. There's different use cases and agentic AI. You think about the agent, and it's a serial process that will go back to the GPU itself and then does this thing. And so it's a serial process. We already have some outstanding systems, liquid cooled CPU systems that are available out in the marketplace. So we think it will be quite helpful because we'll be able to package the entirety of this together, whether there's the model portion and the inferencing portion or the agent CPU portion on top of just general purpose enterprise workloads. So this is a boon to the business and a boon to of the market opportunity.

But are you seeing any of that in your orders?

Yes.

Got it. Customer concentration. In fiscal 2Q, one customer, I think, was 63% of revenue. In fiscal 3Q, you had two customers, 1 27%, 20%. How are you thinking about diversifying the business longer term? And which are the most customer -- important customer verticals that you then need to pursue to deliver on customer diversification?

Yes. So the customer concentration element will be fairly easy to solve. We're already solving it today. We're solving it in a forward look, basically where -- we're designing systems for our customers that we know that will span out. We've already discussed the fact that we'd see customer diversification. We see product diversification. There'll probably be limited -- the pricing competition in those ancillary or CPU-based systems will be a little bit different. In smaller sizes, you won't get the pricing power. So I'll just add another point with -- we have a large 63% customer that I think is now going to be a large customer for another private entity. And so there's -- the demand is pretty intriguing if you catch what I'm saying. It doesn't look like you're catching what I'm saying, but a whole entire data center just went to some. And so it kind of speaks to the volumes of what's going on and the opportunity set here. So as that expands out, we expect diversification to occur. That's the answer to the question. Diversification is going to occur. It's just when, right? The back half of the year, yes.

Okay. So Anthropic becomes a customer is what you're saying?

Somehow indirectly because it's going to lead to your potential next question.

Yes. Because this is one of the pushbacks that I see a lot from investors, which is NeoClouds are the driver of demand that you are seeing some of your competitors are seeing as well. But the argument that they sometimes make is that the durability of the demand is difficult to underwrite. How do you think about derisking when you are particularly taking in orders from NeoClouds, how are you looking at it from the aspect of derisking some of the demand or demand indications they're providing you?

So from a demand perspective and NeoClouds, it's quite -- I think it's quite evident to everyone that the NeoClouds are getting large contracts from the hyperscalers, and there's a dispersion of these workloads across the NeoCloud base that's fueling the NeoClouds and the advent of more and more, so on top of the -- some of the model builders expanding. So the NeoCloud as a class will get bigger. We're seeing that now. Sovereign is coming through, I wouldn't say pretty quickly, but they're also using the NeoClouds and the enterprises are starting to wake up. So the productivity that can be obtained with these tools is significant, and the customers are seeing that. And so again, it kind of speaks back to the diversity of demand and the durability of demand. So I don't know if the -- what the numbers are, but of the amount of people that are using AI or to potentially could use AI is significantly small. The use cases are large. We can all pontificate of what that would look like, but I think it's going to be pretty impressive. And we're here to support all those different types of use cases and make the cost of tokenomics very optimal for our customers. And so we're moving at the speed we like to do that. And again, part of that portion here is system optimization on top of wrapping the solution around that, and that lowers tokenomics for our end customers. And so we're all in on that.

Got it. You did mention towards the end there that the enterprise is waking -- what are you seeing on that front? Is it starting to be material? Is it focused on certain use cases?

So I like the enterprise discussion because the intriguing part about it is the NeoClouds, I mean, the argument is, is CoreWeave an enterprise? I mean, I argue it's an enterprise, right? So from that perspective, the top tier of users in the global corporate base are enterprises, and they're all seeing the higher end -- the higher segment here or the more foot forward enterprises are signing contracts and moving forward, and we're seeing that through our customer base. But the block and tackle enterprise, that's going to buy a couple of servers on a yearly basis. We can still serve them, but we're pretty focused on the customers that are very technology focused that want to drive greater economics in their business. So there are plenty of those customers out there.

Good. You had certain challenges in your fiscal 3Q, there was customer readiness issue that pushed out some revenue into fiscal 4Q and potentially some beyond that, are you expecting to recover completely recover the revenues that were pushed out? And are you seeing as you engage with your customers, this becoming more of a systemic issue in terms of data center readiness that you might have to navigate going forward?

So readiness has been an issue for the past many quarters, right? And there's been some pushouts and movement around, as you know, to architect. Pretty easy to deliver a single server back in the good old days, right, but to deliver an entire data center or a data hall is a little bit of a different emotion and it takes a little bit more time planning. So in some of these cases, the customer -- the power is not ready, whether it's the shell, whether it's power to the building, those types of things that have occurred. They're likely to occur in the future. We try to determine based upon what customer discussions are that when these things will be ready and the time and the window of shipments. The other factor would be a customer that was looking for revenue recognition on the verification of the system working, but the networking switches might not be on site, so you can't verify the system. So from that perspective, those types of things have happened. We try to mitigate those. So it's one of the reasons why our guidance parameters are fairly -- I wouldn't say loose, but have a little bit of leeway to them. And typically, we've made up any -- all the pushouts in 1 or 2 quarters on a go-forward basis. So it's not perfect, but the growth rates that we're putting up are pretty material, and we're trying to match those or meet those.

Okay. Just a heads up, if anyone in the audience has a question, please feel free to raise your hand, and we'll get a mic over to you. But in the meantime, let me just continue to sort of go down this question that I have -- question list that I have. Supply chain constraints. Clearly, that's becoming a bottleneck in a lot of the industry. what's your planning in terms of supply? And what are you doing in terms of either purchase commitments or inventory to be able to have supply assurance? And is it then easy or how are you finding it whether it's easy or tough to pass that higher input cost that you then pay to your customers?

Well, I think that prior to us reporting a recent quarter, people were pretty focused on the memory side of the house and to some extent, the CPU side of the house. And our margins kind of reflected that we didn't have an issue there. We tend to pass through costs and work around in any of those pricing challenges on a real-time basis, and we try to match them to order flow or forecast. So we've had a lot of experience in the supply chain management side. And we're just working through whatever shortages that we have. And in some cases, our customers are helpful in helping secure some of that supply depending on needs and requirements. So it's a challenge, and we're working through it, and so far hasn't really been a material headwind. A lot of work internally to make sure it is not, but, yes.

Any questions from the audience?

Just following on supply and memory. I think the follow-up would be a number of your customers are in a cohort that maybe have more leverage or not quite as much margin to take those price increases over time, especially if supply remains scarce. So just how do you think about that? And is there any risk that there might be friction to passing these price increases along going forward?

Do you want to restate the question?

Question is a lot of customers might not have the -- necessarily the leverage to take that price increase from you. So how does it impact demand?

Well, I think the cadence at the customer level has been that if they don't take the price today, it's going to be worse later. And so they're trying to work around it. So the behavior of the customer, the average customer, I would say, not scale customers is markedly different than what most investors are used to. So I think they're realizing that they have a need, they're going to have to pay for it, at least in the short term. So I think that's a pretty shocking statement of view that that's the case, but that's the behavior we're seeing.

Maybe moving to gross margins. Your execution in F 3Q on that front as well as the guidance for F4Q is much better than at least what the Street was thinking and we were thinking. How much of that is a structurally better execution towards margins or just one-off mix driven that these 2 quarters are a different mix than your usual quarter?

Well, recall that we were absorbing a 63% customer. So you can imagine pricing and et cetera, was a factor and the customer and product mix, what we were doing there had a little bit of a headwind. So I think the majority of the change was working off of that customer. Expedite fees and inventory charges and tariffs were lower in the quarter, and they were lesser of a factor. As we move forward into the current quarter, the customer mix and product mix is generally where we're landing with the current margins. And so that's the driver there. And to be clear, the ultimate goal and the direction that we're heading in is to walk margins up. We're doing that on multiple arenas. One, we have a broadening of the customer base. We have a broadening of the platform. So the pricing pressure on those platforms should be different because, quite frankly, there will be a few competitors that will be able to deliver some of the systems that we're going to be delivering to customers in bulk. And so that -- those things, those factors will be supportive of a double-digit gross margin element as we move forward.

Okay. Maybe just quantify that a bit more. When you say double-digit gross margin long term, is that different from the progress you used to have in the past about 14% to 17%?

So those numbers kind of pinned to traditional compute platform kind of numbers. And so the product set is materially different now. And if and when, as we do a mature DCBBS, those margins can be materially higher. If the platforming the different various platform providers expand and those use cases go up. We're absolutely 100% positioned to deliver to those customers' solutions set. So I don't want to put an upper bound on it, but there's a lot of possibility that the margin profile in a year out would be much different than what we've talking about today.

Got it. And as much as you said, the long-term margin target is double digit, how much of a discipline around what you do to pursue market share versus manage margins? Are we going to see more on a consistent basis? Because one would argue like having a concentrated customer with 63% of your revenue is probably more that you're trading off share, market share instead of taking margin, a better margin, right? So do we see that consistency come in? Or is it still going to be, okay, if there's a big customer on offer, we would rather go for that and focus the margin to be a more margin instead?

So as we all know, is a founder-led organization. He is customer first. So there could be opportunities that pop up that move the needle in either direction from a customer perspective. So what I'm trying to say is we may not walk away from something fairly large if we're able to do it. But we did call out or he did call out that we're focused on a balance of revenue growth and margins. So we'll see. It remains to be seen. But I think the trajectory is the double digits and supporting the customers, realizing the customer base is going to be expanding, the product set is going to be expanding, won't be reliant on one particular build that could have a material impact to the numbers.

Okay. Moving to operating expenses. You had high 20% year-over-year growth in F 3Q. Guidance for F4Q implies further acceleration relative to that. What are you spending on? Where are you directing these OpEx investments towards?

I mean I think the operating expenses is for scaling the business and we haven't made any significant outsized investments on a relative basis. So I don't think there's anything major to call out there. We've been very lean or maybe slightly too lean and pretty focused on not having any outsized purchases in the mix. So I think continue to be pretty focused on controlling operating expenses as we move forward.

I mean is part of the operating expense increase related to DCBBS? Because you mentioned DCBBS is a better margin, but is it a better margin on the operating margin profile as well once you include hiring cost?

It would be implied that it would be, and we may give more color at the year-end when we provide guidance for fiscal '27.

Okay. Good. Maybe just on the corporate side, can you provide us an update on the internal investigations related to some of your employees being involved in selling unauthorized compute to China? Do you expect any financial restatements due to those investigations?

I think we were pretty clear on telling the market that if you read the indictment, that we're not the target. We separated those involved quite quickly. The Board acted quickly, set up a special -- the results have not yet come in from that perspective. We recently filed our Q. So from our view of the incident where there was a falsification of documentation at almost every level -- at every level and even to the federal level, that's an outside very extreme case. And the company is moving forward and serving customers and shipping systems and engaging in new builds. So when there's more details, if there are more details, we'll make them known. But just very fresh situation and we're fortunate again, we were not indicted, and we did not know about the situation until the indictment dropped.

Okay.

Hang a second. And we stated that we expect that there will be no restatement. Obviously, with filing Q is a pretty good indication there.

Maybe a couple more. At the end of 3Q, your total debt was $8.8 billion, cash balance of $1.3 billion, which to us looked like it's the lowest since you've had for a while. How do you then address your working capital requirements with that level of cash balance? And can you continue to support robust growth with that level of cash?

Yes. So we did call out that in the quarter that we had anticipated some payments in the quarter that kind of were pushed out into the April quarter. And so that was, I think, $2.7 billion. So we executed $12.6 billion and change in revenues in the prior quarter with a $4 billion ending cash balance. So you kind of get the sense that those levels should be fine in the zone where we guided to this quarter. So from that perspective, the other element to this is as the customer base broadens out and the deal sizes are a little bit more manageable bites, the turns will be quicker. So there'll be less of a stress on working capital. But swallowing a 63% customer in a period has some ripple effects, and we're going to get past those shortly.

Got it. And last one, have you seen any recent change in credit terms from your suppliers just given the investigation, what's been going on? Have you seen suppliers change at all?

So we're fully engaged with all of our suppliers. And I would say we tend not to discuss those types of things and disclose those. So -- but we're fully engaged with them in helping them bring their products to market.

Okay. Great. I'll wrap it up there. Thank you. Thanks for coming to the conference. Thank you, everyone.