Penguin Solutions, Inc. (PENG) Earnings Call Transcript & Summary
April 1, 2026
What were the key takeaways from Penguin Solutions, Inc.'s April 1, 2026 earnings call?
In the second quarter of fiscal 2026, Penguin Solutions, Inc. reported net sales of $343 million, down 6% year-over-year, while non-GAAP diluted earnings per share remained flat at $0.52. The company raised its full-year guidance for net sales growth to 12% and non-GAAP EPS to $2.15, driven by strong demand in the integrated memory segment, which saw a 63% year-over-year increase. However, advanced computing sales were lower than expected, prompting a downward revision in guidance for that segment, reflecting a transition away from hyperscale customers.
What topics did Penguin Solutions, Inc. cover?
- Revenue Guidance Increase: Penguin Solutions raised its full-year net sales and EPS outlook, now projecting 12% growth and $2.15 EPS, up from 6% and $2.00, respectively. Management stated, "We are raising our outlook for our integrated memory business, fueled by AI-driven demand, strong execution by our team and favorable pricing dynamics."
- Advanced Computing Segment Decline: Advanced computing net sales were $116 million, down 42% year-over-year, attributed to the wind down of the Penguin Edge business and timing of large deployments. Management noted, "Our revenue lags about 3 to 6 months from the time of the bookings," indicating future revenue may improve as bookings grow.
- Strong Memory Demand: The integrated memory segment achieved net sales of $172 million, reflecting a 63% increase year-over-year. Management highlighted, "AI-driven demand remains strong across networking, telecommunications and computing market segments," reinforcing the segment's growth potential.
- AI Infrastructure Transition: Management emphasized the shift towards AI inference workloads, stating, "AI is moving from experimentation to production with workloads increasingly shifting towards real-time inference." This transition is expected to expand the company's addressable market.
- Customer Base Diversification: Penguin Solutions is transitioning from a hyperscale-focused customer base to a more diversified one, with non-hyperscale AI/HPC net sales growing 50% year-over-year. Management noted, "We expect further diversification in the second half of the fiscal year," indicating a strategic shift.
What were Penguin Solutions, Inc.'s April 1, 2026 results?
- Net Sales: $343 million (vs $365 million last year, down 6% YoY)
- Non-GAAP EPS: $0.52 (flat YoY)
- Integrated Memory Net Sales: $172 million (up 63% YoY)
- Advanced Computing Net Sales: $116 million (down 42% YoY)
- Non-Hyperscale AI/HPC Growth: 50% (year-over-year growth for the first half)
- Non-GAAP Gross Margin: 31.2% (up 0.4 percentage points YoY)
Penguin Solutions is strategically positioned at the intersection of AI infrastructure and memory, with a strong focus on product innovation and customer diversification. While the raised guidance for memory is a positive catalyst, the decline in advanced computing sales and potential margin pressures present risks that investors should monitor closely.
Earnings Call Speaker Segments
Suzanne Schmidt
executiveThanks, operator. Good afternoon, everyone, and I appreciate you being with us for today's Earnings Conference Call and Webcast, where we'll go over Penguin Solutions' Second Quarter Fiscal 2026 Results. Joining us are Kash Shaikh, our Chief Executive Officer, and Nate Olmstead, our Chief Financial Officer. The slide presentation and press release that go with this call are available in the investor relations section of our website, and we'd encourage you to visit that site during the quarter to get the latest information about the company. I also want to remind all of you to review the note regarding the use of forward-looking statements contained in the press release and the earnings call presentation. Please be aware that on this conference call, the company will make projections and forward-looking statements, which include, but are not limited to, statements regarding market demand, technology shifts, industry trends, and the company's growth trajectory and financial outlook, business plans and strategy, including investment plans, product development and roadmap, anticipated sales, orders, revenue, and customer growth and diversification, and existing and potential strategic agreements and collaborations. Forward-looking statements rest on current beliefs and assumptions, are not guarantees of future performance, and are subject to risks and uncertainties, including, without limitation, the risks and uncertainties reflected in the press release and the earnings call presentation filed today, as well as in the company's most recent annual and quarterly reports. The forward-looking statements speak only as of the date they are made, and except as required by applicable law, we assume no responsibility to publicly update or revise any forward-looking statements. We will also discuss both GAAP and non-GAAP financial measures. Non-GAAP measures should not be viewed in isolation from, as a substitute for, or as superior to our GAAP results. We encourage you to consider all measures when analyzing our performance. A reconciliation of the GAAP to non-GAAP measures is included in today's press release and accompanying slide presentation. With that, let me now hand the call over to Kash Shaikh, CEO. Kash?
Kash Shaikh
executiveGood afternoon, and thanks for being here on our Q2 FY 2026 earnings call. Taking part in my first earnings call as Penguin Solutions' CEO is something I'm excited about. To begin, I'd like to express appreciation to Mark Adams for the leadership he provided and the solid base he established. From the time I came aboard in early February, I've invested a great deal of time with customers, partners, and our teams across the globe. The company's strength has been evident to me, in both our technology and the relationships we hold with customers. Here's what stands out clearly: AI is transitioning out of experimentation into production, and workloads keep moving toward real-time inference. That shift is already showing up as customer demand beyond hyperscale throughout enterprise, Neocloud, and Sovereign AI markets. Our addressable market should grow because of this transition, and demand for integrated AI infrastructure ought to rise, a space where Penguin is already winning. We observe this directly in how broad our deployments are, ranging from a Sovereign AI factory, Hain, in South Korea, to enterprise voice AI with Deepgram, to large-scale research systems with Georgia Tech, plus an expanding pipeline spanning all three market segments. This opportunity matters so much because AI's architecture is shifting too. Model training was mostly compute bound. Inference, though, agentic AI is memory bound and latency sensitive; our belief is that this is pushing a re-architecture of the data center spanning compute, memory, interconnect, and software. We likewise see AI fueling memory demand, not just for the High Bandwidth Memory or HBM paired with GPUs or other accelerators, but for general-purpose memory as well. General purpose compute surrounds every GPU build-out, and whether it's reinforcement learning pipelines or inference serving. That workload operates on processors supported by substantial memory content throughout the entire system. Memory markets are cyclical, yet our view is that AI is layering in more durable demand for memory. As AI factories grow, I anticipate customers will place growing priority on partners who deliver with speed and precision. That comes alongside full stack AI factory platform capabilities, such as compute, scalable memory systems, cluster management software, end-to-end services, and a partner ecosystem, to provide a differentiated solution. Time to deployment is now directly tied to time to first token. Given this backdrop, we are shaping Penguin into an AI factory platform company. Six core elements form the basis of our AI factory platform. First, Penguin ICE ClusterWare, our AI infrastructure management software. Second, our new Penguin Memory AI line of systems, purpose-built for AI inference workloads. Third, Penguin Advanced Computing systems tuned for AI workloads. Fourth, Penguin OriginAI factory architectures, our reference designs for AI factories. Fifth and sixth, end-to-end services and our partner ecosystem. Production-grade AI factories call for full stack design spanning compute, memory, storage, networking, and software. We work with leading AI companies, among them NVIDIA and SK Telecom, and partners like Dell. We also provide complete end-to-end services covering design, build, deploy, and managed services. We sit in a strategic spot where AI infrastructure meets memory, and we've built a long track record in both areas. Hardly any companies, if any at all, bring these capabilities together at scale. Our view is that combining our expertise in AI infrastructure and memory puts us in position to satisfy the changing needs of AI infrastructure as it moves toward inference workloads. That underpins our capacity to create differentiated solutions. With the momentum visible in our AI infrastructure business and the sizable market opportunity in front of us, our attention is squarely on this space; we intend to put more investment into our AI factory platform to speed up growth in our AI business, particularly in product innovation, go-to-market, and customer engagement. At the NVIDIA GTC conference in March, we unveiled two solutions centered on AI inference that fit this strategy. The first is the Penguin Memory AI server. Leveraging our Compute Express Link, or CXL-based memory expansion capabilities, we launched a new family of scalable memory systems dubbed Memory AI. CXL is a high speed interconnect allowing scalable shared memory across GPUs and CPUs; we also announced that our new Memory AI KV cache server is immediately available. In this context, KV, or key-value cache, holds inference context in order to speed up large language model responses. The second is the broadening of our OriginAI factory architecture portfolio, which now features blueprints tackling the bigger workloads and the low latency requirements of AI inference. We're also continuing to broaden ClusterWare's capabilities on the path toward a unified control plane for AI factory infrastructure, bringing in the open ecosystem to provide repeatable production scale deployments. To speed up innovation and reinforce our leadership team, we recently named Ian Colle as Senior Vice President and Chief Product Officer. Ian carries over twenty years of experience constructing AI infrastructure platforms and scaling high-performance computing, most recently with Amazon Web Services. HPCwire recently put him on its People to Watch 2026 list, which speaks to his standing in the industry. Let me now speak briefly about how we did in the second quarter. During Q2, net sales came in at $343 million. Non-GAAP gross margin reached 31.2%. Non-GAAP diluted earnings per share totaled $0.52. Those results show robust demand and execution in memory along with ongoing progress in our AI/ HPC business. Before I move to the segments, I want to discuss our revised outlook. As Nate will explain in more detail, after our solid Q2 net sales and EPS performance, we are lifting the midpoint of our full year net sales and EPS outlook. We are raising our outlook for our Integrated Memory business, driven by AI-related demand, strong execution from our team, and favorable pricing dynamics. Even though our outlook for Advanced Computing net sales in the second half is below what we previously expected, we're encouraged by the strong year-over-year bookings growth in Q2 for our non-hyperscaler AI/ HPC business, which featured five new AI/ HPC customer wins, bringing our first half total this year to seven new AI/ HPC logos, versus three in the first half of last year. With that context, let me walk through each of the segments in more detail. Beginning with Advanced Computing, net sales for the quarter came in at $116 million, which was 34% of total company net sales and was down year-over-year. Advanced Computing net sales in the second quarter reflect both the timing of large deployments and our move away from hyperscaler concentration; they also reflect the wind down of our Penguin Edge business that we previously disclosed. We believe that diversifying net sales and winding down Penguin Edge will strengthen the long-term quality of the business. As I mentioned, we are shifting our AI infrastructure business away from hyperscaler concentration toward a more diversified customer base spanning enterprise, Neocloud, and Sovereign AI. This shift is showing very encouraging progress, but there is still more work for us to do. Non-hyperscale AI/ HPC net sales rose 50% year-over-year in the first half of the year, accounting for over 40% of first half segment net sales, supported by strong non-hyperscale year-over-year booking growth during the quarter, including five new AI/ HPC logos across financial services, biomedical research, and energy; we expect additional diversification in the second half of the fiscal year. Our AI/ HPC pipeline keeps getting stronger, with opportunities to acquire additional logos in the second half of the fiscal year across enterprise, Neocloud, and Sovereign AI customers.
Kash Shaikh
executiveLike we've said before, these deals usually move through several stages — prospecting, design, award, then contracting, and finally building and deploying the system — and that takes quite a while. This sales cycle can stretch out, frequently running 12 to 18 months, and it can make quarterly net sales fluctuate, but it also helps us build deeper ties with customers, generate repeat business, and achieve growth that holds up better over the long term. The direction the business is heading, along with the signals coming from the market, gives me encouragement. Outside of the figures themselves, we're noticing more activity in particular enterprise verticals; take, for instance, the collaboration we announced not long ago with Deepgram and Dell aimed at supporting enterprise voice AI deployments. That win underscores rising demand for inference infrastructure that offers low latency and production scale in real-time applications. For this engagement, Penguin handled the design and deployment of an optimized inference environment, which was built on Dell PowerEdge servers along with NVIDIA RTX PRO 6000 Blackwell GPUs. This solution supports Deepgram's speech-to-text, text-to-speech, and voice agent capabilities for applications in the healthcare and retail sectors. This case study further shows how expertise in design and integration produces value that sets us apart. As inference workloads grow, we anticipate deployments of this kind will drive AI infrastructure demand to an increasingly significant degree. Georgia Tech's AI makerspace, which we developed alongside NVIDIA, stands out as a strong illustration. Our ties with Georgia Tech keep expanding, and that confirms Penguin's capacity to help organizations transition smoothly from concept to production-grade AI infrastructure; now moving on to Integrated Memory. Net sales during the quarter reached $172 million, which works out to 50% of total company net sales, and rose 63% year-over-year. Demand fueled by AI stays robust across the networking, telecommunications, and computing market segments. Pricing dynamics worked in our favor, and even though supply stayed tight, we kept managing the constraints well thanks to our supplier relationships and disciplined procurement. Looking at the bigger picture, our AI/ HPC and memory segments combined let us integrate compute and memory architecture in ways that satisfy what production AI environments require. Memory architecture is playing an increasingly central role in AI performance, especially as inference workloads grow. The early investments we made in CXL put us in a good position as customers assess more dynamic memory architectures. On top of that, we're starting to see this demand show up in customer deployments, among them a recent sizable order for CXL cards from a generative AI company that builds solutions for inference workloads; this strengthens our strategic position where memory and AI infrastructure intersect, letting us capture the next phase of AI, one centered on inference that powers agentic AI workloads; these solutions go to enterprise AI infrastructure buyers, the very customers we serve in our AI /HPC business; for instance, we sold our CXL-powered KV cache servers to a tier one financial institution for its on-premise AI factory. At the same time, we keep pushing forward on development of our Photonic Memory Appliance, or PMA, previously called OMA, which is built to extend memory capacity and bandwidth for large scale AI environments. We got in early as an investor in Celestial AI, a Photonic Memory business, which speaks to how long we've concentrated on innovating in memory architecture and how soon we recognized that optical interconnects would matter greatly for the next generation of AI systems. Marvell has just bought Celestial AI in a deal worth multiple billions of dollars. In addition to the share of the proceeds that came to us as an investor from that acquisition, we are setting ourselves up to grow in this market going forward. As inference workloads grow larger, things like PMA can assist in tackling the major memory scaling problems that next-generation AI systems face. And finally, LED. Net sales came to $56 million for the quarter, which equals 16% of total company net sales, and they fell 7% compared with the prior year. Focused leadership and dedicated operational discipline continue to characterize how this business runs. Even though conditions in the market are still mixed, we keep taking a disciplined stance on investment and on allocating capital. Our attention is on getting the most value out of the portfolio while directing resources toward the places where we foresee the strongest returns over the long term. To wrap up, demand for data center AI infrastructure and memory is growing quickly. AI factories are turning into infrastructure that delivers artificial intelligence across many industries. With AI moving toward inference and agentic systems and growing across large enterprise, Neocloud and Sovereign AI environments, we anticipate that demand will pick up speed. Meanwhile, memory is turning into both a defining constraint and a defining opportunity. Penguin is positioned right where AI infrastructure and memory innovation meet, and we think that's a strong place to be. Our focus is unambiguous. There are four areas we're putting first. First, putting money into product innovation throughout our AI factory platform, especially where AI infrastructure and memory converge, to generate profitable growth. Second, carrying things out quickly and accurately. Third, strengthening customer relationships and our ecosystem to underpin long-term growth. Fourth, keeping our customer base diversifying while working toward growth that is steadier and easier to predict. This focus, we believe, leaves us well placed to perform in a market that's changing fast while we keep building a business that lasts and can scale. Having said that, I'll hand things over to Nate.
Nate Olmstead
executiveThanks, Kash. 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 site. So, moving on to what we delivered in the second quarter. Penguin Solutions posted total net sales of $343 million for the quarter, a 6% decline from the same period a year ago. Non-GAAP gross margin reached 31.2%, an improvement of 0.4 percentage points relative to Q2 of last year. At 13.2%, non-GAAP operating margin was 0.2 percentage points lower than a year ago. Non-GAAP diluted earnings per share came to $0.52, unchanged from the prior year. During the second quarter of fiscal 2026, net sales across our services business as a whole were $64 million, representing 1% growth over the prior year. Product net sales for the quarter reached $279 million, an 8% decrease compared with the prior year. Here's how net sales broke down by business segment: Advanced Computing recorded Q2 net sales of $116 million, equal to 34% of total company net sales, and that figure fell 42% year-over-year. The drop in sales stems from two things: the continuing wind down of our Penguin Edge business, plus hyperscale hardware sales that occurred in Q2 last year but did not repeat in Q2 this year. Looking more closely at what happened in Advanced Computing, our non-hyperscale AI /HPC net sales declined 35% year-over-year during the quarter, yet they rose 50% across the first half of the year. Because this business is project-based, with sales that can be lumpy from one quarter to the next, we think examining the multi-quarter trend offers a useful way to assess growth in this portion of our business. Alongside solid first half growth in our non-hyperscale AI/ HPC business, we keep making good progress toward diversifying our net sales into new customer segments; during the first half of the year, the non-hyperscale AI/ HPC business accounted for more than 40% of total Advanced Computing net sales, compared with approximately 20% in the first half of last year. We anticipate that our mix of net sales from enterprises, Neoclouds, and Sovereign AI customers will rise further over the second half of this fiscal year. Turning to Integrated Memory, Q2 net sales reached $172 million, amounting to 50% of total company net sales and growing strongly at 63% year-over-year; in Optimized LED, Q2 net sales were $56 million, equal to 16% of total company net sales and down 7% versus the same quarter last year. Non-GAAP gross margin for Penguin Solutions during the second quarter was 31.2%, which represents a 0.4 percentage point gain year-over-year and a 1.2 percentage point gain sequentially, with strong margin performance in each business driven primarily by product mix in Advanced Computing, favorable pricing in memory, and tariff recovery in LED. For the second half, we currently expect lower gross margins, driven by a higher mix of lower margin AI hardware and memory sales, rising memory costs in our AI factory solutions, and less tariff cost recovery in LED. Non-GAAP operating expenses for the second quarter totaled $62 million, which is 3% lower year-over-year and roughly flat sequentially. For the second half, we anticipate a slight quarter-over-quarter rise in operating expenses, which mirrors typical seasonal patterns along with greater spending on R&D, such as for our ClusterWare software and Memory AI solutions. Non-GAAP operating income in Q2 came to $45 million, a decline of 8% from the same period last year and a gain of 9% from the previous quarter. Compared with a year earlier, operating margins fell 0.2 percentage points, yet they rose 1.1 points on a sequential basis, helped by stronger quarter-over-quarter gross margins in both Memory and Advanced Computing. Non-GAAP diluted earnings per share in the second quarter reached $0.52, unchanged from Q2 a year ago and 7% higher than the preceding quarter. Adjusted EBITDA in the second quarter stood at $50 million, 6% lower year-over-year and 11% higher than the prior quarter. Now moving to the balance sheet. On working capital, net accounts receivable came to $371 million, versus $330 million 12 months earlier, with that rise attributable to larger memory sales volumes and shifts in how sales were distributed across the quarters. Day sales outstanding remained healthy at 50 days, in line with the same period last year and one day lower than the previous quarter. Inventory reached $322 million by the close of the second quarter, an increase from $200 million 12 months earlier, driven by higher memory costs, expansion in our memory business, and strategic buying to meet Memory and AI demand during the second half of the year. Days of inventory came in at 51 days, compared with 37 days a year earlier and 38 days in the prior quarter, largely because of our strategic memory purchases along with the timing of receipts and shipments. Accounts payable finished the quarter at $401 million, rising from $238 million 12 months earlier, mainly as a result of higher memory costs, growth in our memory business, and the timing of purchases and payments. Days payable outstanding was 63 days, versus 44 days a year ago and 55 days in the prior quarter. Those year-over-year and quarter-over-quarter changes stemmed from the timing of purchases and payments. Our cash conversion cycle was 38 days, five days better than Q2 a year ago and three days higher than the previous quarter because of the timing of purchases and payments. As we have done in the past, days sales outstanding, days payables outstanding, and inventory days are computed using a gross sales and gross cost of goods sold basis, which in the second quarter were $672 million and $578 million respectively. As a reminder, the gap between gross and net sales mostly relates to our memory business' logistics services, which are recorded on an agent basis, so we recognize only the net profit on logistics services as net sales. Cash, cash equivalents, and short-term investments ended the second quarter at $489 million, $158 million below Q2 a year ago and $28 million above the prior quarter. That year-over-year swing was mostly the result of proceeds from issuing preferred shares in Q2 of last year, offset by repayments of debt on our term loan in Q4 of last year. Taking the cash movement in order, the rise came from money produced by operating activities plus roughly $32 million in proceeds tied to disposing of our Celestial AI investment as part of that company's sale to Marvell Technology. Working against those inflows was the stock buyback we carried out during the quarter. We finished the quarter carrying $450 million in debt, which is $20 million lower than the prior quarter after retiring our 2026 convertible notes. All told, we ended the quarter sitting in a net cash position, and given the debt maturity schedule we currently have, nothing further is scheduled to come due until 2029. Cash flows from operating activities in the second quarter came to $55 million, against $73 million from operating activities a year earlier. The year-over-year decline in quarterly cash flow stemmed mainly from net working capital investments aimed at supporting growth across the back half of this fiscal year. If you're among those watching capital expenditures and depreciation, capex came in at $2 million for the second quarter, while depreciation registered $5 million during that quarter. To close out the cash flow picture, we used $32 million buying back about 1.7 million shares during the second quarter through our stock repurchase program.
Nate Olmstead
executiveAs of February 27th, 2026, $64.5 million in the aggregate was still open for buying back our common stock under the authorizations currently in place. Let's move on to the outlook now. Because of how well we did in half one and a better half two view for our Memory business, we're lifting our outlook for the whole company's net sales and non-GAAP diluted EPS for the year, and at the midpoint that now means 12% net sales growth and $2.15 of non-GAAP diluted EPS, compared with our earlier outlook of 6% net sales growth and $2 of non-GAAP diluted EPS. Just to remind everyone, our full year outlook takes it as a given that we'll keep broadening the mix of customers we sell to, and it leaves out any Advanced Computing AI hardware sales to hyperscale customers. In line with the assumptions we laid out last quarter, our FY 2026 financial outlook captures the continuing wind down of our high margin Penguin Edge business. By the close of fiscal 2026, we expect sales from that business to basically come to an end. Taken together, those two assumptions in our FY 2026 outlook still amount to roughly a 14 percentage point unfavorable year-over-year hit to our total company net sales growth and roughly a 30 percentage point unfavorable hit to Advanced Computing. That being said, our full year net sales outlook comes with these full year growth ranges by segment; for Advanced Computing, we now expect full year net sales to move between -25% and -15% year-over-year. Even though our Advanced Computing net sales outlook for this fiscal year sits below what we previously forecast, our AI/ HPC bookings give us encouragement, among them several new logos and growth in our pipeline. As was true before, this outlook incorporates the Penguin Edge and hyperscale hardware sales effects I mentioned a moment ago. For Memory, we now expect net sales to rise between 65% and 75% year-over-year, on the strength of robust demand and a pricing environment that's favorable. For LED, our expectation is unchanged: net sales will fall between -15% and -5% year-over-year. Our non-GAAP gross margin outlook for the full year now stands at 28% ±0.5 percentage points; we took that gross margin outlook down by one percentage point to reflect a larger mix of memory sales, which carry a gross margin below our company average, along with higher memory costs in our AI hardware business. Our full year expectation for total non-GAAP operating expenses is still $250 million, and we've tightened that range to ±$5 million. For FY 2026, we now expect a non-GAAP diluted share count of roughly 53 million shares, lower than our prior outlook, chiefly because of the effect of our recent share repurchases. Our non-GAAP full year diluted earnings per share is now projected at approximately $2.15 ±$0.15. Our forecasted FY 2026 non-GAAP tax rate is still 22%. Although we anticipate applying this normalized non-GAAP tax rate across FY 2026 and after, over the long term the non-GAAP tax rate could shift for many reasons, among them the fast-changing global and U.S. Shifts in tax rules, major adjustments to the mix of where we earn profits, or modifications to how we run the company or what we sell could all play a role. When we look ahead to fiscal year 2026, our projections rest on conditions as they stand right now — conditions that take into account, alongside other factors, the worldwide macroeconomic picture and the supply chain pressures that continue, particularly those tied to our Advanced Computing and Integrated Memory businesses. That covers longer wait times for some parts that go into our complete solutions, which affects the pace at which we can scale up both existing and new customer projects and ship what customers have ordered. Memory costs being higher across the industry are also factored into what we expect, and that could dampen customer appetite for what we offer and could bring down gross margins in our Advanced Computing and Memory businesses. On the whole, we think that staying focused on execution, keeping tight control of expenses, and having a strong balance sheet give us a solid base for ongoing profitable growth. We believe these strengths will help us keep making progress as we go after chances to build long-term shareholder value. For more specifics, please see the non-GAAP financial information section and the tables reconciling GAAP to non-GAAP measures in our earnings release and the investor materials posted on our website. With that, operator, we're ready to take questions.
Katherine Murphy
analystI'd like to ask about the higher memory segment guidance, the 65%-75% growth. How much of that comes from better pricing that's working in our favor versus demand tied to new product categories? And then, as a follow-up, what should we make of how this affects the operating margin outlook for that segment, plus the investments required in new technologies such as CXL and Photonic Memory Appliances? Thanks so much.
Nate Olmstead
executiveHi, Kath, this is Nate. Regarding the memory outlook, we're quite happy with the demand we're witnessing and the favorable pricing environment we're seeing. I'd say that for the lift we're seeing in the second half, pricing accounts for most of it, though demand is also extremely robust across telco and networking. Demand driven by AI is simply very strong; actually, reaching the top end of that outlook basically comes down to whether we can secure materials, and that's truly the sole obstacle we currently see to lifting that outlook in the second half. We're hunting for materials. Where possible, we're leveraging the balance sheet to make strategic purchases ahead of time. Demand in memory is extremely strong. When it comes to the investments, we've accounted for them in the outlook, so I've held OpEx for the year at $250 million ± $5 million. As we always do, we're balancing the portfolio to seek opportunities that let us speed up our investments in innovation in AI or in the Memory solutions we've discussed. All of that is baked into the outlook. I anticipate operating margins for Memory will stay fairly healthy in the back half of the year. I do anticipate some gross margin pressure in AI, given a greater mix of new hardware shipments in the second half, along with accounting for some of the elevated memory input costs in that business.
Katherine Murphy
analystThank you very much, Nate.
Nate Olmstead
executiveYou bet.
Brian Chin
analystHello there. Thanks a lot for giving us the chance to pose a handful of inquiries, and good afternoon to everyone. Perhaps to start things off, I suppose regarding Advanced Computing, what was it that shifted, leading you to bring down the midpoint from the guidance you previously provided to the fresh range you've now shared? Also, could you walk us through how much of that new range's midpoint is already booked?
Kash Shaikh
executiveA key element here is the delay separating when we book business and when it shows up as revenue. There's roughly a three to six month gap between the booking and the revenue recognition. What mainly causes that is how deployments are timed in certain situations. Availability of materials, that kind of thing. Considering our position within the fiscal year, only five months are left. Looking ahead, the majority of bookings we anticipate probably won't turn into revenue during this fiscal's second half, yet we're confident it will clearly provide a beneficial effect as we move into the first half of the following fiscal year. That's among the reasons we're reducing the outlook for Advanced Computing because of the deployments; we're witnessing robust momentum in our pipeline and in bookings. For the non-hyperscale AI /HPC business, which matters a great deal strategically to us, bookings increased quite dramatically in Q2, and seeing this progress encourages us; five new logos were closed with AI /HPC in Q2 and, across the first half, that brings the count to seven new logos, versus three new logos a year ago; we have strong confidence in our capacity to deliver. Timing is the core problem right now.
Brian Chin
analystAll right. Yes, I appreciate that, Kash. It seems you're quite fully committed through the lowered guidance for the fiscal back half, and certain of those fresh bookings extend past what would be a 6-month horizon. Also considering growth within the business. Clearly there's that sort of obstacle you assisted in making clear regarding the decline in hardware revenue tied to the new hyperscaler, plus the winding down of Penguin Edge. You know, a 30 percentage point effect if we offset that against the guidance, perhaps 10% growth for this year, net of that within that segment. Going ahead, as you look over the business, and you haven't held the position very long, and you consider what that sort of like-for-like growth rate was or is trending toward for this fiscal year, how are you viewing sort of desired growth rates for the Advanced Computing business going forward?
Kash Shaikh
executiveLet me share one data point with you. During the opening six months of this fiscal year, our non-hyperscale AI /HPC business saw net sales rise roughly 50% compared with the same period a year earlier, and it made up 40% of the total Advanced Computing mix — nearly double what we finished the prior fiscal year at. That growth is meaningful both in bookings and in the revenue we're seeing, and we anticipate it carrying on. As we keep closing bookings and moving the pipeline through, we're seeing a robust pipeline across all three segments I referenced: enterprise on-prem AI deployments, a lot of activity with Sovereign AI customers, and Neocloud customers as well.
Brian Chin
analystOkay. Thank you.
Kash Shaikh
executiveThanks, Brian.
Matt Calitri
analystHi everyone. This is Matthew Calitri calling in from Needham. Appreciate you letting us ask. When it comes to those fresh Memory launches, does that signal a change in direction there? I ask because previously, the company discussed the smaller, specialized segments within the Integrated Memory business more, and pointed out that areas such as CXL were still at an early stage. Now it sounds as though memory will be a bigger contributor within this AI factory platform. So I'm just wondering whether anything has shifted, and what makes you confident that demand here will hold up over time.
Kash Shaikh
executiveYes. This does form part of our strategy. The Memory AI appliances we brought to market roughly a month back, beginning at GTC, are part of us putting more investment behind our AI factory platform strategy. That strategy has six components, and Memory AI is among the strategic ones, and it's very well timed given how AI is moving away from model training toward inference. With workloads centered on inference, you know, memory turns into a bigger necessity, thanks to reduced latency along with a larger context size for inference that powers agentic AI. This matters a great deal strategically for our business; as a matter of fact, we hold the leading position in this space, capitalizing on our distinctive spot where Memory and AI infrastructure meet, and bringing together deep knowledge and architecture; we rolled out this Memory AI KV cache server as one of the offerings within the Memory AI line. Additional products are in the works, and we'll keep investing and, as a matter of fact, put in more money here to seize the market opportunity, since the timing could not be better and we lead in the Memory AI family of products. As a proof point, a new logo we brought in, a tier one financial institution, isn't just having us deploy AI infrastructure and an AI factory deployment for them, they additionally bought our CXL-based KV cache server, which serves as evidence that customers are moving beyond training and bringing AI on premise into their factories, deploying on site, concentrating on inference and powering agentic AI. This is highly strategic for us, and the timing is exactly right. We anticipate seeing this demand, and we intend to keep putting investment into this space.
Matt Calitri
analystFantastic. That's wonderful to hear. Nate, now that there's a new CEO in place and with certain elements shifting regarding sales cycles and supply chains, did the way you approach guidance shift at all, or did you build in any extra conservatism? Any detail on the pluses and minuses would be useful.
Nate Olmstead
executiveYeah. Hey, Matt. The philosophy hasn't changed. You know, Kash and I got on the same page quite quickly, I believe, regarding how we approach monitoring the business and evaluating things. Actually, I believe that with our new CRO, who joined a couple quarters ago, he's done a good job of bringing greater rigor into the planning process for our AI business and just enhancing the visibility there somewhat; from a supply chain perspective it's a difficult environment, and we've, naturally, accumulated substantial experience handling supply chain within our Memory business, and I think that, you know, in conditions such as these that works in our favor.
Matt Calitri
analystAwesome. Thank you guys so much.
Nate Olmstead
executiveThank you.
Manmohanpreet Singh
analystHi there. I appreciate you fielding my question; MP here, speaking for Samik Chatterjee. To kick things off, I'd like to dig deeper into the Advanced Computing outlook you provided. You noted a three-to-six-month delay affecting revenue that will be recognized during your second half. That said, did anything shift when it came to bookings made in the first quarter, or was there any variation compared with what you had anticipated for 2Q? I've also got a follow-up.
Nate Olmstead
executiveYeah. Hey, MP. What I'd say is that bookings came in strong during Q2, with quite solid sequential and year-over-year increases. My sense is that the deployment timeline has stretched somewhat because of certain supply constraints. Specifically where Memory is concerned, lead times have extended a bit further. Still, we're very happy about the five new logos. And, as I see it, demand remains healthy. The pipeline is showing solid momentum, and it's broadening well across the non-hyperscale categories, including enterprise, Neocloud, and Sovereign. Our confidence in the demand is quite high. What we're dealing with here is simply a matter of timing as we work to turn bookings into revenue.
Manmohanpreet Singh
analystAll right. For my follow-up, I'd like to ask about Advanced Computing as well as the business tied to your AI factory. For instance, when NVIDIA develops its own reference designs aimed at factory-level solutions, how does that stack up against what you offer? Is it something that gives you a boost, or does it work against you? Could you walk us through that?
Kash Shaikh
executiveOur view is that this works in our favor. We collaborate extremely closely with NVIDIA, and take one of the wins I brought up — the tier one financial institution from not long ago — where our Memory AI product was part of that deal. NVIDIA was deeply involved alongside us, and we're engaging with NVIDIA, making use of their reference design, then pairing that with our AI factory platform and building on NVIDIA's NVL, just to give one instance, so we can deliver a full stack to our customers. The blueprints they offer actually reinforce our AI factory platform and the pieces that constitute it. Those blueprints genuinely have us enthusiastic, and we're coordinating tightly with NVIDIA to seize the opportunities, particularly given that NVIDIA keeps sharpening its enterprise focus. That fits with our strategy and how we go to market.
Manmohanpreet Singh
analystI appreciate that, and many thanks.
Nate Olmstead
executiveThank you, MP.
Ananda Baruah
analystThanks for letting me ask. Good afternoon. I've got a couple, if that's alright. Kash, and perhaps Nate too. In your earlier comments, you said momentum is picking up in Neocloud, Sovereign, and Enterprise. You brought up one or two of the recent wins. Kash, I believe you made some remarks about inferencing specifically, or at least in general terms, agentic included. Can you offer us any concrete context regarding what your customers say their push in inferencing looks like at the moment, and perhaps how much agentic figures into that? Say we simply wanted a feel for the behavioral tone of customer activity across the past 90 to 180 days. Is there anything you could share that would make it a bit more experiential for us? I have one quick follow-up. Two things.
Kash Shaikh
executiveSure. With these customers, our view is that we're at an early stage of inference adoption, yet it's being deployed more and more as customers progress toward agentic. The chance to power agentic comes through inference. On the subject of inference, here's an illustration of why the architecture is shifting and why memory matters more and more for inference relative to model training. Suppose, for instance, you're writing a book, and you must compose a new sentence without memory serving as a supporting element — then you'd need to reread the whole book prior to writing that next sentence; with inference, you know, you're inferencing over plenty of data you already possess, and if there's a component where the book you've written up to that point is kept, then ahead of writing the new sentence, rereading the book isn't necessary. That's roughly how things are shifting for enterprises and other segments. We already see customers putting it into deployment, and the architecture is shifting, which explains why we not only have the chance and the edge in supplying them our AI infrastructure along with the services — more and more, we're witnessing demand for our Memory AI portfolio, where they're deploying AI infrastructure and increasingly inference. To deliver that memory element for inference, they require products of that kind, so that LLM responses can be far quicker than they'd be otherwise.
Ananda Baruah
analystUnderstood, that helps. Since others might be waiting, I'll squeeze in one final quick follow-up of my own. Regarding the CXL product, based on that earlier question, it seems you folks are somewhat more bullish on the CXL sleeve right now compared to roughly 90 days back. You've launched new items at GTC. Is that a fair characterization? Is it that thanks to those new offerings, and definitely a few of the NVIDIA reveals at CES too, you anticipate somewhat greater revenue somewhat sooner than you might have expected on the CXL front 90 days back? A quick second piece to that: does photonics have to function before you can genuinely achieve CXL amplification? In other words, must CPO or photonics work before you're truly able to amplify CXL and scale out or scale up? Thanks. That's all from me.
Kash Shaikh
executiveYes. I'll take your CXL question first. My view is CXL adoption is well-timed, considering the shift toward inference. As I noted, with inference, you require more memory to get faster LLM responses. What CXL, meaning Compute Express Link, gives you is the ability to share memory across GPUs and CPUs. It enables memory pooling now, and that's a benefit for inference workloads. Although CXL had clearly been around for the past, I'd say several quarters, it's inference uptake that's propelling CXL adoption, along with this deal I referenced where an order came in. That company is in fact an enterprise Generative AI firm focused on inference workloads, so it stands to reason CXL cards are a fit for them, since such workloads demand additional memory, and the pooling of memory that CXL enables between GPUs and CPUs benefits that type of customer. As for the photonic memory appliance we're developing through our partnership with Celestial AI, which is of course Marvell now, it delivers greater capability since, clearly, once you have photonic connectivity, your capacity to share memory grows. It elevates things to another level. That said, CXL on its own is a benefit. The photonic appliance can elevate it further. There's one more piece, the KV cache I brought up, the Memory AI KV cache server, which basically delivers far better responsiveness for larger context workloads, again, applied in inference. Different needs exist—you can view inference as having assorted memory-related requirements and workload types, with some being about latency; these pieces, whether CXL or the CXL-based KV cache delivering improved responses plus bigger memory sizes and larger context sizes, and then the next level up, Photonic Memory, constitute different use cases for inference. As inference becomes mainstream, we'll benefit from this fair portfolio supporting assorted inference use cases.
Ananda Baruah
analystThat background is really helpful. Many thanks.
Kash Shaikh
executiveAppreciate it. Thanks to you.
Kevin Cassidy
analystHi. Yes. I appreciate you fitting me in. Just wanted to ask about the gross margin for Memory. Your gross margin rose during the quarter, and Memory revenue increased substantially. I'd simply like to get a handle on what was driving that?
Nate Olmstead
executiveYeah. Of course, Kevin. We experienced a slight tailwind when it comes to memory margins. Part of that relates to mix. Demand came in somewhat stronger for flash, which for us represents a product with somewhat better margins inside the portfolio; beyond that, certain price increases allowed us to pick up a bit of extra margin, purely because of when we bought our inventory versus when, you know, shipments and sales to customers took place.
Kevin Cassidy
analystAlright. So when you're looking toward the back half of the year, you expect that to catch up with the price increases, I guess—relative to.
Nate Olmstead
executiveYes. Since the price increase is gradual, correct? If you're working with that assumption—that price increases will moderate—then we'd anticipate seeing reduced margin benefit from it, because there'd be a smaller timing gap between, or rather a smaller price fluctuation between when inventory is bought and when it's sold. You know, we've been leveraging the balance sheet to lock in inventory wherever possible. The market's tight, so supply isn't endless. Wherever we're able, we're drawing on the balance sheet to try to get a slight edge.
Kevin Cassidy
analystAll right. And perhaps while we're on the subject of memory, once you reach those CXL systems, do you anticipate that will carry a bigger margin compared with the module business?
Nate Olmstead
executiveYes, that's our expectation. It genuinely amounts to a solution. There's software involved, along with certain elements. There's also solid differentiation when it comes to the hardware. I view that as a good margin opportunity for us in the future.
Kevin Cassidy
analystOkay, great. Thanks.
Nate Olmstead
executiveThank you.
Kash Shaikh
executiveThanks, operator. It's clear to us that AI is moving in the direction of inference, and demand keeps widening past hyperscalers into enterprise, Neocloud, and Sovereign AI customers. This transition remains at an early stage, yet our customer demand, product innovation, and booking momentum together give us confidence about the road ahead. We think we're positioned well where AI compute infrastructure meets memory, and our efforts to diversify the customer base are progressing nicely. Strong execution is my priority — across product innovation, customer engagement, and diversification, along with disciplined capital allocation and investment in our AI/ HPC business to back the long-term growth. We're eager to share updates on how we're progressing.
Operator
operatorThis concludes today's call. Thank you for attending. You may now disconnect.
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.