Penguin Solutions, Inc. (PENG) Earnings Call Transcript & Summary

April 9, 2024

NASDAQ US Information Technology Semiconductors and Semiconductor Equipment earnings 52 min

Earnings Call Speaker Segments

Suzanne Schmidt

executive
#1

Thanks, operator. Good afternoon, and I appreciate everyone taking part in today's earnings conference call and webcast, where we'll go over SGH's results for the second quarter of fiscal 2024. Joining us today are Mark Adams, our Chief Executive Officer, Jack Pacheco, our Chief Operating Officer, and Ken Rizvi, our Chief Financial Officer. The slide presentation and press release that go along with this call can be accessed in the Investor Relations section of our website. We'd encourage you to visit that site during the quarter for the latest information about the company. I'd also like to remind everyone to review the note regarding the use of forward-looking statements contained in the press release and the earnings call presentation. I want to remind everyone that during today's call, we'll be sharing forecasts and forward-looking remarks. Those may cover things like how the company expects to grow and what our financial picture looks like going forward. Such statements are based on what we currently think and assume — they aren't promises about how things will actually turn out. A number of risks and uncertainties could affect them, including those described in the press release and the earnings call presentation we filed today, along with those in our latest annual and quarterly reports. These forward-looking statements reflect circumstances only at the moment they're issued. Unless the law demands otherwise, we take on no obligation to revise or publicly refresh any of them. Today's discussion will also cover financial measures under both GAAP and non-GAAP. You shouldn't view non-GAAP figures on their own, treat them as a replacement for GAAP results, or rank them above those results. When evaluating how we performed, we urge you to weigh every measure. Today's press release and the slide deck that goes with it contain a reconciliation between the GAAP and non-GAAP figures. Having said that, I'll hand things over to Mark Adams. Mark?

Mark Adams

executive
#2

Thanks, Suzanne, and I appreciate everyone being here for our earnings call covering the second quarter of fiscal 2024. We turned in strong financial performance during Q2 and kept moving forward significantly with our shift into being a supplier of high-performance, high-availability solutions that enterprise clients require for deploying AI on-premise, at the edge, and in the cloud. With decades of expertise in high-performance compute and specialty memory, SGH stands among just a handful of industry participants and is uniquely suited to assist companies in handling the complexity that comes with implementing AI at scale. As a provider of total solutions, we give our customers and partners inventive technology, hardware configurations that are agnostic, software that oversees AI systems to deliver maximum output and availability, plus a professional services suite allowing our customers to reach best-in-class performance and reliability. Allow me to recap our operational results for the quarter. Revenue came to $285 million, matching the midpoint of our guidance range. Even though non-GAAP gross margin landed at the lower end of our guidance because of a larger hardware mix, we posted non-GAAP earnings per share of $0.27, surpassing the midpoint of our guidance thanks to improved operating expense controls. We finished Q2 with a robust balance sheet. Cash and short-term investments reached $466 million; now, I'll begin our business line review with the Intelligent Platform Solutions Group, or IPS. Our IPS team provides a strong solution set encompassing industry-leading hardware, sophisticated cluster management software, and best-in-class professional services. This solution portfolio supports our design, build, deploy, and manage solutions framework for HPC and AI applications on-premise, at the edge, and in the cloud. During Q2, IPS revenue reached $141 million, climbing 19% versus our prior quarter, which accounted for 50% of total SGH revenue, making IPS the biggest piece of our overall business in Q2. Our vision is unmistakable: partner with our customers and work alongside them to build the future of AI. The market keeps showing robust investment in deploying AI infrastructure solutions from hyperscalers and large-scale cloud service providers, or CSPs. The opening months of 2024, though, have made clear that AI is no longer only for early adopters; we're observing indications of AI adoption among larger enterprises in sectors including financial, oil and gas, defense, education, and digital media, along with tier two CSPs, with projects spanning proof of concepts to large-scale deployments. Our engagements with existing and prospective AI customers have risen noticeably in volume across the last few months, mirroring this market dynamic. In our discussions with both current and targeted new clients, they've described the challenges they encounter when deploying AI, attempting to handle the complexity that emerges as they integrate advanced compute, memory, networking, storage, and cooling in large-scale data center rollouts... What our customers ultimately need is high-performance compute running workloads at scale within an environment that delivers maximum uptime and overall efficiency. Serving as a provider of complete solutions, we stand prepared to take on that challenge. What we give our customers is an end-to-end offering: inventive hardware engineering, software that oversees AI infrastructure so it delivers peak output and uptime, plus a collection of professional services — every piece intended to help them reach top-tier performance and dependability. Because we put the customer first, their concerns sit at the center of all our efforts, which means every solution we hand over is customized for their particular needs and positioned to back their success. In our view, AI inferencing at the edge is going to be a major market opportunity too, since it moves intelligence nearer to the place where information carries the most value and nearer to the spot where choices get made. At the edge, we've broadened what we can do with our freshly launched next-generation fault-tolerant computing platform, the Stratus ztC Endurance Server. Demand for this platform is strong — it lets our customers run applications with an unplanned target downtime estimated in minutes per year — and we're building out an approach that will let them deploy AI at the edge using a platform that's both high-performance and high-availability. The ztC Endurance Server stands as yet another illustration of the investments we've made and keep making to serve our customers' needs, whether that's on-premise, at the edge, or in the cloud. We're also unveiling a new addition to our management team today. It's my pleasure to share that Pete Manca is joining us in the role of President of IPS. Pete carries deep experience in growing businesses that deliver high-availability, high-performance solutions to enterprise customers. Before he came aboard, Pete spent five years as Senior Vice President and General Manager at Dell Technologies, where he oversaw several large businesses, among them Converged Solutions, OEM Solutions, and APEX, Dell's end-to-end cloud portfolio spanning storage, high-performance computing, and AI services and solutions. Before Dell, Pete held the roles of President and CEO at Egenera, a top provider of wholesale cloud computing solutions, which highlights how broad his experience and expertise are. I have every confidence that Pete is the right person to drive the team forward and capitalize on the opportunities lying ahead. Pete will collaborate with the previous IPS President, Dave Lorello, who is moving into an advisory position. Dave has been an extraordinarily valuable partner in reshaping IPS; under his guidance, IPS grew more effective and efficient in every area, from go-to-market, to engineering, to manufacturing. He's a leader of great integrity with an execution mindset, and we will miss him. We send Dave our very best wishes. Now shifting to memory, which runs under the SMART Modular brand name. For specialty markets including supercomputing, networking and telecom, storage, data centers, industrial, and other specialty applications, we supply customers with memory solutions that are high-performance and high-reliability. Revenue for Q2 landed at $83 million, which equals 29% of total SGH sales. As we anticipated, sales were slightly lower than Q1 levels, mainly because inventory stayed elevated at a number of our large customers. We keep observing indicators that the memory cycle has begun moving higher. That said, during our prior earnings call we noted that unit demand over the short term is still difficult at certain longstanding enterprise clients of ours. Even so, we stay confident that the business will recover as we enter the back portion of our fiscal 2024, and we anticipate sequential revenue growth in the third quarter. AI is likewise transforming the memory market's structure, since requirements for denser capacity and wider bandwidth are growing ever more essential to the system performance needed for the most sophisticated compute workloads. By taking advantage of Compute Express Link, or CXL, memory expansion alongside switching technology — which enables various portions of a computer memory system to communicate with greater speed and efficiency — we are broadening our product lineup to seize on where compute and memory converge along with system-level solutions; our CXL product development efforts keep advancing. The design of our 8-DIMM DDR5 CXL add-in card has been finished successfully, and we expect to sample this novel offering to our customers in the latter part of this year. Delivering 512 GB of memory, this high-density approach makes the system quicker and better equipped to manage the intricate tasks that large-scale AI and high-performance computing workloads demand.

Mark Adams

executive
#3

In Q2, we delivered first engineering samples of our 4-DIMM DDR5 CXL add-in card to several enterprise clients. A patented, unique technology in this product keeps its footprint inside one single-width PCIe slot. That layout is extremely advantageous for 1U servers, since it frees space for other PCIe devices, such as accelerators and network interface cards. Revenue from this product should start ramping in early fiscal 2025, we anticipate. And we rolled out our Zefr Z-DIMM memory models featuring ultra-high reliability for demanding compute applications, while demand keeps building for our DDR5 SO-DIMM products, driven by our exclusive I-Temp and Zefr testing offerings, which boost reliability considerably... All of this together demonstrates how we keep committing to innovation that leads the industry, enabling us to serve what our customers require. Turning now to Cree LED, maker of application-optimized LEDs for products and markets including specialty lighting, video screens, outdoor, horticulture, and architectural lighting; LED Solutions revenue came to $60 million in the second quarter of our fiscal 2024, representing 21% of total SGH sales. Second quarter sales declined on a sequential basis, as expected, largely because of seasonality. Since the demand picture for LEDs stays fairly subdued over the near term, we keep running our Cree LED operations carefully and are matching our expenses to today's business conditions. We feel good that demand trends will start to strengthen and right now anticipate that revenues will rise sequentially in the third quarter. Driving the technology development Cree needs to stay ahead in high-value specialty applications has kept our R&D team hard at work. In the quarter just past, we brought out the XLamp XP-G4 high-intensity LEDs, built for indoor directional, aftermarket auto, and portable applications; we likewise launched an addition to our XLamp S product line, aimed at horticulture with products made for settings such as greenhouses, where precise lighting can reshape crop growth and yield. As a technology and brand leader holding a robust intellectual property portfolio, the Cree LED team keeps spearheading lighting innovation. With our strong R&D and IP portfolio, plus our capital-light outsourcing model, I have confidence in Cree LED's ability to compete and its prospects for success ahead. I'll pause there and pass things to Ken for a closer look at our Q2 financial performance and our guidance for next quarter. Ken?

Ken Rizvi

executive
#4

Thanks, Mark. My comments will center on the non-GAAP figures, and you can find reconciliations to GAAP in the tables accompanying our earnings release as well as in the investor materials posted on our website. Let me now move on to how we performed in the second quarter. SGH's total revenues came to $285 million, right at the midpoint of the guidance we provided, while non-GAAP gross margin landed at 31.5%, the low end of our guidance range, largely because hardware made up a bigger portion of the mix. For Q2, non-GAAP diluted earnings per share reached $0.27, surpassing the midpoint of our guidance and benefiting from tighter control over operating expenses. During the second quarter, services revenue across the company came to $49 million in total, which is lower than the $55 million recorded in the same period a year ago. Product revenues amounted to $235 million. Here is how second quarter revenue broke down by business unit: IPS brought in $141 million, memory contributed $83 million, and LED added $60 million. That works out to a sales mix consisting of 50% IPS, 29% memory, and 21% LED. For SGH in Q2, non-GAAP gross margin was 31.5%, which is below the 32.1% posted in the year ago quarter, mainly as a result of reduced memory volumes, though an improved mix within IPS provided a partial offset. Looking at the quarter-over-quarter picture, gross margin came in lower than the 33.3% we posted last quarter. The main driver there was a bigger contribution from hardware within the overall revenue mix. When we turn to non-GAAP operating expenses, Q2 landed at $63.2 million. That's below the $64.6 million figure from Q1, and the improvement largely reflects lower variable spending as well as the cost-cutting steps we took. Compared with the same period a year earlier, when the number stood at $68.7 million, operating expenses were also down. On the bottom line, non-GAAP diluted EPS for Q2 2024 was $0.27. A quarter earlier it was $0.24, and in the year-ago quarter it was $0.87. In Q2 2024, adjusted EBITDA was $33 million, which worked out to 12% of sales. That compares with $34 million, or 13% of sales, during the prior quarter, and $65 million, or 17% of sales, in the same period a year earlier. Turning now to balance sheet items: on the working capital front, net accounts receivable totaled $170 million, slightly under the $171 million recorded in the previous quarter. DSO came in at 41 days, down from 44 days in the prior quarter, largely driven by the timing of invoicing and collections. At the close of Q2, our inventory was $173 million. That's down from $208 million at the end of Q1. Turns for Q2 came in at 6.8 times, up from 5.8 times the previous quarter. Consistent with our past practice, gross sales and cost of goods sold form the basis for calculating net accounts receivable, DSO, and inventory turnover. For Q2, those two figures were $375 million and $294 million, respectively. Just a reminder that the gap between gross and net revenue stems from our logistics services, which are accounted for on an agent basis, so the only thing we recognize as revenue from logistics services is the net profit. At the close of the second quarter, our cash, cash equivalents, and short-term investments came to $466 million, which is $88 million lower than the $553 million recorded in the prior quarter. During the second quarter, we made a $50 million payment for the Stratus earn-out and paid off roughly $37 million under our term loan facility. Cash flows used in operating activities for the second quarter came to $22 million, whereas the prior quarter generated $60 million from operating activities. When you look at the $22 million in operating cash flows for the second quarter, that number includes the $29 million we paid out as contingent consideration — that was a piece of the $50 million total cash payment tied to the Stratus earn-out. During Q2, we also repurchased roughly 106,000 shares of common stock, spending $1.9 million on that. Since we first got repurchase authorization back in April 2022 and running through the close of our second quarter, the total spent on buying back 4.1 million shares comes to $72.3 million. As of the end of Q2, we still have $78 million left that we can use for repurchases under our existing authorizations. I want to remind everyone that when it comes to capital allocation, our strategy remains the same. To start, we're going to keep putting money into the business because we believe there are meaningful chances for more organic growth. Next, we'll stay disciplined as we look at possible acquisitions, like the one we just completed with Stratus. Then there's the share repurchase authorizations, which give us room to send capital back to shareholders in a way that's opportunistic and sensitive to price. Last, we want to pay down debt so our gross leverage remains at sensible levels. During the second quarter, we paid off $37 million of our term loan, and once the quarter closed, we paid down another $75 million of term loan. Since the first quarter, those payments have cut the principal by $112 million, leaving $425 million outstanding. If you're keeping an eye on capex and depreciation, here's where things landed: capital expenditures came to $5.2 million during the second quarter, while depreciation totaled $7.2 million. Let's shift now to what we're projecting for the third fiscal quarter of 2024. At the midpoint, we're looking at roughly $300 million in revenue for Q3 2024, give or take $25 million. Several things shape that third-quarter outlook: within IPS, the midpoint suggests revenues holding steady or rising modestly, and there are further opportunities that could land in either the third or fourth quarter, subject to when deployments actually happen. Regarding memory, at the midpoint we anticipate sequential revenue growth in the high single-digit range; similarly for LED, our current projection is a sequential revenue increase in the high single-digit range at the midpoint. For the third quarter, we expect GAAP gross margin of roughly 29% at the midpoint, plus or minus 1.5%. Non-GAAP gross margin for Q3 should come in at about 32% at the midpoint, plus or minus 1.5%. Non-GAAP operating expenses for the third quarter are projected at approximately $66 million, plus or minus $2 million, which represents a slight increase versus the previous quarter, driven mainly by variable costs tied to the anticipated higher revenue. Looking ahead to Q3, we expect GAAP diluted EPS to be a loss of about $0.07, with a margin of plus or minus $0.15. If we exclude share-based compensation, amortization of intangible assets, debt discount, and other items — which gives us our non-GAAP measure — diluted EPS should be roughly $0.30, again plus or minus $0.15. For GAAP, our projected diluted share count in Q3 is about 52.6 million, based on where the stock is trading now, while on a non-GAAP basis it should be approximately 54.4 million. Cash capital expenditures for Q3 are expected to fall in the range of $4 million to $6 million. A reminder here: the long-term projected non-GAAP tax rate we're applying is 28%, and that captures information available right now, among it the divestiture of SMART Brazil, finished during the first quarter, plus additional factors and assumptions. Through 2024, we do expect to apply this normalized non-GAAP tax rate, though the long-term non-GAAP tax rate could shift for numerous reasons — among them the fast-moving global tax landscape, major shifts in the geographic mix of our earnings, or alterations to our strategy or business operations. Our Q3 2024 forecast rests on the present environment, one that accounts for today's global macroeconomic headwinds and continuing supply chain constraints, particularly those touching our IPS business. That encompasses longer lead times for specific components built into our overall solutions, which affects the speed at which we can scale both existing and new customer projects. We keep running our operations prudently as we steer through a difficult environment while continuing to invest in our long-term growth. For additional specifics, please look at the non-GAAP financial information section and the reconciliation of GAAP to non-GAAP measure table included in our earnings release; with that, I'll hand things to Mark for a few comments ahead of Q&A.

Suzanne Schmidt

executive
#5

Thanks, Ken.

Mark Adams

executive
#6

Since fully operational AI infrastructure at scale is still in its early innings for the majority of enterprise customers, the market's demand keeps accelerating for a trusted advisor who can help address the difficulties of getting this new AI infrastructure deployed. With more than 25 years of expertise in HPC and memory, plus our deployment know-how, we are positioned to lead in this market by working to solve the toughest AI infrastructure needs our customers have. As our transformation continues, I want to thank our global team members for what they did this quarter. We feel we are well-positioned for the exciting market opportunities ahead. Operator, we are now ready for Q&A.

Quinn Bolton

analyst
#7

Hey, everyone. Nice work on the strong results and guidance. I'd like to begin with the IPS business, and Mark, perhaps you could share your perspective on demand trends across the customer base — large CSPs, tier two CSPs, and enterprises — specifically where you're seeing the strongest demand for new deployments as we head into the back half of calendar 2024. I have a few follow-ups after that.

Mark Adams

executive
#8

Oh, absolutely. Thanks for the question, Quinn. So when we consider how we view the market, it's rooted in our executive-level discussions with those three segments you mentioned. Clearly, the first movers in this space were the big hyperscalers, and one of them is obviously a significant customer of ours. Then there are the other two segments you brought up — large enterprises that have an infrastructure need in place for AI to support their future success; and then those tier two players, which we refer to as tier two CSPs, meaning companies that possess their own infrastructure to deliver services enabling AI implementations across a variety of different verticals, if you will. So those are the three segments through which we assess the market, and as I said, the initial investors were hyperscalers. That said, what we've witnessed a significant increase in comes in two forms. Large enterprises are assessing their AI approach, and within that approach the real question is: do they run it on-prem in their own environment? Do they pursue a colocation model, leveraging an outsourced data center provider for space and power? And do they bring in a company like Penguin to assist with deployment? Or do they instead team up with a partner through some kind of JV structure or co-investment structure for a future build? We're observing all of those models, and the demand signals coming from those kinds of customers on the enterprise side are extremely strong; plus, the hyperscalers are mostly building to suit their own needs. The tier two CSPs are primarily putting money into infrastructure for enterprises lacking their own infrastructure, largely involving data centers; and one thing I want to emphasize — if you speak with anyone in the power and data center space, there's an enormous shortage right now of available data centers for AI, and that enormous shortage centers on the power requirements needed to run AI. So at present, there's substantial pent-up demand for new data center capacity to power AI with gigawatt, megawatt-type performance, which lets you install these kinds of large-scale cluster AI platforms. That gap is driving a great deal of hustle around which providers can get these systems enabled for customers fastest, or within the tightest window. And that is exactly where many of the engagements we have come from — customers telling us, "Hey, we need this capacity, and we have to be operational by X." Those customers, in the enterprise world, generally lack the kind of infrastructure that one of the big four hyperscalers has when it comes to deployment, and they truly don't even know what they know regarding performance, uptime, reliability, design uptime, and optimization. Every one of those things matters. So the volume of engagements we've had, the designs and the proposals, producing greater clarity around bookings and even future deliveries — it's all moving the right way.

Quinn Bolton

analyst
#9

Great. My follow-up is this: you brought up AI at the edge and appeared to put some emphasis on your Stratus product. I'd like to know whether there are also chances for edge AI within the legacy Penguin side of things, where someone might be standing up bigger clusters to serve LLMs or other AI workloads. In that case it wouldn't only be about high reliability; you might simply be saying, "We need inferencing placed as near to the end user as we can get it," and you could imagine fairly big rollouts of inferencing hardware instead of only training, which seems to have been the main emphasis of some of your older deployments.

Mark Adams

executive
#10

... I'd agree with that; your question basically contains its own response, since we're currently building solutions and putting money into development that will later be folded into our solution roadmap to deliver precisely that. Like you said, looking back at earlier AI spending, training outweighed inferencing. And training will not disappear, because with large language models, there will be plenty of them that organizations draw on to pull out their own tailored version of whatever they require from such a model. So that training component will remain. As a share of the total, though, inferencing will clearly grow larger going forward. And because of that, as you pointed out, when greater speed and access to information exist nearer to where decisions get made, that can determine whether a company stays competitive or not. So the solutions on our minds are: how can we, beyond merely using the platform as you described, create and deliver to our customers the best-performing, most dependable systems in settings where IT staff may be absent? That's central to our strategy. Yet a day could come when we're building small cluster platforms or single cluster platforms using these platforms, judged on cost, reliability, and performance, enabling them to sit in those end markets — oil and gas, perhaps on a rig, perhaps a retail store, perhaps an ATM, perhaps a healthcare regional office, and so forth; it's just... This is how we're approaching AI. Without question, edge AI has a strong future ahead. It's simply so early. I mean, AI overall is early, so we're putting money in for tomorrow as we consider that solution platform we're building.

Quinn Bolton

analyst
#11

Great. Then one final quick item regarding the memory forecast. You had projected high single digit expansion. We've been hearing from numerous people within the memory sector that costs for DRAM as well as NAND are climbing quite rapidly at this point, particularly following the Taiwan earthquake. Curious, when you consider that high single digit figure, is it primarily average selling price increases while volumes remain fairly subdued, or are you beginning to observe a volume rebound, though pricing gains might arrive in later quarters? How exactly are you weighing volumes against ASPs? And I'll return to the queue. Thanks.

Mark Adams

executive
#12

Yeah, prior to providing the particular response, I simply want to point out, if you look back at memory cycles historically, in contrast to the pure play semiconductor firms, we've generated profits throughout those cycles, correct? That happens because our business is differentiated; plus we observe that if you chart this historically, we tend to be somewhat delayed on the initial side of experiencing the downturn's discomfort, while on the latter side, our rebound comes a bit more slowly owing to those inventory adjustments of certain inventory that customers might be holding or, or, or refraining from regarding the, how they consider purchasing memory. Thus, we are beginning to notice pricing advantages across our business generally, and we're observing isolated instances of demand inquiries. I'd say our demand has been relatively level recently, yet that beats shrinking during the previous couple of quarters. And if you revisit the prior earnings call, I speculated, I forecasted that my perspective was we might likely begin seeing additional demand during our Q3, Q4 period, since I merely believed it would be somewhat slower. And, you know, we're beginning to witness that reversal. It's simply, it isn't present in the, in the Q2 figures. I believe during Q3, we're starting to observe somewhat more demand. And I think we'll keep seeing that across Q4 and early fiscal year 25.

Ken Rizvi

executive
#13

Yes, and Quinn, to really drive home what Mark was saying about margins—take a look at the operating margins for this business. Even in Q2, a quarter that came in on the softer side, op margins landed slightly above 7%, and we're beginning to see some growth expectations take shape. That's exactly what we guided toward for Q3, on a sequential basis for the memory business too.

Quinn Bolton

analyst
#14

Great, thanks.

Kevin Cassidy

analyst
#15

Yeah, thanks for fielding my question, and nice work this quarter. Could you expand a bit on IPS — you mentioned it would be flat to up, hinging on at least a single order. Does that figure take in services, and is it tied to the gross margin guidance of 1.5%, either direction? Could you just-

Mark Adams

executive
#16

Yeah

Kevin Cassidy

analyst
#17

... could you expand on that a bit more?

Mark Adams

executive
#18

Thanks for asking. I'll kick things off, and then Ken can add his thoughts. Like we told you at the start, back on our Q1 earnings call, we anticipated a stronger second half, and that is exactly what's happening now. And with some of these engagements, the way margin shifts is relatively muted. Mix issues could be the cause, or it could be hardware early configurations that we later convert services to. When you take all of that into account, we're sitting in a good spot, and margins are going to fluctuate. As you mentioned, I'll have Ken put some numbers to that, but margins do move around somewhat in our business. Let's keep in mind, gross margins have climbed from 19.5% three years ago to where we stand today. When a margin swings by a point in a quarter, it's mostly due to mix — perhaps hardware deployments in one quarter, and then later quarters where software or services might be stronger, and that's simply how this business works. We're very happy with how the team has executed and very confident that good progress will keep coming.

Ken Rizvi

executive
#19

Okay, Kevin — so when it comes to margin, what we did was leave ourselves a bit more headroom in Q4. Some of that comes down to timing — not only on the hardware rollouts, which we've discussed before, but also on making sure the customer signs off the way they should, and things like that, and those can shift between quarters, as you know. That's the reason we laid it out that way. But when you look at the Q4 margin and the guidance, we've got point in time services plus ongoing managed services, and we simply wanted a little extra flexibility. My guess is most of the Street will probably land close to the midpoint, though this gives us some wiggle room around that.

Kevin Cassidy

analyst
#20

All right, excellent. Appreciate the additional detail. And like you mentioned, as your pipeline visibility or interactions grow, could you elaborate a bit on, you know, the opportunity to widen the client roster specifically for IPS?

Mark Adams

executive
#21

Yes, certainly. Kevin, I believe, as you're aware, this is something we've discussed many times — it's a segment we've described as having concentration among a few customers and being uneven in nature. We've put a great deal into this during the past fiscal year, across the most recent 9 to 12 months, toward adding fresh go-to-market personnel. And frankly, I'm extremely happy with how things are advancing on that front, particularly around handling key accounts and securing executive-level involvement to bring our offerings to these big enterprises and tier two CSPs. The volume of proposals we've produced keeps growing alongside the opportunities we're identifying, and we're eager to reach a point where those turn into bookings and revenue eventually. Consider it — we get them signed, but then we must go out and align that with supply chain verification regarding what's achievable in our configurations, after which we discuss rollout timelines and similar matters. So I want to highlight one thing, since I feel I haven't done this well enough on earlier calls. We don't view our customers through a buy-sell lens. That's not how we operate. And I realize it's tricky because, you know, we previously were solely a memory module business, and back then it was very much, very much an industry that operated transactionally. But the transformation we keep discussing publicly isn't merely about shifting from memory into AI infrastructure and HPC. That's not the endpoint. In terms of how we approach our customer relationships, it's truly more about clients and engagements, since we stay with them for quite a while. We aren't just making a sale and vanishing. We design it, we install it, we manage it, and then we ensure it performs over time; so that relationship, you know, I think once more, is that we have clients and engagements. We don't have transactional customers. Yes, we have customers, clearly, but those get handled as a client and an engagement, and that's where it appears on the gross margin line when I stack myself against some of the bigger competitors we, we go up against, if you will; and so, I'm quite happy with how that process is developing and unfolding. I'd also like, Kevin, if you don't mind, to just seize the moment. While we're doing that, what's become really clear is that the demand for what we provide is getting more and more confirmed daily. So what I'm getting at is this: sure, a ton of GPUs have been sold during roughly the past 12 months, yet once customers actually put them to work, they hardly overlook how complicated it all gets. Where does that complication show up? Take data center design — how do I bring power into the building, into the transformer, into the main systems? How do I lay out every piece of that data center, from a cold aisle to a hot aisle? What does my rack design look like? In what way are the GPUs managed? How do I get the most uptime and availability? Can I detect future failures ahead of time, sparing myself downtime? And beyond that, hey, how do I scale later on? So the sale we're making now is far more of a consultant advisory one, which means those executive engagements I mentioned keep confirming that this kind of trusted advisor relationship is needed, and I believe our position there is very strong after... You know, nothing could set the stage for AI infrastructure better than HPC. Think about HPC — its real role was helping people build big multiprocessor architectures within data centers, though not for AI back then, yet it developed in that direction as we moved closer to 2018, 2019, 2020. So very few companies possess 25+ years of deployment knowledge across hardware, software, and the future add-on services. And I think we're articulating that value to these customers better than before.

Kevin Cassidy

analyst
#22

Okay, great. Thank you.

Mark Adams

executive
#23

Thank you.

Analyst

analyst
#24

Hello, this is Scott, filling in for Tom O'Malley. I'd like to ask about the services line. It appears that services experienced a fairly significant decline during the quarter. Should we view that as roughly the new baseline level, or do you anticipate it will rise again? Thank you.

Ken Rizvi

executive
#25

... so when I think about Q4, my expectation is that services will increase. As we've pointed out previously, among the various services we offer, some are point-in-time services — design, implementation, and similar work — while others are managed services. So, during Q3 we had fewer point-in-time services. My expectation is that figure will rise again in Q2, sorry, I mean my expectation is that figure will rise again in Q3, and our guide accounted for that.

Tom O'Malley

analyst
#26

Thanks, that's useful. Then one additional thing, if you wouldn't mind revisiting CXL. Can you share your view on when that market will start turning upward in a bigger way, plus which kinds of buyers appear interested — AI-related, or broader general purpose?

Mark Adams

executive
#27

Let me handle the first portion, then Jack can address productization and revenue. When you speak with top technology and engineering leaders about AI performance curves and AI infrastructure, the number one item by a wide margin involves latency and performance problems stemming from insufficient bandwidth and memory availability for the GPU and CPU. A huge amount of capital is flowing toward early stage businesses, and several bigger firms also are putting money into solutions addressing this bandwidth constraint. Here's an analogy I enjoy using: sharing storage across a network is possible right now. Doing the same with memory is not. So down the road those capabilities will arrive — I believe it falls under CXL 3.0, at some future point — and memory spooling will become possible, yet that alone won't suffice. Memory's transport layer represents a separate challenge, which folks are attempting to solve using an optical transport layer with CXL layered above it, purely to maximize speed. High bandwidth memory isn't the sole answer, and additional approaches will emerge regarding how to boost throughput, latency, and overall compute system performance through more inventive hardware options such as CXL and optical transport folded into such an approach. Now Jack can discuss the market dynamics regarding when we expect that to begin contributing meaningfully.

Jack Pacheco

executive
#28

Hello there. Yes, just like Mark said — he brought up the close of this calendar year, and we anticipate meaningful CXL shipments around then. So on both fronts: we'll be delivering into your typical server manufacturers, and beyond that, we've been holding discussions and providing samples to plenty of different players in AI who are looking to add more memory to their AI servers. What I mean is, with HBM expanding, you have to keep adding more conventional memory into the server, and as you do that, space starts becoming scarce because DIMM slots are shrinking while the memory has to run faster. That leaves really only one route to obtain that extra memory — turning to CXL and attempting to fit additional memory into the server. We're noticing chances across both phases of the market.

Tom O'Malley

analyst
#29

That's very helpful. Thank you.

Mark Adams

executive
#30

Thank you.

Denis Pyatchanin

analyst
#31

Hello, Denis here filling in for Brian. I appreciate you taking a few of our questions. When it comes to IPS, in what ways has demand for generative AI during roughly the past 12 months altered how you bring customers on board and, down the line, the way products get delivered and rolled out? Has that gotten longer or more complicated? Also, you brought up choices such as on-prem, co-location, or co-investment. What effect does a customer picking one among those three have on either how long a rollout takes or on your business broadly?

Mark Adams

executive
#32

Yeah, excellent question — allow me to pull back a bit. My belief hasn't changed that we're at the very beginning of what AI will become, and all the chatter regarding GPU purchases and similar things, sure, that serves as a trigger or a gauge pointing to what lies ahead, yet what I'd share is that big enterprises, in my view, are still genuinely figuring out their approach to AI. My sense is that a great many of those GPU purchases went to parties in the middle who were constructing the infrastructure, and I don't believe large-scale enterprises have truly nailed down in full how AI will reshape their operations and hand them a strategic edge. The reason I bring that up is that all of this is unfolding live right before us. On top of that, this represents an entirely new architecture for any enterprise apart from the major hyperscalers, who have spent several years in this space. Still, what I mean is, there exist businesses that are household names within the sectors they operate in, and they're truly wrestling with something as basic as how to lay out a data center, or how to build the correct solution, and the reasoning behind choosing A instead of B. So when I consider what your question is driving at — what effect is this having? — it has, in my view, actually turned adoption into more of a longer-term journey, yet I also believe, as I said earlier during this call, that what I've confirmed via my executive-level interactions with our sales team is that big enterprises are beginning to grasp it. They're beginning to recognize that a divide exists between what they've accomplished on the information technology side and the abilities they'll require to roll out AI effectively, and that divide truly represents Penguin's opportunity. From where we sit, sure, it could take somewhat longer, there could be a steeper learning curve, yet that's simultaneously causing us to become more entrenched and genuinely able to pinpoint the value we can deliver to our partners, our clients, who require our assistance with AI deployment. And that, I believe, is where you're picking up on the confidence and the optimism regarding the back half of our fiscal year.

Denis Pyatchanin

analyst
#33

Thanks. As a second question, word has reached us lately that every major memory manufacturer is raising prices on both DRAM and NAND. Should those costs keep climbing, what effect do you expect that would have on your gross margins down the road?

Mark Adams

executive
#34

Our practice is not to project beyond the usual one-quarter horizon, and Ken already shared that outlook with you. That said, a few points: any upside of the kind you're describing would typically appear in our revenue at the top line. Yet, as we've stated on earlier calls, our gross margin will drift either upward or downward based on the direction pricing takes. What's the reason for that? It's because a shift in pricing doesn't alter what we bring to the table, and our clients understand this. So strangely enough, when our top line falls, profitability remains fairly steady—it might even edge up slightly—simply due to the arithmetic of a constant value spread across shrinking revenue. The opposite also holds; we could potentially observe a hit of one point, two points, or something like that once things start recovering. I don't expect that to occur all at once. From my perspective, that won't really shift how the business operates, since in a market where pricing is recovering like this, the added revenue and the gross margin dollars usually outweigh whatever effect shows up in gross margin percentage.

Ken Rizvi

executive
#35

And what I would say in addition to that is that, as we observe, you know, once revenues begin increasing, driven by both units and ASPs over time, what we'll see reflected in the bottom line is an operating income percent that gets better. And it's visible that within just the last 2 years, for the memory-specific business, this operation was delivering op incomes above 10%, and right now we're under that level. So once we begin to see unit demand get better, along with ASPs, that ought to flow through to the bottom line.

Denis Pyatchanin

analyst
#36

Great. Thank you.

Mark Adams

executive
#37

Thank you.

Mark Adams

executive
#38

Okay, that's great. Thanks, everyone, for being here. Just before we wrap up this call, I wanted to mention that we're planning an IR Day sometime in the next few months, and you'll be getting the specifics before long. Until then, I hope to see you at shows, NDRs, and similar events, and once more, thank you for taking part today. Have a wonderful week.

Operator

operator
#39

That brings us to the end of this session. We appreciate everyone joining us, and you're free to hang up at this point.

Read the full transcript via the API

You're viewing the first half of this call. Get the complete Penguin Solutions, Inc. transcript — plus 132,000+ transcripts from 7,000+ companies, speaker segments, AI summaries and full-text search — through the EarningsCalls.dev API.

Get the API View API docs →

This call discussed

For developers and AI pipelines

Programmatic access to Penguin Solutions, Inc. earnings transcripts and 132,000+ others is available through the EarningsCalls.dev REST API. Plans from $39.99/month — full transcripts, speaker segments, full-text search, and the recently-added /api/v1/transcripts/recent polling endpoint for ETL pipelines.

More tools for stock research

OptionsBell: an unusual options activity scanner that tracks options flow across all US stocks and emails you when large trades appear in Penguin Solutions, Inc. or any ticker on your watchlist.

StockMarketScan: a stock screener with alerts: 24 pre-built screens refreshed every 5 minutes, watchlists and email notifications. Both offer API and MCP access and fair monthly pricing.