Let's go ahead and get started. Um good morning, everyone. Welcome to day one of Bank of America's Global Tech Conference. I'm Wamsi Mohan. I cover IT hardware and supply chain here for the bank. I appreciate all of you being here today. I'm especially delighted to welcome Dell again to our conference. This year, we have Arthur Lewis, who's President of Infrastructure Solutions Group, ISG, which covers all of servers, storage, networking services so. Amazing portfolio, which has been doing incredibly well. So we're super excited to have Arthur over here, especially at a time when there's so much talk around and excitement around AI. And Dell has definitely been in the lead over here with some of its products, recent announcements at Dell Tech World. Super excited to have you over here, Arthur.
Thank you, Wamsi. Great to be here. Good to see everybody. Good morning.
Well Fantastic well. I know we've got just only 30 minutes over here, and there's so much to cover. But maybe to kick it off, Arthur, can you just talk about some of the things that, as you just closed out last quarter, what you saw, um some of the puts and takes from the quarter first, and then we can get into other topics?
Yeah. So, let me um you know break down the quarter for you, Wamsi. From an overall ISG perspective, $10.3 billion in revenue, which was 12% growth year over year. That represented our fifth consecutive quarter of double-digit revenue growth, which was pretty exciting. Operating margins came in at about $1 billion, growing 36%, or three times faster than revenue. Um and that' s four consecutive quarters of double-digit growth. Right so great top-line and bottom-line growth for Q1. If I break it down into the three main components there, you know clearly a blockbuster AI quarter. Three things I would call out for you guys here is, one, you know record orders bookings of $12.1 billion. To put that number in perspective, that is greater than everything that we shipped last year, just in our first fiscal quarter. Number two, you know we have a backlog of $14.4 billion, which is another record for us.
Importantly, the third point is that we take a look at the next five-quarter pipeline as being indicative of demand. Even though we took in $12 billion of orders, our next five-quarter pipeline grew sequentially and continues to be multiples of our backlog, even at $14.4 billion. So again, very indicative of the demand that's out there around what is a very revolutionary technology. The server business also continued to grow six consecutive quarters on a demand basis, five consecutive quarters on a P&L basis. The overall server networking business all in grew 16%. As we said on the call, Wamsi, there was a little bit of slowdown that coincided with April 2nd. I think something happened on that day. We saw kind of a slowdown from week 10- 12. We're monitoring that very closely. The storage portfolio was great. It grew 6%.
We expanded profitability. It was a third consecutive quarter. This pushed the third consecutive quarter of growth. This push that we've had towards Dell IP is really working out. You can see the proof points with PowerStore in particular, growing five consecutive quarters, double digits. Our growth rate in Q1 was the highest growth that we've seen in PowerStore in the last 12 quarters. Importantly, something that I pay a lot of attention to is that um PowerStore drove 15% new buyers, uh I'm sorry buyer growth. 45% of those buyers were new to PowerStore, which is great for share of wallet expansion with existing customers. But 18% of those buyers were new to Dell. So it is also a great acquiring engine. Um so you know we had PowerProtect Data Domain, the PowerProtect portfolio, excuse me, growing double digits as well.
This push into Dell IP storage is really doing well for us. We have the headwind with VxRail. The Dell IP portion actually grew faster than P&L average, which is something that we like to see, especially when the market is growing in the 3%-4% range.
No, that's amazing. Thank you for that recap. Clearly, a very strong print over here. As you think about you know just the server revenues that you just accomplished, your backlog and pipeline, you just noted extremely strong, guiding to very strong revenues here in the near term. Um help us think through the revenue trajectory as we go forward around AI and also the incrementals from a profitability standpoint that investors should think about as you're getting this revenue stream.
Yeah. Let me touch on your first point first, or your last point first, and your first point second. You know from a profitability perspective, we kind of take a look at the three independent businesses, whether it's the AI business, the server business, traditional server business, the storage business. And each of these businesses, sequentially, we see expansion of gross profit, of operating margin dollars and rate. It kind of comes down to a mix at the end of the day. When you have a higher mix of the AI portfolio, you might see a lower rate. Again, I kind of emphasize Q1, we saw operating margins growing 36%. If you take a look at the guide for Q2, you have revenue sequentially growing $5 billion, but operating margin growing $500 million. Right so.
There's clearly margin accretion when it comes to the AI portfolio, but we're also expanding margins in traditional server and storage. On the volume piece, we guided at the beginning of the year 15%+ on AI with $2 billion of shipments and $14 billion of backlog. I think it's safe to say that we're on the plus side of the $15 billion. A lot of people say, "Well hey, why didn't you just call a new number?" The reality is it's really hard to do because these are very technical engagements that we have with customers because a lot of this revenue is being generated by these large tier two CSPs. I think a little bit of color on how the relationship works kind of helps you guys understand the spikiness in the business.
When we engage with these customers, they're spending a lot of money, and they have very specific outcomes that they're looking for. When we engage with them, we will do anywhere from three to five designs for them. These designs will be very different from each other. The customer will have to, and we go back and forth for quite a while to hone in on the designs, on each of these designs.
The customer will then wait until the very last minute to make a decision to say, "Hey, I'm ready to make a decision where I'm spending billions of dollars of CapEx that I need to know that I'm making exactly the right decision on exactly the right technology." That has been a key differentiator for us in addition to the ecosystem, in addition to our services, in addition to our captive finance with Dell Financial Services. It is this technology back and forth that makes it hard to predict when is the order going to land, and therefore, what is the delivery schedule going to look like, and then when is it going to materialize into the P&L.
I think as we go through our second fiscal quarter, we'll learn a lot more about the deals that we're working on and have a better look at what the second half looks like from an AI perspective. Again, clearly, we're on the plus side. We recognize it. We also understand the frustration around, "Hey, if you know a number, call it." We have a number, but it's really going to depend on customer choice and technology decisions that they make. When we put a number out there, we want to make sure that we're going to hit it. That's why we kind of stick to the 15%+ for now.
Yeah. I know it was called out on the call that the plus was a capital PLUS. All right.
Underscore, bolded, italicized. Yeah.
Yes. So all right. Well maybe just to think through product transition, because you noted one point, which is you're doing so many different designs, and it's kind of quite complicated. You got to get all these components, supply chain. You guys have been terrific at managing supply chain. Maybe if you could just spend as 30 seconds talking about how you handle product transitions in some way. I think the industry has actually struggled a lot going from Hopper to Blackwell. You guys were super early in delivering that. As you're looking from now, maybe GB200- 300. How do you think about the cadence of this? It feels like maybe this transition should be easier, but would love to get your high-level thoughts on that.
Look, I do not really want to jinx myself, but when you are engaged at a very technical level with customers and you get an indication of what it is they are going to buy, you have an indication of what you need to buy. When you are kind of speculating as to what might happen, you might get yourself kind of like in a sideways sort of type shift. This is, again, the pace of technology here is like, I have been in the industry for a little bit over 30 years. I have never seen anything like this, Wamsi. You guys have super chips coming out now every six months. You have the software that goes along with it, the networking advancements. You have gone from 200 to 400 to 800 to 1.6, soon to 3.2. I mean, the amount of innovation here is just staggering.
We talked about NVL72 has got 1.2 million parts. I kind of was joking with Paul, envisioning like, "Hey, why don't we just lay out all the parts on the floor?" Who wants to judge? Who wants to come and partake in putting this puzzle together? I mean, these are not easy things to engineer. People who think that there's a reference design out there that you can just go copy and get a bunch of people and kind of put it together, that's a fallacy. Putting these systems together is extremely complicated. Um you know deploying them, connecting them, getting them up and running, turning them over into production, it's very complicated. We have differentiated ourselves. We were the first to deploy Hopper. We deployed 100,000 GPUs in six weeks last year. I have not heard of anybody being able to do that. We deployed the first NVL72 in November.
A lot of our competitors are still having trouble deploying NVL72. We are now in a position where we turn over systems to customers in production within 24 hours. We just keep getting better and better as we hone our skills. Customers need it because the pace of technology, I mean, typically, you are on sort of a TikTok, like a major thing every two years, a midlife kicker in between. You have major transitions happening now every six months. You are going from last year, we were at 62 GPUs per rack. Now we are at 256, heading to 576. It is just like within 12 months apart. And it's.
Yeah, the pace of product introduction. I tuned in into DTW.
Yeah. So, remember Jeff, he has a new acronym, UFB?
Yes.
Yeah. That's what this is.
Yes. Yes.
I'm comfortable with the, yeah.
Yes. It's quite amazing. Maybe a little bit on Dell's differentiation here in handling the supply chain. We heard from one of your peers, and they had a material guide down to their numbers because of issues associated with tariffs. Um how did you guys navigate? I know you absorbed some of the impact. I know you called out on the call and said, "We're not making any price changes at the moment." How are you able to do that? What is the secret sauce here that you are able to deliver such solid results in managing the supply chain?
Look, I mean, kind of like a four-part unsatisfactory sort of answer. Number one, I mean, Dell has been managing geopolitical issues for the last 40 years. And there have been any number of issues that we've shown a certain agility and adaptability around. Specifically on tariffs, we've been working in a situation since Trump v1 in 2016. We've become really good at it. We've become really agile about it. Everything that we know about tariffs is embedded in our guide. I know it surprises a lot of people that everything that we know about tariffs is embedded in our guide, and we talk about Q2 being deflationary, but that's just kind of like what we do.
Yeah. No, it's actually really impressive, especially when you consider how other companies are operating. And I know ever since Michael started the company, supply chain and negative cash conversion cycle, these have been hallmarks for you guys. Um pretty amazing. Maybe back to AI servers a little bit. You guys, you just noted, Arthur, that you had $5 billion incremental on revenues. You've got about $500,000 million incremental sequentially on gross profit dollars, $500,000 million sequentially on operating profit dollars. Maybe just help us think through how that flow through is as high as it is on the one hand, and what other components are contributing towards that sequential increase outside of AI servers.
Well I'm going to include AI servers just to kind of tell you the full story, but I think there's four components to think about. One is storage is incredibly important in the P&L, and we continue to push more Dell IP. That's incredibly important because we talk about the fact that we want to push more Dell IP because obviously we make more money on it. I think the why is incredibly important here. The reason is there's been a significant trend over the last couple of years back to disaggregated infrastructure. You start thinking about, why is the world moving back to disaggregated infrastructure? The reason basically comes down to the fact that more and more customers are moving into what we refer to as a multi-hypervisor environment. Why are customers doing that?
They're doing that, number one, because they want to prevent vendor lock-in. They want to have flexibility in terms of the cloud operating system that they use. More practically, though, they're also looking at the new workloads that are coming online, some of it due to AI, that are more container and bare metal-based. They need to be thinking about that. There has been an industry shift over the last two years away from perpetual licensing and to subscription-based pricing. Subscription-based pricing. That subscription-based pricing is based on CPU cores right? In a traditional HCI, your cluster might run at 30%-40% CPU utilization. That's completely ineffective in a situation where you're paying per core. You need the ability to scale compute and storage independently. More and more customers are now pushing back towards disaggregated as the way to go.
It gives them more flexibility, and it's a lot more efficient from a cost perspective. You see that in PowerStore. We just talked about the numbers there. You see that in the private cloud and Dell Automation Platform announcements that we had at Dell Technologies World. We leveraged this technology trend. We have the data trend for the unstructured portfolio, the security trend for the cyber resilience portfolio. We talked about the fact that PowerProtect grew triple digits. We like that storage trajectory, and we expand margins as a result. A big contributor there, number one, is storage. On this traditional server portfolio, we continue to maintain very strong price discipline, even with the mix where we have, which is probably more indexed to the enterprise than the market itself. We have industry-leading margins there.
In our guide, you see that we have margins expanding there. We even have margins expanding as we go up throughout the year in AI. We couple those three business objectives with very strong cost controls from an OpEx perspective. That is how you get to your $500,000 million of incremental OpEx.
Okay. That's helpful. Would you call out anything that was one-time in nature at all in the quarter? Okay. That's great. As you look at sort of the pipeline of opportunity here for AI and think about the customers, different customers, where that opportunity resides, how would you characterize this between maybe tier two CSPs, enterprise and sovereign, and line of sight into that?
Yeah. So you know from a market segment perspective, we basically look at it as three separate customer segments. Maybe a little bit more when we get to enterprise, but I'll get there in a second. The one that we talk a lot about is sort of the tier two CSPs. They're the dominant portion of the portfolio right now. There's kind of two flavors of these CSPs. You got a couple of companies out there that are looking to change the world, and they're driving to artificial general intelligence or artificial superintelligence. Then you got the rest of the Neoclouds that are really out there vying for the GPU as a service business.
And that is the lion's share of the revenue, and it's the lion's share of the pipeline because there's an unabated race to a pot of gold at the end of the rainbow that these guys are chasing. Um the second customer set is sovereign. We've talked a lot about sovereign. We talked now that, hey, this is an area that's starting to ripen. You're starting to see sovereigns materialize in the pipeline. A lot of people talk about the announcements in the Middle East. Yeah, those are interesting.
When are you booking your flight there?
I was there earlier this year. A lot of people talk about the Middle East, but there are sovereign opportunities all over the world in the United States, U.K., France, Germany, in Northern Europe, uh in Asia, Japan, Malaysia, Singapore, Korea. Um there are sovereign opportunities all over the place. Last Thursday, we announced one with the Department of Energy in building out their new flagship NERSC- 10, which was an incredible win because traditionally, we have not paid a lot of attention to the HPC space. Now, very interestingly, HPC is moving towards AI as you get model and simulation to go along with the traditional machine learning, deep learning, and generative nature of artificial intelligence, which is placed naturally to our strength. That was a pretty nice win there. Jensen was there with the [audio distortion] Department of Energy Secretary Chris Wright.
That was an incredibly proud moment for us in winning that business. You will see more of those opportunities ripen, get into the pipeline, and start to book orders. There is the enterprise where we talked about the fact that we now have 3,000 customers. Um you know what I really like about the enterprise is it continues to grow, but the maturity level is still very nascent compared to the other two segments. The other, the CSP, the ball game is underway, right and it is early innings. The sovereign, it is probably the top of the first inning. In the enterprise, the game has not even started. Players are still warming up on the pitch. You were pulling into the parking lot, and we are getting ready for the game.
What I really like about the enterprise and what's really exciting is that you know we are now helping customers define their future data strategy. This is not something that we have traditionally done as an infrastructure provider. Um companies really do not understand artificial intelligence, how to deploy it, how to use it, let alone how it is going to impact their data center going forward. We sit down with customers at a very early stage, and we talk about very strategic things like what use case and ROI are you going after? How do you think about model selection? Jeez, I got this Llama thing. I got this Cohere thing. I got this Mistral thing. I got this Gemini thing. How do I match up model against use cases? They get to the really hard question around, okay, I understand my use case or a couple of use cases.
I understand my model strategy, but now my data is tied to hundreds of applications against thousands of databases. These databases are very siloed. They do not talk to each other. The data is not clean. It has not been prepped. It is not tagged. There is no easy way for me to ingest it into the AI. How do I do all of this, Dell? Because if data is the fuel that feeds AI, and you are building a 400-horsepower engine, but you are going to feed it non-premium-grade fuel, you are not going to get the optimization and the value out of the engine itself. We are now at a very strategic position with a lot of these customers where we are helping them not just define their current architecture, but the architecture, their infrastructure for their future data center. That is a position that we have not been in.
I love the fact that we're leading with innovation here. We're not fast-following anybody. If you take a look at our compute portfolio, our storage portfolio from private cloud to the AI data platform to the cyber resilience, I mean, we're innovating like it's nobody's business. DTW last week was an incredibly proud moment. I have the privilege of speaking to the innovation that the team is driving, 40 major you know product announcements across everything. The biggest question, I go, "Hey, you didn't talk about this. You didn't talk about that." I'm like, "I have 15 minutes, man. I got some stuff you want me to talk about." That's how much innovation we're driving. We're in the driver's seat, Will, not in the driver, but we're right front row into one of the biggest you know technology transitions of humankind. It's just incredibly exciting.
Arthur, just on, I think when you started talking about enterprise, you mentioned ROI and it's important. So can you give maybe just some examples that can help us think through where these efforts are being deployed at enterprises and what kind of ROI targets and timeframe on those, any perspective that you could share on that?
Yeah. Let's go through a couple. Content is always an interesting one because you know generative AI in particular is very good at document summarization. Right so if you, like us, have thousands of people that are generating tens of thousands of pieces of content, there's a very clear ROI to say, "Hey, you can build a content generation engine and really help streamline and accelerate and drive consistency and perfection in your content creation." So that's been a pretty easy-to-understand use case for a lot of customers. A coding assistant has been another really good use case. I mean, we're using that internally. We've targeted, I think, conservatively 30% um productivity gains using coding assistants. Then I got a choice to make. I can go deliver 30% more stuff or deliver the same stuff with 30% less OpEx. Now, it's a choice.
Do I want to be more of an innovator, or how do I kind of think about the business? Again, that's a conservative number. Customer service has been a great one because at Dell, we've been driving what we call the next best action, which we use generative AI to help the call center reps understand what the problem is that the customer has to get to an action that's going to get the situation resolved significantly faster than in the old world. Obviously, your sales chat assistant would be sort of the top four use cases. There are other use cases around supply chain, finance, things like digital twins. Any number of use cases, but the content creation, code assistant, sales chat, customer service, I think are the top four that customers are playing around with right now.
Okay. That's super helpful. Um one of the things that, since you have such a large portfolio, um you know can you share some color around just storage attached in AI? I know you've spoken about Project Lightning. Where do you see deployments of that on a go-forward basis? How should people generally think about the attach of storage and other services when it comes to the ramp of AI servers?
Yeah. So, we already see the attach of network and storage to the compute, right? Because the AI, I would think of it as a system more than anything else. Last year, artificial intelligence was a place where you had these one-shot inferencing models. With these one-shot inferencing models, the model could sit on the HBM. And so customers were really focused more on the server node itself. As you move now to TRIA Thought Logic and some of these long-thinking reasoning models, the model still sits on the HBM, but it's not going to have all of the information to respond to all the queries. It needs to go to fetch information from the network attached storage. Now you need very quick connection.
You need fast scalable storage that's got a very quick connection to the compute to be able to generate the tokens to solve the problems for AI. As enterprises move and adopt more of these reasoning and long-thinking models, it's not just a compute game anymore. It's a compute plus the network plus the storage. Our value proposition, I think, is very strong here because in this system, we are unique in that we are the only entity in the world that can actually engineer, optimize, and fine-tune the compute, the network, and the storage under one roof. Essentially, we become the integrator of the system versus the customer saying, "Well, I'm going to buy the compute from vendor A, then the network from vendor B, the storage from vendor C.
Now, I got to put this puzzle together and make it work, r ight? We have not only put the infrastructure together, but we are building on very salient components. We talked about the parallel file system, which will be the fastest parallel file system in the world. We believe it will be twice as fast as our nearest competitor, allowing for 64% greater data access, which is really what you want in a parallel file system. For customers that have the very high-performant file in that tier zero space, Lightning is going to be perfect for them.
We're also talked about the partnership with NVIDIA and Dynamo in creating this key-value caching layer so that when you kind of have to go back and fetch from your network attached storage, you can do so at very accelerated speeds because these long-thinking models, and this was kind of like the deep sea conversation we were having before, it was really more about the reasoning model because the compute needed to respond to a reasoning model is about 100 times more than in this one-shot inferencing models. In fact, Jensen said in earnings last week, it's 1,000 times more. I haven't seen that math, but I know it's at least 100 times more. If he's saying 1,000, then maybe we'll go with his number, 1,000 times more. You have the underlying PowerScale, PowerStore.
It's the breadth of the portfolio that's coming together, whether it's private cloud, whether it's the unstructured portion of the portfolio, or the cyber resilience portion of the portfolio, we kind of got you covered from an enterprise perspective.
Okay. Can you talk about, for let's say, if you sold servers for $100, how much of what percenr or how many dollars of attach could you get on any additional parts of the portfolio? And how should we think about the margin structure of that as well?
Yeah. I don't know that I have a good kind of, it's too early to say, "Hey, for every dollar of server revenue, you should get X," because not every dollar of server revenue is the same. See, I'll give you an example. I'm an enterprise customer. I might have a use case that requires an eight-way server node because that's the performance I need. That's going to cost one thing. It might be that I need a PCIe form factor that I'm going to scale up over time. That's going to cost something else. Both are going to require the same storage, but they're different costed compute nodes. The ratios will be off.
Yeah right so I'm still trying to figure out if there's a good attach metric, but we're still trying to kind of figure out what we want to do is attach storage and network to every opportunity. That's the attach metric I'm looking for. Then we can get into, am I attaching sort of the right amount of storage to it? Like I said, we see it today. You know where we are selling the compute, we have a really good opportunity to sell the storage, and we're doing that. The margin profile should be within the profile of what we're talking about when we sell Dell IP storage.
No, that's great. Those are very good margins for those who do not know. When you think about, as you have more and more of these such contracts and attach and services, especially, some of that portion will get deferred over time. One of the interesting things that is happening, I think, is you have this dynamic on your deferreds where obviously you are adding to the deferreds with all the AI servers, but there is also an element of deferred that is coming off your balance sheet. It is not fully apparent to investors what is going on. Maybe if you could help us think through, when does the materiality of that start to kick in from your perspective?
I'm going to defer to Paul on the materiality of it, but this is a very nascent business. We shipped a little bit close to $10 billion last year. We know we're going to ship over $15 billion this year. When we book these deals, there's a percent of the services that we do that we follow generally accepted accounting principles, and we defer a portion of that consistent with the term of the contract. That stuff moves off to the balance sheet and then over time comes back to us. It's still very nascent. So I don't think we've reached all the level of materiality, I would say, yet, just given the newness of the business. But again, at minimum, we're going to do $25 billion sort of in the first two years. That's an incredibly fast-ramping business.
I can't imagine we're not going to get to materiality quicker than normal.
Okay. All right. That's really helpful. Um maybe just, I know we've got very little time left and we still have so many pieces of the business that we haven't even touched on, but maybe just to very quickly touch on industry standard servers and sort of the cycle replacement cycle dynamic that's going on over there.
Yeah. Like I said, we've had six consecutive quarters of demand growth, five consecutive quarters of P&L growth. A lot of that has been driven by this consolidation. There's clearly a refresh that's going on, but then there's a huge opportunity to consolidate and refresh. That's kind of like what we see for the year. We did see right around April 2nd, demand slowed for two or three weeks. Again, the world got turned upside down where we thought one thing on tariffs and then the next day it was like, "Wow," thought something completely different. I think a lot of companies were like, "Hey, what does this mean from a macro perspective? What does this mean to my business? How should I be thinking about spend?" We definitely kind of saw a slowdown.
Internally, we brought down our internal model forecast by about a point. We were thinking the market was going to grow 5-7%. We thought now it is more in the 4-5% to be a little more conservative, but it is still a growing market. That is sort of factored into our guide. What is really cool is that we have a pretty big install base, and 75% of that install base is sitting on 14th generation and older servers. We just launched our 17th generation server. So I mean, depending on the workload, you can consolidate, say, one server to four to seven servers, kind of depending on the workload. That is a pretty significant consolidation for enterprises that are looking to optimize for space and power.
No, that's awesome. I know we're out of time. Maybe, Arthur, just to close out, what do you think investors should most be focused on about Dell and any parting thoughts from you?
Yeah. I think the most important, what excites me about what I do. And I love technology and I love artificial intelligence. I think this is a technology that's going to revolutionize the world for the better. I think this is something that's going to drive human progress in so many ways that we can't envision. We have a front row seat to helping customers navigate this very challenging transition. We're helping the largest of the CSPs. We're helping sovereigns. We're helping the enterprise. We're taking all of the learnings across all the different customer segments and sharing those learnings. Everything that we do in the largest of the CSPs, we take that and we help customers of all sizes and shapes to kind of really understand the technology.
Being there and being that trusted advisor for the enterprise through this technology transition and leading the way, not being a fast follower, I think is something that we do not get enough credit for.
Awesome. Thank you so much, Arthur. That was really insightful. Really appreciate your time.
Thanks, Wamsi.
Thank you.
Thanks.