Hello, everyone, and good afternoon. Thank you for joining us for the second day of Needham's 27th Annual Growth Conference. My name is Nick Doyle, and I'm an analyst on the semiconductor team here at Needham. It's my pleasure to host this fireside chat with Penguin Solutions, the company designs, builds, deploys, and manages high-performance, high-availability enterprise solutions across its computing, memory, and LED lines of business. Joining me from the company is Nate Olmstead, CFO. Nate, thanks for joining us.
Hey, Nick.
A couple of introductory questions. Nate, you're relatively new to the company. Can you provide a quick overview of your background and what compelled you to join the company?
OK, sounds good. Yep, so I've been here about six months now. Prior to this, I was a CFO at Logitech for four years. And prior to that, I was at Hewlett Packard and Hewlett Packard Enterprise for about 17 years, 18 years. You know, when I was looking for a new place to be CFO, I was really interested in someone who was going to be addressing the AI market in some shape or fashion, and someplace that was undergoing a business transformation, because I find that that type of work is really interesting. And anytime there's change taking place, there's a lot of opportunities for value creation. And so Penguin, at the time, was called SMART Global Holdings . It kind of fit both of those things very, very nicely. We were transforming from a holding company to an operating company, so a lot of business transformation.
And at the same time, our largest and fastest-growing business was addressing AI infrastructure, as you said. So that's what kind of got me excited about it.
OK, makes sense. For any new investors out there, can you explain what the company does with a focus on that advanced compute segment? What capabilities do you have that enables you to grow in this highly competitive data center market?
Yeah, so as you kind of mentioned, we are kind of a mixed bag of things because of the holding company history. But if we focus on the compute side, which makes up, like I said, just over 50% of the revenue now, we do focus on each element of that value chain, so from design to deploy, manage, and no, I missed one in there.
Build.
Build, sorry, I missed build right in the beginning. This is DBDM. And the thought is that we have expertise of over two decades with high-performance computing. The history of Penguin Computing, which was an acquisition the company did a few years ago, was in high-performance computing. And these are the most complex compute environments that existed kind of historically, really large clusters for things like weather forecasting or Department of Defense use. And it turns out that that experience is really applicable to what's taking place in AI these days, and that these AI clusters, which are GPU-centric computing, are really complex, especially around the networking, the power, the cooling requirements.
And so this history that we have of doing high-performance computing for more than a couple of decades is very applicable and puts us in, I won't say unique, but in rare company in terms of who can really help customers solve the complexity of these large AI deployments. So again, I would think about us more as a solutions provider, a systems integrator than a hardware company. We're pretty hardware agnostic. We do have our own servers that we sell, but they're really, to be honest, not differentiated versus what else you would find out in the industry. We are just as happy to be selling a Dell server as our own server. We sell Pure Storage, VAST Data, Weka, Arista networking. So we're really a solutions provider, customer-centric solutions provider, very agnostic on the hardware side.
I think our key differentiator is, again, that experience and expertise in highly complex infrastructure integrations.
Yeah, and I think, something we just met with Celestial AI a couple of minutes ago. And so, just speaking to this integrator capability that you have, I mean, they're out kind of showcasing this unique optical memory appliance, and you've kind of been the spearhead of the integrator. So, I guess, why would they choose you and using that to kind of show how you're capable compared to your competitors?
Yeah, and that really takes us over to the memory portion of our business, which is about 30% of our revenue. And it's kind of the legacy of the company, SMART Global or SMART Modular. With Celestial and the work they're doing, we sell specialized memory, so we're not in consumer memory, specialized enterprise memory solutions. Think about like inside Cisco routers and switches and the boot drives or the memory components that are inside there. Those would come from SMART Modular. And in the case of the work you're mentioning around the OMA or the optical memory appliance, this is really a totally different way to think about how memory is used in a data center. This would be a way to pool memory and have it be addressable across multiple GPUs, servers.
So rather than being on the boards themselves with the GPUs, you can separate the memory, connect it optically to the GPU or to the motherboards for really low latency, but you can pool that memory so that it's addressable across multiple servers. And that's kind of a unique way to think about memory. It's similar to the transition years and years ago to network-attached storage or thin provisioning in the storage space, if you will. So this is something we're pretty excited about. It's a new area for SMART
We're in our fiscal year 2025 right now. But I think it's something that we're pretty interested in. In between now and then, there's some first- and second-generation products that we talk about around CXL technology, which is Compute Express Link. Again, it's a little bit different than what we've done historically, which is exciting for us. It's an opportunity for us to drive some incremental growth and incremental margin from what we're doing today. Here, the CXL, before we get this optical link set up with Celestial, is we can address some of the density problems that exist in compute today around memory. There's a limited amount of DIMM space or memory space on the board close to the GPU to really for the GPU to access it at really low latencies.
And with an add-in card that we've been working on and developed, customers can increase the density of memory that's in their servers dramatically at a much lower price than having to buy an incremental server. So it has a real TCO or total cost of ownership benefit. Again, something that's different than what we've done in the past. It's a higher margin opportunity for us and incremental growth for us. And we have a little bit of revenue from that this fiscal year, but more meaningful next year. So in 2026, we'll get some of the benefit of these add-in cards. And then in 2027, I think we can get all the way out to the memory appliance.
Great. You rebranded the company in July of last year. What's been the biggest change, expected or unexpected, from how your customers and partners view the company?
I think one, if you just think about the name. So we were Smart Global Holdings. Penguin Solutions kind of sounds like you're solving problems for penguins, but it's better than Smart Global Holdings, which just didn't really have much buzz to it.
It wasn't buzzed.
Yeah. And so I think and Penguin Computing is the legacy brand. So within the industry, we had some brand recognition and awareness around Penguin Computing. But I think now that we've rebranded the entire company as Penguin Solutions, it really indicates our focus on changing to an operating company, because we've gotten rid of the holding name, and our focus on Penguin Computing as being really the core strategically of what it is we're trying to do. So it's just part of a transformation, and I think it just gives us more awareness. I think it's gone actually spectacularly well so far. Again, probably coming from a place where we didn't have a lot of brand recognition to begin with, so it's not a heavy lift. But really pleased with how it's gone so far.
As a CFO, I'm pleased it hasn't cost us a lot of money to go do it, and so I think we're getting a good ROI on it.
Great. You just reported fiscal 1Q25 results last week. We came away pretty positive. My one-liner would be fiscal first half numbers are largely de-risked, and that lets us feel better about a more conservative fiscal second half. Would you summarize it differently?
I think that's true, and I think it comes down to what visibility we have, so obviously, we reported Q1, so that's in the books, first 90 days of the year. We do guide on an annual basis rather than quarterly now. It aligns better with sort of the flow of business that we have, which can be kind of lumpy. As a smaller company, we have some large deals. We could book something one quarter and not the next, and so the quarters can look different, and we don't have a ton of seasonality in our business. So quarterly guidance has a lot of fluctuations, but it smooths out over the year, and so we put the first quarter in the books, a good quarter, kind of as expected, as we thought, pulled forward a little bit of revenue, but really as kind of we expected.
Q2, we've got really good visibility to essentially have everything booked today for that quarter and a good amount of that work done. But the second half of the year, we still have some bookings to go generate. And so we're only going to give an outlook that considers what it is we have really high confidence in. And so the first half of the year, we said we thought would be probably stronger than the second half. And based on what we can see today, I think that's true. There's opportunity for the second half to certainly be better. Over the next eight weeks or so, we'll have probably the remainder, the great majority, if not the remainder of the bookings for the second half complete.
And so at that point, when we go out after Q2 earnings, we'll be able to give a crisper update, probably tighten the range on the outlook for the full year and indicate whether or not we think we'll be at the midpoint of that, maybe the low end or the high end. We'll have that information available by the time we get to Q2 earnings.
Yeah, that's great, and then just as a little background, I believe you were giving annual guidance at Logitech as well, so a little more experience there than usual.
Yeah, that's the way we did it there at HP. We gave quarterly and annual at both. HP is a much different business model than what we have here. It's the same industry, and it's hardware and services and some software, but they had a run rate of transactional server business that might have made up 70% of the revenue. We don't have that kind of business here. We're really deal-centric, more binary. And so we don't have the same linearity, if you will, that HP does.
OK. A couple of broad questions we're asking our management teams here. It's NGC. If the Trump administration increases tariffs gradually on Chinese imports or really any other country, how would this affect your business?
So to China specifically, it really only impacts our LED business. So we're facing that today with the existing tariffs. And like most companies that have that sort of exposure, we've been trying to move supply chain operations to other parts of Southeast Asia and have been doing that pretty successfully. We had less than $10 million of tariff expense annually for LED. And some of that may even be recoverable. Like most companies, again, will petition to try to recover some of that based on actual country of origin for the majority of the value add and things like that. So it's a fairly contained and somewhat limited risk for us today. If the Chinese imports grew, again, it would impact the LED business. If the tariffs were applied more broadly, then it would have certainly more of an impact on us like for other companies.
We have a lot of manufacturing of memory. It takes place in Malaysia, for example, so that would be one area that we would look at. We also do have manufacturing for memory in the U.S., however, so we have some ability to shift where some of the work is done, if necessary, to try to offset some of those tariff impacts. The other thing I would say is, to the extent that the administration's tariff strategy goes hand in hand with reducing corporate taxes, right, so instead of taxing corporations, they want to tax, in theory, other entities. With tariffs, we may see some favorability on the tax rate, so trying to position ourselves to take advantage of that if that's the case.
The advanced compute capacity, is that in the U.S. as well?
Yeah, we do. It's U.S. manufacturing. It's actually in California.
Great. So would it be fair to say this Made in America trend would be beneficial to the business overall?
I think that's a relative question. I think yes, relative to a lot of other players in our industry in that we don't have operations in Mexico or in really, we don't do compute activities out of China. It's tricky, though, because you could have small components that come out of China, as it does in a lot of the technology supply chains, and depending on the structure of the tariffs, you could have impacts, but I don't really think I have a major risk to manage in advanced compute, and I think made in America, yeah, I think relative to other companies that I can immediately think of, I think it's probably a favorable item for us. We've done a lot of work in the past with the federal government. It's less of a target customer for us today than it was five or 10 years ago.
But as a result of that, we kind of have some operations that are set up to serve government customers. And we could potentially expand that if necessary.
Yeah. Across your three end markets, where do you still see pockets of inventory? And how do you see that inventory at direct customers versus the distribution channel?
Each of the three businesses has kind of a different inventory structure. Advanced Compute, we don't really move anything through disties, so to speak of. It's mostly inventory that we're sitting on our books rather than out in a channel. And I think that is healthy. Because the deal life cycles are pretty long, from the time that we win some business till actually shipping, that can easily be six months, maybe even nine months. And so we aren't really having to pre-position inventory and hold things in raw materials waiting for deals, right? It's build to order, not build to stock. And so inventory is pretty lean on that side. We grew inventories this last quarter in anticipation of having some shipments in Q2. So we mentioned that on the earnings call. Memory, we've actually seen there, we do work through distributors.
A lot of our customers, we have large enterprise customers like Cisco or someone like that I mentioned, they sort of play the arbitrage game. So when memory gets cheaper, they'll stock up. And they did that a year or two ago. And they've been working down that inventory. We're now getting to a point where they've largely consumed that excess inventory that they acquired. And so we're starting to see purchases now more aligned with end user demand, which is a good thing for us. We've seen sequential growth in the memory business each of the last three quarters. And we think it's going to grow 10%-20% this year, year-over-year. So we're really kind of getting to a normalization now on memory. And then in LED, again, it's a distie multi-tier kind of channel model.
The channel inventory there, we have in good shape. It was high a couple of years ago, I would say, for similar reasons. There was, if you remember back during the COVID days, there was kind of limited visibility to really what the true end user demand was. And so a lot of companies ended up with excess inventories out in the channel. And I think our LED business had similar problems, but that's now normalized itself. And when I look at the weekly data, I can see that my sell out and my sell in sort of match each other. So I feel good about the inventory across the board.
Yeah. Yeah, much more positive sentiment here versus a lot of our industrial or automotive type companies. So something we talked about CES last week that really got my attention was that you haven't lost a customer in advanced compute. So I mean, why do you think customers stick with you? Is there more stickiness in the services and software portion of the business? Is that maybe a bit of an explainer there?
Yeah, I think it's a couple of things. I think, as I mentioned at the outset, what we do are the most complex things in the computing space, and so where there's complexity, I think the value proposition of services in general, whether you're talking computing or elsewhere, I think the value proposition of services is stronger, and we do see ourselves, like I said, as a systems integrator and a solutions provider, and so the services piece of our business is nice and sticky. Helping that is that we have some software assets that we use as part of our service delivery, which makes the services even stickier, so not only are we adding a lot of value with the services we provide, but the service delivery that we do utilizes some software to do predictive and proactive support, which are hard things to replicate.
Our number one competitor is probably in-house IT departments, and again, I would say the number one competitor for most services businesses are something being done in-house, and that's true in IT. Most in-house IT departments don't have the skills to do what we do at the most complex levels, which is even why a customer like Meta, who is our largest customer, who has obviously great resources in IT, has year after year renewed with us and actually grown their business with us, so I think it's a good testament to the quality of the work that we do and the value proposition we have, and I think those combined become the reason why we haven't lost customers in advanced computing. We don't have really a services churn problem. We get strong renewals, and again, I think the software that we have is one of the things.
Why we're investing in software organically is to just strengthen that moat, if you will, around the services business.
Yeah. Yeah, again, it is just really interesting because we get asked that a lot. We see this lumpiness in the advanced compute business. So investors are wondering which customer fell off and who's leaving for the competitor. But really, it's just the lumpiness of the hardware business specifically that's kind of driving those results.
Yeah, so the hardware gets recognized upfront. And every hardware company has to, if you do $100 million of hardware revenue one year, for the next year, the first thing you've got to do is find $100 million of hardware revenue before you can start growing. And so I think in markets where things are growing rapidly, like we have right now in AI, the good news is that $100 million I did last year, you might repeat with those same customers next year because the technology is evolving so quickly that the refresh cycles are shorter. When industry slowed down and the refresh cycle lengthens, that $100 million in year one, I may not refresh again until year three or year four, right? So we're in a refresh cycle right now that is favorable in that the technology is rapidly evolving and the demands are increasing.
So repeating last year's performance is perhaps a little bit easier than it has been in other places I've been. But yes, the lumpiness comes from I may land a $50 million deal one quarter. And if I don't have a $50 million deal the next quarter, then I've got a comp issue. But each time we sell hardware, we're selling services as well. We never sell naked hardware. It always comes with a services component. And that services revenue is recognized ratably over time. So the hardware revenue is recognized upfront. That $50 million gets recognized in Q1. The services revenue on that, let's say it's 10%. So a $5 million services deal will get recognized over a 12-month period. And so the more hardware we can sell, the more services annuity we're building up. It'll grow over time. It'll drive up the gross margins.
Right. And you just mentioned Meta earlier. And right, they were their largest customer in fiscal 2024 at 18% of the revenue. So that was also a huge question from investors all year. And you delivered a great result in line with your messaging. What can you tell us about the relationship going forward?
So it's very strong. They've been a partner, a customer since 2017. We don't do all of their data center work by any stretch. We do a very interesting amount for us. As you said, it's approximately 20% of our revenue. So that would make it $200 million a year, which is really interesting for us. They're spending billions of dollars on data center work. So a lot of that is handled in-house. But for the most complex things, again, they've chosen us as a partner and continued to grow their business with us. It's double-edged, right? They're a great customer to have. They're a fantastic reference customer to have. But we also have a lot of customer concentration. And so one of the things that we're really focused on is how do we diversify our customer mix.
Our number one sales metric this year, and I'm sure next year will be as well, is new customers. And large or small, we're interested in adding new customers. Obviously, prefer the large ones. But small customers can grow over time. So the first half of this year, our revenues, we have relatively few customers, mostly existing customers. But the second half of the year, the forecast includes a lot of new customer bookings, which is really exciting and portends very well to the future, FY 2026 as well. Because once we get them into our installed base, as you mentioned, we don't lose them. So we should be able to build off of the new customers that we get.
Yeah, just following up there, how has the go-to-market strategy evolved that enables you to kind of go after more customers? We've kind of been steady at 10 large customers, 20-30 customers total in advanced compute. We're really hoping to finally grow that. What have you and Mark done over the past year or two years to get that to happen?
Well, I think two things. Internally, we've brought in some new leaders. The gentleman who's running that business now came from Dell. The head of sales came from Dell. So great experience there. Associated with that, we've just signed a new partnership deal with Dell as well, which I'm pretty excited about. We did have an agreement with them where we could resell their hardware. They can now resell our services. So I think it's a natural evolution of business that has been occurring where they've, again, we're agnostic. We've sold their hardware. In some cases, they've had customers where they've sold their hardware and haven't been able to get the clusters to operate at the level of performance and output that the customer expected. And so they brought us in.
Dell brought us in as a sort of last-mile solutions provider to clean things up and to raise the output levels of the clusters. We're now just formalizing that. So they can sell our services on their paper, on their contracts. So from a customer standpoint, it's a Dell relationship that they have. They're not bringing in really necessarily a third party. But we can get access to a lot of customers that we, as Penguin Solutions, back to the branding question, we don't have to jump through as many hoops to get someone to trust us. We're leveraging Dell's relationships in that example.
That can increase our reach to market and also improve our time to revenue because we're not going to spend two to three months trying to convince somebody, "Hey, you're safe going with Penguin." Dell's doing that job. And Dell's got the great brand and the great history to do that. So I think that Dell partnership is potentially a big value driver for us. I don't think it'll probably materialize into much in FY 2025. But I think in FY 2026, I would expect that to generate some business for us. And we'll try to quantify that for investors maybe the back half of this year.
Right. Yeah. And just kind of last customer-esque question, just because it is, it's the biggest question that we get is how to think about the advanced compute growth rate. I think a lot of investors try and do this by customer. And so we get questions around that. Who are the current customers? Who are the potential ones? How big are they? How big are the contracts, the overall pipeline? So just how would you like us to think about underwriting the AC mid to longer-term growth rate?
Perhaps the number one question I get, and one I wish I had just a really crisp answer for you, but the challenge is each customer looks different. Some are $5 million. Some could be smaller, but let's say there's five. It might be $15 million-$25 million, and then onto Meta up into the $hundreds of millions, and so it's not just a matter of five customers at $50 million each. There's your $250 million and go from there. I think that we are working to provide more forward-looking visibility to investors to help answer that question. One of the reasons why we went to an annual outlook was to give more forward-looking, and that forward-looking outlook includes the sum product or the sum total of all information that we have access to. Obviously, we can't share all of that externally.
When you start sharing information externally, you want to make sure it's repeatable. You got to make sure it's clean. You deal with a lot of imperfect information as a management team to manage your business. And so, but hopefully, as we continue to mature, and it's a journey, we'll be able to give some more forward-looking metrics, which I think would be seen very favorably by investors. At this point, I would say we're focused on attracting new customers. And we're focused on growing the revenues with our existing customers who are very sticky. And I think if we just execute those two things well over the next year or two, you'll see that this business has continued to migrate away from being a memory module company and towards an AI compute company, which is going to unlock some value.
That metric for the new sales or the new customers, is that built into the sales organization as part of the bonus structure?
Yeah. Like I said, number one metric that they have, there's an accelerator on their quotas when they bring in new customers. They have to be of a certain size, obviously. You can't just sell them one thing.
A different yeah.
Yeah. But we're pretty excited about when Mark Adams, our CEO, talked about the pipeline being really healthy or healthier and it had grown year over year. He's talking about a couple of things there. One is the overall size of the pipeline, and that's coming primarily from new customers rather than existing, and then the second thing is that things are progressing, so if you know the Salesforce stages, right, you're moving from stage one, which is like prospecting, where you've sort of very high-level conversation, through stage three, stages four and five, and I would say we've seen the pipeline shape improve, so it's gotten bigger, but the shape's also improved, so things have migrated from stage one to stage two and into stage three, which is now you're actually to the configuration and the bid portion of the pipeline, so it's maturing.
It's growing at the same time.
Great. I'll stop there before going to SK Telecom for any audience questions.
This side of the room. Yeah.
OK.
Talk about moving to an operating company, dropping the operating or the holding company to an operating company. I mean, what does that mean in reality? You still got the memory business. You still got the business. How has that changed internally?
Yeah. Good question. Listen, I think on the one I could go very directly to one thing. LED is not really strategic for us anymore. It's a good LED business compared to other LED businesses that are out there. But there's really no argument for why that remains in the portfolio if our focus is on compute. I think memory you can make a pretty good case for, certainly a much stronger case than you can LED. But LED, I think the best possible outcome for us would be to find a strategic buyer and one that makes sense for us financially and remove that from the portfolio. So that's a big change and a transition that we might see internally. Not a lot of sexy stuff. But I mean, there's a lot of integration work that didn't occur at the time of some of these past transactions.
You sort of operated as individual companies, and so some of the work that I've been doing as part of transformation, say in finances, migrating everyone onto the same SAP instance, streamlining the reporting structures, looking for areas of synergies across things like sales operations or in manufacturing or places like that. Whereas before, they would have operated in vertical silos, you start to look at all these functions as horizontals, and you can take out costs and streamline things by doing that. That frees up investment capacity for us. Our OpEx is only growing low single digits, and revenues are growing mid-teens, so at the same time, we're making investments while we're doing that, so as we free up resources and spend on other areas by doing some of this integration work, which is hard work, it creates investment capacity for us.
We had another.
A bigger picture question on the Advanced Computing's growth opportunity: So, like, if Meta moves to like these 300,000 XPU GPU deployments this year, starting this year, the complexity is inherently more challenging. So, what's the growth potential? When you say you're more into complex projects, like, can you tell us the trajectory of the complex projects, how they're growing? Then a second question is, as you start to see these more advanced optical interconnects that allow you to disaggregate compute from memory, so memory starts to grow at a much faster rate, shouldn't that translate into an accelerated growth for you, what we've seen in the past?
Yeah. On the first part of your question, think about our target customer is actually not hyperscale. It's our largest customer. But it's really not our target customer. So hyperscalers, if you take the big four, they have very complex needs. But they also have a lot of sophisticated capabilities from an IT perspective. They roll their own, right? We happen to have landed a really good customer there. But that's not our target customer. They're good at the complex work that we do, by and large. It's their business model. Our target customer are enterprises that have complex needs but don't have the sophistication to go do that work themselves. So we think we're pretty early days in that part of the market, that life cycle where the enterprises are starting to do their own in-house AI work, some fine-tuning. We're seeing more and more pilots and small deployments.
But we think that those can grow into much larger deployments. We had one large consumer gaming company that we announced in not this quarter, but the prior quarter. It was kind of mid-single-digit millions of revenue this year. We've already got follow-on business with them lined up that if we win, we're in kind of late stages on that bid. It's double-digit million dollars, right? So a lot of enterprises are kind of starting, toes in the water. But as the value proposition for AI and their business models starts to evolve, we can see those really grow. So I think the complexity with something like Blackwell, with the cooling requirements that it has, these are huge investments that companies are making, really, really large IT investments, and they're fixed-cost investments.
So if you think about the financial model that that entails, with the fixed cost, you've got to get your output up to a certain level before you really break even, and so what I mean is, if you think about these clusters across a couple of dimensions, performance and availability, and the product of those is the output of what that cluster can generate, OK?, well, what we've seen is that industry standard is that these clusters are running at like 60%-70% across both of those dimensions. You actually multiply those two things together, so if you're running at 70% availability and performance, your output is only 49% of total potential, and if you think about a high fixed-cost base, you may be running below the fixed-cost investment at that sort of outcome.
And what we've managed to do at Penguin is, and this is documented at Meta, where we're achieving 95% of total potential, is that we're allowing companies to have more availability and higher performance, which then pushes them above that break-even line. So the reason why the services are of really high value is each incremental chunk of output that you're getting from your cluster drives you further and further above that break-even point. And so really high value-added services. So complexity is absolutely good for us, and especially complexity in the enterprise where they're less well-equipped to solve those problems themselves. And then you asked on the memory side. I think the growth there could potentially be really interesting for us. I think it could be in FY 2026.
We could see tens of millions of dollars of incremental revenue in our memory space, I think, by the time we get to FY 2027. Certainly, there are scenarios where we could even get to triple-digit incremental memory revenue. Early days. So it's not a commitment at this point because there's a lot of work that has to happen. But I think that the potential is of that nature.
OK.
Thanks, gentlemen.
Yeah.
You alluded to it throughout. But the hardware business itself, it seems like you guys are really a consultancy, a system integrator. And the hardware business is sort of commodity that you pass to. Is that the right way to think about it? I mean, why even have your own at all? I know there's a revenue issue if you don't stop selling it.
Yeah. I mean, we're really interested in gross profit dollars and operating profit dollars and cash flow. So yes, it would be a revenue hit if we completely stopped selling hardware. It's really a customer decision. So the customer will come to us in a lot of cases. And they want one throat to choke. And so you can think about us as like a general contractor, if you will, like in a construction space. We'll often do the bidding and the procurement for a customer. They want us to be the one negotiating with all the different vendors and handle all of that for them. And in that case, we're happy to do it as long as it comes with the services revenue as a follow-on. We don't take deals that are just hardware only. That's just not our business model. And it's not interesting to us.
Interestingly, with faster revenue growth, we'll probably see some gross margin compression because we've got more hardware coming in, and to your point, it's lower margin. It's kind of low teens, just like what it would be in the industry, but our services margins are three times that, four times that, and so with faster revenue growth comes some pressure on gross margin. It still flows through positively to operating profit, but as the revenue growth slows, we'll actually see gross margin rate improve because we'll have a higher services mix and lower hardware mix, and I still think really strong flow through to operating margin and operating profit in that case, so business model kind of works that way. The way when I put together the outlook, on the high end, we've got 20% revenue growth for the year and 31% gross margin.
Those numbers could go hand in hand with each other. At the high end, we've got 10% revenue. Or excuse me, at the low end, we've got 10% revenue growth. And at the high end of the gross margins, 33% because you'd have a higher services mix. Those things could go together. So trying to balance those things. Net net, I'm looking at gross profit dollars because that's what's going to fund the investment in the company and generate the cash.
All right. I think we have time for one, maybe two. For the SK Telecom transaction, I guess I'm most interested is this more of a partnership where you're utilizing their connections for lead generation? Or is this more of a direct customer opportunity?
I would say both are very much in play. And there's a third dimension as well. So we see opportunities to sell directly to them. They've talked about, I think, at least four large data centers that they're working on at this point. And too early to say whether or not what business we'll get from that. But certainly, we're in deep discussion with them on how we can help them. Also, they have real aspirations to be an AI service or solutions provider. They're building out really the entire stack from an AI infrastructure standpoint, from GPU all the way up through the application layer. And we're kind of the infrastructure component of that value chain for them, the infrastructure integration component of that for them. So they can use us kind of like Dell as a part of their go-to-market.
We can step in and be the integrator of choice. Then the third thing is just around core technology work that we can do with them. That's true in the memory space. But it's also true with some of the software that is used to manage these AI deployments. Like I said, we have our own software today. They have software as well that's really complementary to what we do. Ours is more focused on providing solutions for on-premise infrastructure. Theirs has some nice cloud features to it. You can combine those two things together and provide a more complete solution for a hybrid environment.
Yeah. Yeah, that was my next question. What's the difference there? So thanks for that. Yeah, that's a super interesting partnership. They're talking about this AI infrastructure superhighway, building multiple 100-megawatt data centers going to a gigawatt. So yeah, look forward to seeing how that impacts the P&L.
Yeah. I don't know if you saw them at CES.
Incredible.
Big presence, all focused on AI. They've been moving some of their top executives into new roles centered around AI. So I think we've got a good opportunity with them.
Cool. We're out of time. Thank you so much, everyone. Thanks, Nate.
Yep. Thanks, Nick.