Q3 2023 SMART Global Holdings Inc. Earnings Call

Participants

Ken Rizvi; Senior VP & CFO; SMART Global Holdings, Inc.

Mark W. Adams; President, CEO & Director; SMART Global Holdings, Inc.

Suzanne Schmidt; Head of Investor Relation; SMART Global Holdings, Inc.

Brian Edward Chin; Associate; Stifel, Nicolaus & Company, Incorporated, Research Division

Kevin Edward Cassidy; Senior Semiconductor Research Analyst; Rosenblatt Securities Inc., Research Division

Mark John Lipacis; MD & Senior Equity Research Analyst; Jefferies LLC, Research Division

Rajvindra S. Gill; Senior Analyst; Needham & Company, LLC, Research Division

Shek Ming Ho; Director & Senior Analyst; Deutsche Bank AG, Research Division

Thomas James O'Malley; Research Analyst; Barclays Bank PLC, Research Division

Presentation

Operator

Good afternoon. Thank you for attending today's SGH Third Quarter Fiscal 2023 Earnings Call. My name is Hannah, and I will be your moderator for today's call. (Operator Instructions) I would now like to turn the conference over to our host, Suzanne Schmidt, with Investor Relations. You may go ahead.

Suzanne Schmidt

Thank you, operator. Good afternoon, and thank you for joining us on today's earnings conference call and webcast to discuss SGH's Third Quarter fiscal 2023 results. On the call today are Mark Adams, Chief Executive Officer; Jack Pacheco, Chief Operating Officer; and Ken Rizvi, Chief Financial Officer. You can find the accompanying slide presentation and press release for this call on the Investor Relations section of our website. We encourage you to go to the site throughout the quarter for the most current information on the company. I would also like to remind everyone to read the use of forward-looking statements note that is included in the press release and the earnings call presentation. Please note that during this conference call, the company will make projections and forward-looking statements, including statements about the company's growth trajectory and its 2023 financial outlook. Forward-looking statements are based on current beliefs and assumptions and 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 are representative 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 considered in isolation from, as a substitute for or 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 with that, let me turn the call over to Mark Adams, CEO. Mark?

Mark W. Adams

Thanks, Suzanne, and thank you all for joining us today for our Fiscal 2023 Third Quarter Call. Over the past 3 years, we have made tremendous progress on the transformation of SGH. We have grown our top line, increased our gross margins, delivered strong earnings per share and strengthened our balance sheet by generating positive operating cash flow. Earlier this month, we announced that we had agreed to divest a majority stake in our Brazil Consumer Memory Business, enabling SGH to shift our business towards high performance, high availability solutions for enterprise customers. Although macroeconomic headwinds persist, we believe we are well positioned for long-term success, driven by secular tailwinds in AI, machine learning and data analytics. While Ken will review the financials in more detail, I would like to call out some highlights from our third quarter performance. Our team achieved strong results for the third quarter in what remains a very challenging global economic environment. Non-GAAP gross margins was 28%, up 230 basis points from the year ago quarter and non-GAAP diluted earnings per share totaled $0.66 on sales of $383 million. We generated strong cash flow from operations of approximately $41 million in the quarter and exited Q3 with a strong balance sheet, including record cash and cash equivalents of $401 million.

Now let me review each of our business lines. Starting with IPS, which is comprised of our Penguin Computing and Stratus Technologies brands. We design, manufacture, deploy and provide managed services for high-performance computing for the data center, the cloud and the edge. AI, machine learning and data analytics are becoming foundational technologies for the success of enterprises across a growing set of industries. We believe that Penguin Computing, which has more than 25 years of experience in the deployment of HPC Systems and Solutions is well positioned at the forefront of the generative AI revolution. Our team has designed, built and managed some of the largest and most powerful supercomputers in the world, including managing a total of more than 50,000 NVIDIA GPUs for AI training at scale. Our engagement model is focused on developing tailored solutions for our customers. We architect a system that integrates technologies such as compute, storage and networking that optimize for our customers' unique HPC and AI workload requirements. Our skilled ClusterWare software provides an intelligent suite of management functionality, including node provisioning, image customization and cluster monitoring while serving as a platform for additional software components. What we believe differentiates Penguin Computing from the competition is our ability to work with our customers from the design phase through integration, all the way through to implementation and management of large HPC and AI clusters.

Additionally, we are able to provide solutions on-premise, in the cloud and at the edge. In the third quarter, IPS sales totaled $171 million, which represented 45% of total SGH sales, reinforcing the transformation we are going through. Compared with the year ago quarter, IPS sales increased by 31%, excluding Stratus Technologies and were up 79% by including Stratus. Our service revenue, the majority of which is generated in IPS, represented 15% of total SGA's revenue in the third quarter. and reflects the incremental value we are providing to our customers. Our services include point-in-time services such as design and implementation as well as longer-term managed services. During the quarter, we continued to introduce new technologies and capabilities in the market. Penguin announced a new version of its flagship skilled ClusterWare software platform, which is a powerful management and monitoring tool that empowers customers to expand their own HPC clusters. This new version provides greater scalability, boost performance and delivers greater ease of use for users. Stratus announced the next generation of its fault-tolerant computing platforms that enable our customers to run business-critical applications, such as transaction and payment processing to industrial automation at the core, in the cloud and at the edge.

During the quarter, Meta announced the completion of the second phase build-out for their AI research SuperCluster known as RSC. As our implementation partner, Penguin worked with Meta to design the AI optimized infrastructure that powers their RSE, improving the overall cluster management and implement the system. The Navy DoD Supercomputing Resource Centers newer supercomputer named Nautilus is a Penguin True HBC system featuring 48 GPU nodes. This HPC system, which completed its final testing in April, enables scientists and researchers to use complex applications to unlock weather patterns and ocean modeling, providing predictive insights about the Earth climate and sustainability. Today, we are also announcing that Dave Lorello, who joined us as CEO of Stratus at the time of the acquisition will take on an expanded role as President of IPS, replacing Thierry Pellegrino. We believe that Dave's extensive experience building and leading technology teams while serving Fortune 100 enterprises will help us scale our customer engagements and overall operations. Prior to Stratus, Dave also worked in executive roles at Lucent and Digital Equipment Corporation. While there is a heightened level of customer engagement with regards to HBC-AI opportunities, we recognize that we are still in the early stages of generative AI and machine learning. As we have noted on prior calls, we expect IPS sales will be lumpy due to the deployment time line decisions and cycles of large customer installations. Over the long term, under Dave's leadership, we believe we are in a position to capitalize on these emerging market trends.

Shifting to Memory. Our Memory Solutions group is currently made up of 2 businesses: Specialty Memory, which is focused on specialty memory applications in the enterprise and our Brazil module business, primarily serving the consumer market. Overall, third quarter Memory revenue came in at $148 million or 39% of total SGH sales and was relatively flat with the second quarter. We have a long history of serving enterprise customers in the networking telecom and industrial end markets. An area of focus going forward is to leverage this expertise to help drive memory subsystem innovation in AI and machine learning. For example, our CXL offerings removed the performance and latency bottlenecks between GPUs, CPUs and memory and HPC applications. We are seeing strong design activity and customer interest for our CXL offerings and the wide variety of CXL architectures required by customers plays to our strength. In addition to our development efforts with CXL, we continue to implement leading-edge test processes to ensure high levels of quality and reliability for our enterprise customers. Our 0 failure rate or Zefr memory offering is gaining significant interest among major customers as it meaningfully reduces field failures and helps to maximize system utilization. During the quarter, we started shipping DDR5 Zefr memory models, which when combined with our DDR4 Zefr memory models give SMART modular key differentiation and high availability memory products. Our memory business continues to demonstrate relative stability throughout the current memory cycle due to the specialty nature of our value-add business model. In the third quarter, this translated into operating margins of approximately 8% for our memory business.

Now turning to our LED Solutions Group, which operates under the CreeLED brand and produces application-optimized LEDs for specialty lighting, video screens, gaming displays, horticulture, outdoor and architectural lighting. For the third quarter of fiscal 2023, LED Solutions totaled $64 million or 17% of overall SGH sales and were up 15% sequentially from what could have been the bottom in the second quarter. While customers are continuing to work down inventory levels, we are seeing customer design activity improving. Typically, as we come through these cycles, we look for this as a precursor to demand improving. And while we are still in early days in the cyclical recovery of the LED market, we are optimistic that the business will have a stronger fiscal 2024. Cree continues to be recognized as a leader in customer-focused innovation. During the third quarter, LED Magazine awarded 3 of Cree's LED's latest product releases with Brightstar awards for innovative LED products. The XLAMP Element G and Pro9 LEDs were recognized in the LED light sources category, while the Photophyll select LEDs were recognized in the horticulture SSL and control systems category. By continuing to invest in advanced LED technology, Cree is able to empower its customers with new and innovative LEDs to achieve remarkable system-level solutions. We believe that the combination of industry-leading technology in IP, a capital-light outsourced manufacturing model and disciplined expense control has positioned Cree well in these challenging times. We are starting to see signs of improving customer demand and expect revenue to be up modestly in the fourth quarter. I'll stop here and hand it over to Ken for a more detailed review of our third quarter financial performance and our guidance for next quarter. Ken?

Ken Rizvi

Thanks, Mark. I will focus my remarks on our non-GAAP results, which are reconciled to GAAP in our earnings release tables. Now let me turn to our fiscal third quarter 2023 results. Total SGH revenues were $383 million, and non-GAAP gross margins came in at 28% at the midpoint of our guidance. Non-GAAP diluted earnings per share were $0.66 for the third quarter. In the beginning of this fiscal year, we began providing the breakdown of our overall revenue by product and services. As a reminder, our services revenue includes longer-term services as well as point in time services such as logistics and implementation services. In Q3, our overall services revenue totaled $58 million, up from $48 million in the year ago quarter and helped by the inclusion of Stratus, which we acquired in the beginning of this fiscal year. Product revenues were $326 million. Third quarter revenue by business unit was as follows: IPS was $171 million. Memory was $148 million, and LED came in at $64 million. This translates into a sales mix of approximately 45% IPS, 39% memory and 17% LED. Non-GAAP gross margin for SGH in Q3 was 28%, up from 25.7% in the year ago quarter, driven primarily by IPS. Non-GAAP operating expenses for the third quarter were $70.9 million, down from $72.5 million in the second quarter of 2023. Operating expenses were down from the prior quarter, primarily due to continued cost containment initiatives as well as lower bonus accruals. Operating expenses benefited in the third quarter of 2023 from $0.4 million in financial credits in Brazil, which was down from $1.4 million in the second quarter of 2023 and down from $3.3 million in the year ago quarter. This credit is expected to provide minimal benefit in our fourth quarter of 2023.

Non-GAAP diluted earnings per share for the third quarter of 2023 was $0.66 per share compared with $0.87 per share in the year ago quarter. Included in our non-GAAP EPS for the third quarter is a reduction in taxes of approximately $6.5 million or $0.13 per diluted share as we anticipate using additional prior year net operating losses to offset U.S. taxable income for fiscal 2023, which also has the benefit of reducing our cash taxes for the year. Adjusted EBITDA for the third quarter of 2023 was $45 million or 12% of sales compared to $64 million or 14% of sales in the year ago quarter. Turning to balance sheet highlights. For working capital, our net accounts receivables totaled $244 million compared with $229 million last quarter. Days sales outstanding came in at 42 days, up 6 days from the last quarter, primarily due to the timing of invoicing and collections for IPS. Inventory totaled $226 million at the end of the third quarter, down from $294 million at the end of the prior quarter. The decrease in inventory was driven primarily by a reduction in IPS and memory inventories in the third quarter. Inventory turns were 7.6x in the third quarter versus 6.3x in the prior quarter. and consistent with past practice, net accounts receivables, sales outstanding and inventory turnover are calculated on a gross sales and cost of goods sold basis, which were $526 million and $428 million, respectively, for the third quarter. As a reminder, the difference between gross and net revenue is related to our logistics services, which is accounted for on an agent basis, meaning that we only recognize the net profit on logistics services as revenue. Cash and cash equivalents totaled a record $401 million at the end of the third quarter, up $25 million compared with $376 million at the end of the prior quarter. Third quarter cash flows from operating activities totaled $41 million compared with $101 million in the prior quarter. And for those of you tracking capital expenditures and depreciation, capital expenditures were $13.3 million in the third quarter, and depreciation was $9.5 million.

Earlier this month, we announced an agreement to divest the majority stake of SMART Modular Brazil. The 81% divestiture of the commodity module business in Brazil will enable SGH to focus on our strategy of delivering high performance, high availability solutions to our enterprise customers. The proposed purchase price values 100% of SMART Brazil and an initial enterprise value of $205 million, and values are 81% disposition at $166 million. We will receive an upfront cash payment of approximately $138 million, subject to certain customary adjustments and a deferred cash payment of approximately $28 million to be paid 18 months post closing. The transaction is expected to close by the end of calendar year 2023, subject to regulatory approvals and satisfaction of customary closing conditions. The parties have also agreed to a put call feature for the remaining 19%, exercisable during 3 exercise windows between 2027 and 2029. The put call values 100% of the Brazil business at 7.5x its fiscal year net income. For example, if the put call is exercised in 2027, the reference net income would be fiscal 2026 and SGH would receive 19% of such value at the time of and subject to the exercise of the put or call. For reference, the Brazil business revenue totaled $153 million through the first 3 quarters of 2023 and approximately $39 million in the third fiscal quarter. Based on current run rates for the business, at closing, we would expect this divestiture to be immediately accretive to our non-GAAP gross margins by over 200 basis points and also neutral to slightly positive to our non-GAAP earnings per share.

Now let me turn to our fourth quarter fiscal 2023 guidance, which includes Brazil. We expect that revenues for the fourth quarter of 2023 will be approximately $375 million at the midpoint, plus or minus $25 million. Our guidance for the fourth quarter incorporates the following assumptions: For IPS, which is more project-oriented with variability related to the timing of hardware sales, we expect sequential revenues to be down slightly at the midpoint. For memory, which includes SMART Brazil, we expect total revenues to be relatively flat sequentially. And for LED, we expect revenues to be up modestly in the fourth quarter. Our GAAP gross margin for the fourth quarter is expected to be approximately 26% at the midpoint, plus or minus 1%. Non-GAAP gross margin for the fourth quarter is expected to be approximately 28% at the midpoint, plus or minus 1%. Our non-GAAP operating expenses for the fourth quarter are expected to be approximately $71 million, plus or minus $2 million. GAAP diluted earnings per share for the fourth quarter is expected to be approximately $0.02, plus or minus $0.15. And on a non-GAAP basis, excluding share-based compensation expense, intangible asset amortization expense, debt discount and other adjustments, we expect diluted earnings per share will be approximately $0.45, plus or minus $0.15. Our GAAP diluted share count for the fourth quarter is expected to be approximately 54.5 million shares based on our current stock price. Our non-GAAP diluted share count is expected to be approximately 52.5 million shares as it includes the benefit of our convertible note cap costs. Cash capital expenditures for the fourth quarter are expected to be in the range of $15 million to $20 million. our non-GAAP taxes for the fourth quarter are expected to be in the 11% range as we get continued benefit from our U.S. net operating loss carryforwards. However, as we look into fiscal 2024, we will have used the majority of our available U.S. tax attributes and our non-GAAP effective tax rate is expected to increase to the low to mid-20% range. We continue to manage our operations in a prudent manner as we navigate a challenging environment while also continuing to invest in our long-term growth. Please refer to the non-GAAP financial information section and the reconciliation of GAAP to non-GAAP measures and tables in our earnings release for further details. Now let me turn it back over to Mark for a few remarks prior to Q&A.

Mark W. Adams

Thanks, Ken. While we are in the early stages of AI, machine learning and data analytics deployment, we feel SGH is well positioned to grow over the next 3 to 5 years. As such, we continue to transform into a business focused on delivering high availability, high-performance enterprise solutions that help our customers solve for the future. We are not only benefiting from the right secular tailwinds in terms of end markets, but we have an operating model that has performed well throughout the cycle. The combination of our shift towards enterprise solutions and the secular tailwinds, coupled with this discipline on how we operate the company, make us optimistic about the future and our ability to deliver long-term value to our shareholders. With that, operator, we are now ready for Q&A.

Question and Answer Session

Operator

(Operator Instructions) The first question is from the line of Kevin Cassidy with Rosenblatt Securities.

Kevin Edward Cassidy

Maybe there's, of course, a lot of questions around the IPS business right now in the interest in AI. Can you say how your RFQ funnel or maybe opportunity funnel has grown, say, since the beginning of the year? And how are you addressing it? I guess how are you choosing which customers to pursue?

Mark W. Adams

Kevin, this is Mark. Well, needless to say, the level of activity is noticeably on the uptick. And the second part of your question is really an opportunity and a challenge for us because our model is a bit different. We're looking for customer relationships that really want us from a trusted adviser role and all the way from the beginning of our engagement on designing systems and solutions to implementing them and then managing them post deployment, if you will, post installation. And so the criteria around how we do it, just given the level of activity is really what's the best fit in terms of a partnership model with the customer and what role we can play. And so that's how we're looking at it because as you might imagine, if it's just a hardware play, that's a game we don't want to be in. And we're trying to select the type of relationships we engage with to be those that are a little bit more full service-oriented.

Kevin Edward Cassidy

And maybe just to add on to that, what interest are you seeing from various vertical markets? It's more than just cloud? Or are you seeing broad interest?

Mark W. Adams

Yes. That's also interesting because it's really taken off in terms of the vertical kind of expansion, if you will. Certainly, the cloud hyperscalers kind of early to market on some of this. And I'll come back to that in a second because it's just not the big names. There's a class of what I would call Tier 2, Tier 3 cloud providers that are building out data centers for custom type application environments, which is obviously very interesting to us. But I think financial and health care are 2 that are very interesting for us. Oil and gas continues to be one that's very interesting to watch. And I'd say, one, we continue to take a look at is just education and research, while that was kind of the original market segment for HPC to expand over the years, AI is taking out a new definition of what that opportunity looks like in terms of science and research and what have you. And so I know that sounds like a lot, but I mean everyone is taking a look at how they can enable AI and their enterprise. So it has been an interesting year to watch the horizontal expansion of customer interest.

Operator

Our next question is from the line of Tom O'Malley with Barclays.

Thomas James O'Malley

I just wanted to hit on the AI topic, again, obviously, very important in these times. But is it possible you obviously have some exposure to "AI" in the Specialty Memory business and then you have some through Penguin? If you were to assign just a percentage in terms of what you view as AI as a percent of total revenue today, would you be able to size that for us?

Mark W. Adams

It's a tougher one for us to kind of do that. Let me -- in my script, I talked about our deployment of GPUs at scale is kind of roughly over 50,000 GPUs and if you kind of play that out in the industry, you could see that level of scale. It's kind of hard because I've got to like take liberty with definitions when we talk about an AI system, does that include the networking component, the storage component service component? Or is it just the platform. And so we don't really have a good metric, and I just don't want to mislead you. But the index I just gave you of 50,000 NVIDIA GPUs under management. That should be telling that we're looking to continue to grow and invest in that area. But Ken, will take a little bit more.

Ken Rizvi

That's a good question. I think if you look at it, and as Mark mentioned, it's tough to decide for one for one. But if I look at systems that either use a lot of GPUs or systems that in Jack's business, in specialty memory that are going into the data center for these generate AI applications and the like, my guess is that number for our fiscal '23 is probably trending north of $250 million in aggregate could be north of $300 million if we look at it on a blended basis.

Thomas James O'Malley

So you're saying fiscal '23, the AI exposure is greater than $300 million?

Ken Rizvi

I would say greater than $250 million approaching the $300million.

Thomas James O'Malley

And then, Mark, you made the comment that Q2 looking back may have been the bottom for the LED business. Could you talk about what you've seen in terms of changing trends there? Obviously, there were some tough times in China. But could you just remark on what's gotten better?

Mark W. Adams

Sure. I think the primary index for us has been, despite some continued channel burn, the customer design activity, which is a good precursor in index for us, the customer design activity is up and positive. And look, I don't want to celebrate too much from the bottom. But revenue quarter-over-quarter 3, 2 was up 15%-ish and that at a time where there was still more channel burns. So we're not -- it's not like we're building inventory in the channel to achieve revenue growth. We got good revenue growth from the bottom at a time where the channel is still burning some revenues. So I would just say, overall, that coupled with the design activity, it's very favorable. The other thing I would say is this LED market environment is not germane or specific to Cree. As a matter of fact, I would encourage you to benchmark Cree's performance relative to the competitive landscape. Cree's done a great job. And it's just been a tough business, very similar to the memory environment, Tom, but Cree has done a great job navigating some really tough headwinds during the macroeconomic environment they face. And I'm just really impressed with how they've kind of navigated and gotten to this today.

Thomas James O'Malley

Helpful. And then just if I could sneak in one more. Traditionally, you see some very strong seasonality for your IPS business into the November quarter. You commented in your slides for IPS that you just completed the second phase with Meta on the Supercluster. Is there any reason why November seasonality would be any different than what you've seen historically there, just given some of the programs have wound down?

Mark W. Adams

Well, it's kind of a loaded question, but I'm going to try to answer, but it's a tough question being that the timing of these deployments. When you're growing like we've grown, we've got -- we've been able to navigate a lot of this, but timing of some of these deployments, not just the budgeting piece upfront with our customers or even the large-scale customers and how they're allocating budgets. But also just once you get the orders, when those get installed and how they get installed, the supply chain challenges, I mean the supply chain has improved broadly. But if you look at certain key technologies, the networking component of these installations, the GPU availability of the leading-edge GPUs in the industry. I just can't commit that everything is going to be the same as a prior year given all the factors that I just went through, lumpiness, concentration, deployment timing, capital budgets, supply chain. We are long-term believers. We think we're in a really good place, but I want to stop short of calling quarters that far in advance.

Operator

The next question is from the line of Sidney Ho with Deutsche Bank.

Shek Ming Ho

It's great to hear about the demand strength that you guys have seen in AI. So a couple of quick questions here. One, do you see the demand for these AI servers to be crowding out spending on traditional services, meaning it's coming out from a same fixed budget, but that may not be impacting you guys? And the second question I have is if you can give us an update on your Penguin on Demand solutions, that would be great. But what kind of customer appetite are you seeing for consumption model versus like traditional CapEx model, particularly as it relates to HPC and AI market?

Mark W. Adams

Can you just repeat -- restate the first part of the question?

Shek Ming Ho

Yes. The first part is just trying to figure out that everybody is spending a lot more money in AI. And of course, you see a big uptick in demand stream. But do you see that as the -- basically coming from a fixed IT budget that some of the money is being pulled from traditional spending on traditional servers?

Mark W. Adams

I'm not sure that I can really give you a definitive answer there. Here's what I would say. We have seen in fiscal '23, money reallocated when there is commercial success in the deployment and the production side of the house for these types of systems. As far as next year's budget process, and obviously, our fiscal year is off cycle with the budget process of our largest enterprise customers. I think if I had to guess, I would say, yes, probably there'd be some shifting of budgets toward this, but you're talking about different type of customer environments in the UltraScale hyperscale customers, they obviously have a lot of infrastructure they have to build out, not just AI, but just pure cloud resource capabilities. And some of the more newer verticals I think the math adds up. And we got -- let's remember that we're not quite out of the woods on the overall macroeconomic headwinds. And so I'm guessing that there will be money that needs to be shifted appropriately, but it's just such a broad question that I want to just caution my answer a little bit.

And depending on demand piece for us is something that we are kind of shifting slightly our model on demand. We announced this year in fiscal '23, a partnership with Google Cloud as an HP partnership model with them. And so while we do have cycles being used on our Pod Penguin-On-Demand platform, our strategy really is to leverage some of the capabilities in infrastructure with our Google partnership. And in addition to that, leverage our control plane software, the skilled cloud software platform that allows people to run their workloads in a hybrid environment, both on-prem and in the cloud and just really focus on their workload optimization, not necessarily carrying where the workload is being processed.

Shek Ming Ho

Maybe a follow-up question, looking at the Brazil memory divestiture. You talked about EPS being neutral to slightly positive. Can you talk about what's included in that assumption in terms of the use of proceeds from the divestiture? And kind of related to that, now that you have divested the Brazil memory business to focus on enterprise solutions, how are you thinking about other nonenterprise driven businesses? Maybe you can touch on your M&A strategy going forward? That would be great.

Ken Rizvi

So if we look at the accretion dilution, one, we said on the formal portion of the transcript that we would expect the gross margin accretion to be north of 200 basis points. And the transaction based on the current run rates to be kind of neutral to slightly positive to our non-GAAP EPS. That does not include what the net proceeds that we would receive. And so it is just based on the current state and run rate of the Brazil business. If we look at the net proceeds, I think this was your second question. In terms of how we look at and where we will deploy that, we said we really have, I would say, 3 to 4 vectors in terms of how we think about capital allocation. First and foremost, continue to invest in the business. I think Mark highlighted on his formal comments, continued investment in the IPS segment. There's lots of opportunities, as you could imagine with our platform and our ability to grow longer term in that segment. #2, it is to look at M&A and inorganic opportunities similar to Stratus, which we closed here earlier this fiscal year. Those have been great additions to our overall business line. #3 is returning capital back to shareholders. If you looked over the last 12 months or so, we've returned about $58 million in terms of share repurchases. And then also, I would say 4 is looking at reducing our gross debt levels. Our net debt levels are very reasonable, but just looking at potentially reducing gross debt. So in terms of the net proceeds from Brazil, once that transaction closes, we would look at those opportunities.

Operator

Our next question is from the line of Raji Gill with Needham & Co.

Rajvindra S. Gill

A question again on the AI business -- the IPS business. You mentioned it grew 31% year-over-year, excluding Stratus. When we're thinking about the overall market size for this opportunity, is there a way to think about it from kind of a bottoms-up view? How are you kind of valuing the overall market opportunity? Are you kind of valuing it in terms of potential projects based on certain verticals that you're targeting? You mentioned conceptually, some of these verticals you're focusing on financial services, health care. Just any kind of thoughts on long term in terms of how you're identifying and quantifying the size of the market and how that leads into your pipeline of business that you want to try to go after?

Mark W. Adams

Sure. I'll start and I'll let Ken jump in as warranted. Raj, thanks for the question. I think the best way to think about it is that there's a broad market opportunity that you could size up AI. And I would -- one way we look at it is there's kind of those large companies that will deploy the technologies and systems and not really need or want initial outside services and capabilities that we like to provide. And as I've mentioned before, and you can see from a comparison on where our gross margins are relative to our competitors, we're just not going to play that game. That's just not a good game for us to be in. That's not what we're good at. And so I'd like to think of us as a subset of the AI market where we're really playing the role of not just the system designer and developer but the service and trusted adviser consulting type model that we're full service end-to-end. And that's a small subset -- a smaller subset of the bigger pie because there are large companies that are looking to deploy with their in-house capabilities. And that's something that, quite frankly, some of our customers will eventually look for us to do more or they'll build their own capabilities, and we have to kind of continue to bring on new customers that way. And that's the dynamic nature of the business. There's so much activity that way that we're excited about our go-forward business, but we're not the whole AI market, and I want to be transparent. We're going to be disciplined because at the end of the day, I just don't want to take us through a hardware-only business. That's not going to be one we're going to be very profitable nor are we going to be as competitive as we are in the core trusted adviser model.

Rajvindra S. Gill

Got it. I think that's really helpful to try to understand the distinction between your model versus, say, others. Just with respect to the competitive advantage, and again, you mentioned your focus being on more kind of smaller customized solutions. Where do you think kind of Penguin Computing ultimate competitive advantage is relative to the other competitors? And can you talk about who are you competing with head to head? Is that changing at all given kind of the deployment of AI inside companies are you seeing the competitive landscape change? Any thoughts in terms of kind of Penguin Computing advantage? And why are you winning? And if you do lose, what are the reasons why you might lose?

Mark W. Adams

Yes. Let me do my best with that. And then if I don't get to everything, just to kind of re-ask what I missed, but Here's how we look at it internally. If I kind of broke down the pyramid of engagements that we have and that we're talking the type of customers we have, the top tier are the major enterprises looking to deploy AI in the middle tier might be, again, significantly large enterprises, but not necessarily with the spend at the level of a Tier 1 project. And so it might be considered like a middle-tier project. And then, of course, there's kind of, I would say, early stage technology deployment, but really more for test of applications and really kind of getting their feet with. So if you think of 3 different tiers, we look at each of the opportunities and how we play in them. I think the differentiation, and you asked about competitors. The primary competitors we see out there are Dell, HP, Super Micro. And each one of those is a large-scale hardware player with margins substantially lower than us in the business today. Now I'm not here to try to argue whose models per se is better. It's just what we're good at and what we're focused on.

Of course, the 3 companies I mentioned are not the only companies, but they are the examples of what we compete against. And if someone is trying to roll their own internal AI deployments and it's a hardware-only game, well, those folks, that's the line of business, that's the margin structure they're set up to win on. For us, what we're good at, and it comes with 25 years of history of deploying these type of systems. And like we're not talking about one server and a generic software application. Just yesterday, I was over -- we have one of the earliest deployments of liquid immersion technology for a big customer of ours that's in our lab. And our differentiation is we're at front of the technology we're learning about it so we can bring it to market, and we can design it for an environment that we've seen before because over 25 years, there was a lot of organization know-how. So whether it be on the design side, the actual deployment side, when you think of data centers and the complexity of connecting massive amount of compute power with the memory, with the storage, with the networking and making sure the right power infrastructure is in place. This is not easy. It's probably just understated. And with 25 years of history of this type of experience and as Ken mentioned, I mentioned in my script, we've had some of the largest AI deployments in the world to date. And so when you combine all that, our value add, yes, we know how to manufacture these products and systems sure, but our value add is how to design them for a customer environment, how to deploy it and how to manage it.

Rajvindra S. Gill

I appreciate that from that insight. And just last question, Ken, just on the model. You mentioned that the IPS business ex-Stratus grew 31%. So let just kind of do the basic math, it looks like Stratus might have been around like $45 million a quarter. Is that kind of the run rate that you guys are thinking about $45 million a quarter? And other -- how would you describe some of the growth drivers in the Stratus business?

Ken Rizvi

Yes. So I mean that business has been -- just to put it in perspective, last quarter, it did closer to 41% this quarter. You are right. It's closer to that 45%, 46% range. So it does move around a little bit quarter-to-quarter, to some extent, based on some of these hardware deployments relative to Stratus specifically as well. But if we look at it, where has that business been growing? It's been growing primarily at the edge, where they have a leading solution for high availability, fault-tolerant applications. And so what does that mean? And why is that important? If you are trying to run an application and it can't go down for more than 5 minutes a year, you go to Stratus. And those types of applications can be in the oil refining area, oil and gas, it could be in retail, it can be on oil rigs, it can be water treatment plants, but areas where you require a business-critical compute system to run those applications is why customers go to Stratus and that's where that business has been growing over the last 12 months.

Mark W. Adams

And the follow-on, I'd just like to add one more comment, Raji, if it's okay. We've said all along that sure AI and the HPC environment on-premise is where a lot of the early movement is going to be. But over the long term, we believe AI will be more broadly deployed throughout the enterprise and certainly at the edge. And what Ken just described is going to give us an opportunity to extend the AI and the enterprise customer relationships we have.

Operator

Our next question is from the line of Mark Lipacis with Jefferies.

Mark John Lipacis

Mark, for the IPS business, you described it as lumpy. Can you describe the process that you expect to undertake that kind of transitions that to a less lumpy business. And I'm wondering, is it that each time you do a project, it ends up being a kind of a recurring service or management stream over a multiyear period? And then you just kind of stack up on top of one another for each project that you do. And then is that it? Or is it like you become the in-house?

Mark W. Adams

It's a great question. It really reinforces the differentiation, I think we are achieving with our capabilities and the byproduct of that should help us on some of the lumpiness over the long term. But as you said, because of our strategy, our aspiration is to build out these services over time that become more and more kind of recurring and predictable. And remember, we're early on. We're 2 years into this, guys. I mean -- and if you look at the services have grown dramatically, and we're excited about that. But our focus is on if we select the right engagements with our customers, we will be continuing to build that over time. And right now, given that we have been growing so substantially, yes, services are up, but we're only going to be able to do that with continued execution. And that's why when I had been asked earlier on this call, how we differentiate ourselves, well, because of our capabilities and our skill sets, we're going to tend to select customers that value that as opposed to hardware-only transactions that don't give us that stability, they tend to be one-offs. It's going to take us time to build that guys. I mean if you think about where we've come from, I mean, with the divestiture of Brazil, we've got a business that is now a compute story with gross margin percentages dramatically up, operating income performs pretty stellar through a hard difficult cycle. I mean the memory business has been -- this is a better it's been in 15 years in the memory business, generating cash and investing in this new transformation. And we're long-term company builders, guys. We're not trying to run off people on a quarter-by-quarter basis. We are excited about this opportunity. The specialty Memory business really plays well into this enterprise solutions for AI and machine learning and data analytics because the biggest bottleneck right now in high-performance compute AI is the memory connectivity. And I think SMART Modular is a leader. So if you put this all together, we have a pretty bright future. It might not be perfect every quarter because of the lumpiness. But as we design and get our customer relationships more aligned to a full engagement model with solutions and services over the long term, I think we can work out some of the lumpiness and concentration issues that we've had in the past.

Mark John Lipacis

Got you. And I appreciate you're only a couple of years into this. I mean, do you have like the starting base of services and management and would you care to break it out?

Mark W. Adams

I think Ken has talked about it, and I'll let him clarify in terms of the percentage of our top line revenue. Now remember this, I think this -- I better not go, but services are up a multiple -- a big multiple 4 to 5x since we started here 3 years ago.

Ken Rizvi

Yes. I think even if you look over the last 9 months or so, I think services are up about 3 year-to-date, so the first 9 months versus the first 9 months last year, are up about 60% year-over-year. Now within that, there are point in time services like logistics and design and implementation. And then there are the ongoing services that we've been growing over time as well.

Mark W. Adams

Mark, one more comment is we -- you asked the question. I think for us, and this is where I'm super committed in working with the team, the discipline is going to be stay away from the hardware bidding one-offs, do what you're doing, stick to your knitting and do what you're good at. And we can certainly design great systems. We've proven that. But the question is, does the customer value the overall offering, and that's going to be something we continue to focus on as we continue to expand and invest in IPS.

Operator

Our last question is from the line of Brian Chin with Stifel.

Brian Edward Chin

Maybe the backtrack to Brazil. I guess any consideration to reporting Brazil in discontinued ops. But I guess, more importantly, I'm also a little surprised by the limited EPS impact when you strip out Brazil. So basically, Brazil was operating at a slight loss and specialty memory op margins seem to be in the low to mid-teens. But also thinking across the cycle, what were the op margins in Brazil when the business was operating at peak revenue run rates, maybe going back 4 to 5 quarters ago?

Mark W. Adams

Yes. So I think if you look at the Brazil business and where it is today, we tried to outline the first 3 quarters, Brian, and also where the current run rate was as of Q3, which was in that $39 million range. of revenues. So not a surprise in the sense that the margin profile, gross margin profile for that business is lower than where it was, call it, a year ago. But this has always been a business that has had the lowest margins or the lower end of the margin profile for SGH if you looked on a historical basis and well below the corporate average. Even at peak times, it's been well below the corporate average. And where we've done better in terms of both margin profile on the gross margin side and operating margin profile is in the specialty segment where we are targeting enterprise solutions and a higher value-added solution for our customer base. I don't know if that answers the question, but that's where the business is okay.

Brian Edward Chin

Okay. enough. Fair enough. And then maybe kind of to be a bit direct to the point on IPS. I guess, broadly speaking, are you starting to rebuild the funnel there in IPS? And then rather than asking you about a quarter, I'll just go ahead and ask about the full year. So how are you thinking about IPS growth rates in fiscal '24 based on current visibility?

Ken Rizvi

Well, I think -- look, we're seeing a lot of activity right now. Brian, it's difficult for all the factors I mentioned at the front end of the call, meaning the timing of budgets and deployments and how we're working on that. We are long-term bullish. We're not ready to forecast the year at this point because given the timing of our fiscal year and the budgeting process that our customers but the budgets are kind of being worked on. We've got tremendous activity. There are some supply chain issues that we're working through that as it relates to how that would impact future deployment. Obviously, the broader macro trends. I'm not sure we're out of the woods on the macroeconomic environment. So we're trying to be careful just to not overstate where things are. We're excited about our long term, obviously, and we'll be kind of give you more color as we pay closer to our customers' budgeting process and future Q1 and Q2 activity.

Mark W. Adams

Brian, I realized you asked about the disops that we'll take that -- because the deal was announced here in our Q4, we will evaluate that as we head through the end of Q4, whether or not the Brazil business will be available for -- that help is available for sale disops and we'll evaluate that as exit Q4.

Brian Edward Chin

Okay. And then maybe just lastly, to hold in back on the discussion around AI, when you think about the duration of your sales cycle, I guess your sales design and implementation cycle as well as those possible component supply constraints with something like the middle of next year be reasonable to maybe expect some of this upswing in AI engagement to translate into revenue?

Ken Rizvi

I think that's fair as we sit here today, and I'm not trying to overhedge, Brian, but I think that's fair based on what we're seeing in terms of some of the core technology lead times and manufacturing ramps and some of the newer launches in the industry. I'm sure you follow the industry as well as we do, and there's been some technology announcements and early customer shipping in some areas and some of the market leaders in GPUs, for example, I would anticipate in the middle of next year, we start to see a little bit more favorable conditions for availability of supply.

Operator

That concludes the question-and-answer session. I will now turn the call over to Mark Adams for any further remarks.

Mark W. Adams

Well, thank you all for joining us today. We continue to be very positive about the business. If you look back over the last 3 years, we feel like we're way ahead of where the transformation might have been as we looked at starting in August of '20. A lot of good things going on here at the company. We're excited for both ourselves and our Brazil team on the path that we have for each other going in different directions, but supporting each other through the process. And we like the secular terrains we have in AI, machine learning and data analytics to drive our future. And we appreciate you attending today's call. Thank you.

Operator

That concludes the SGH Third Quarter Fiscal 2023 Earnings Call. Thank you for your participation. You may now disconnect your lines.