Infrastructure Strategies: Measuring Profitability in the Age of AI

Speaker A: What we're seeking to understand here is, you know, how well are vendors managing to a capture the opportunity that exists, but also are they doing so at a sacrifice to other parts of their business or profitability and you know, what's the long term viability of these different business strategies? So I think that's really what we want to better understand and bring, bring in a metric that compares the growth potential as well as profit, um, strategy as kind of one of the facets of how we evaluate the OEMs.

Speaker B: Welcome to TBR Talks, decoding strategies and ecosystems of the globe's top tech firms where we talk business model disruption in the broad technology ecosystem, from management consultancies to systems integrators, hyperscalers to independent software vendors, telecom operators to network and infrastructure vendors and chip manufacturers to value added resellers. We'll be answering some of the key intelligence questions we've heard from executives and business unit leaders among the leading professional IT services and telecom vendors. I'm, um, Patrick Heffernan, principal analyst and today we'll be talking with Angela Lambert, principal analyst for TBR's IT Infrastructure practice and, and Ben Carboneau, senior analyst for TBR's IT Infrastructure Practice, about how TBR is approaching our new AI research in the infrastructure sector and our new ramp metric. Angela and Ben, welcome back to TBR Talks. Very exciting to have you back here. And we're talking about all of the research that we're doing now at TBR around the disruption that artificial intelligence is causing to the companies that we cover. So forget about the whole broad disruption that AI is doing to all of society and all of technology in the world. Let's just talk about the companies that we look at, the companies you're looking at in your practice, Angela, and companies you look at Ben and what their adoption of AI is doing to them. So I know in the services space we had a certain kind of way of looking at it, but tell us a little bit about what's been happening in the hardware, the infrastructure space over the last couple of years that has led us to where we are now. Angela?

Speaker A: Sure, yeah, I'll kick off with that. So I think when we broadly look at the IT infrastructure market and what has changed over the last couple of years, there's just been such an explosion of opportunity to sell AI infrastructure, which on the surface sounds wonderful, but there are some implications to kind of diving in and serving a new hyperscale and neo cloud market that's a little bit different than the enterprise market that uh, the infrastructure OEMs are very ingrained in and I think one of the major ways you see the impact from a business model point of view is profitability. So while there is vast opportunity for revenue growth, in short, what we see is that there's also some real concerns about what chasing that opportunity could do to the business as a whole in terms of profit, in terms of the business model overall. So really we're looking at today, how are the business strategies changing within the individual vendors and what decisions are they making based on those dynamics?

Speaker B: Just to set the stage a little bit, this some of the specific vendors that we're talking about, like when you

Speaker A: say are some of the major ones in here, would be Dell Technologies, Super Micro, HPE as well as Lenovo.

Speaker B: Okay, excellent. And it's interesting. So the opportunity is massive in terms of the sort of the market that they have in front of them, but it may not be as profitable as what they're used to. And that could be a huge challenge for them short term and long term.

Speaker A: Exactly. Right. So that's where we're seeing some diverging OEM strategies come in. Right. So on one end of the spectrum you have, you know, dive headfirst into this opportunity of serving hyperscalers and neo clouds. On the other end of the spectrum you have a strategy of doubling down in enterprise or also in the sovereign, you know, opportunities. Well, where the customer needs are obviously very different. Right. There's the enterprise who needs more of an end to end solution. They need more services and they prefer kind of a more all in one solution versus what, what we see in terms of serving the large, huge, you know, billion dollar orders that are on a massive scale.

Speaker B: Right.

Speaker A: But really a very different service profile.

Speaker B: Right. And then Ben, what are some of the trends you're seeing across the companies you cover?

Speaker C: Sure. So covering Dell, hp, Lenovo, Supermicro. I think exactly what Angela's saying is that divergence and strategy has become really apparent, especially between maybe Dell Technologies, who I see as the OEM um, that we cover, that's most willing to take on those really large scale, lower margin deals with NeoClouds versus HPE, who originally did take on more service provider AI systems deals, but has increasingly shifted to being more selective targeting enterprises and sovereigns. I think a lot of that has to do with the company's acquisition of Juniper and their networking focus, which really aligns well with the uh, sovereign opportunity and more distributed AI inference that the enterprises will be doing. But what we're seeing there is while Delt's taking A lot of gross margin dilution in those really large scale deals. HP has maintained, uh, more durable gross margin profile. However, something that we predicted happening and has actually happened more quickly than we had anticipated is the effects of scaling on operating margin. There's always our idea at TBR that if Dell could grow these low margin deals at a rate fast enough, it would wind up offsetting gross margin dilution right from the, from the perspective of their operating expenses not growing as quickly as revenue growth. And we're starting to see that now. And that's really what our metric here, revenue adjusted margin productivity or ramp, is looking at. It's not just looking at whether an OEM's capturing that AI driven infrastructure revenue, but also how effectively the OEMs are taking in that increased demand, um, and that and that opportunity and then either turning it into improved operational leverage and a stronger margin profile or whether they're kind of in the middle. Whereas where I would see maybe more of a super micro or where they're almost at a tipping point of, of growth and having that offset the gross margin dilution or whether they're kind of going down the road that, that HP is going for and kind of staying true to the more traditional customer set, those enterprises that they're used to serving.

Speaker B: And that metric, the uh, say it again, I know it's ramp, but what's

Speaker C: the revenue adjusted margin productivity.

Speaker B: So revenue adjusted margin productivity is a metric that is useful in an AI age because why, like sort of, why would you not have been looking at that five years ago? What is it that's making that metric useful to look at now as a way of reflecting on or looking at the uh, the disruption that AI is causing these companies in particular.

Speaker C: I think the big change has really been NEO clouds coming onto the scene. So an infrastructure as a service provider, you could think of them as a kind ah, of a tier 2 cloud service provider. Smaller than the hyperscalers. Typically a lot of cloud service providers would have gone to ODMs, not um, this OEM industry set, but, but ODMs to uh, serve them and, and deliver those low cost, high volume servers. But now we're seeing this group of NEO clouds go to OEMs to serve them, which has just completely kind of upended what was the traditional OEM infrastructure market.

Speaker B: And so Angela, the ramp is a metric, but what sort of, what are we looking for? What's a good ramp and what's a bad ramp? I'm guessing up and down is the simple way to think of it, but it's gotta be more than that, right?

Speaker A: I think for me the key thing is revenue adjusted. Right. So if we take the most recent quarterly results, for example, you know, Dell's infrastructure business grew by almost 200% um, compared to a year ago, which is incredible. But does that alone, does that say enough about what's going on within the market and the business model? So what we're seeking to understand here is how well are vendors managing to a capture the opportunity that exists, but also are they doing so at a sacrifice to other parts of their business or profitability and what's the long term viability of these different business strategies? So I think that's really what we want to better understand and bring in a metric that compares the growth potential as well as profit, um, strategy as kind of one of the facets of how we evaluate the OEMs.

Speaker B: And then you have to pick a starting point. So how did you come up with the starting point for where you would start looking at the data and devising this metric ramp? And so what was your starting point and why?

Speaker A: I think our starting point, one of our starting points was really sitting together and saying, wow, these companies are harder to compare than they've ever been before.

Speaker B: Huh?

Speaker A: Their, uh, personalities are diverging more so than ever because of the, the AI infrastructure shift in the market. And we would sometimes when you put some of the different metrics side by side, I've looked at for 15 years, and like, well, it's just the compares are a little different and you have to add a lot of nuance when you're describing each of these things. So I think a lot of what Ben and I have been talking about is how can we, I don't know if it's really making a level playing field, but how can we kind of assess the strategic choices that each of these companies are making and who's succeeding within these different customer segments, such as the enterprise space sovereigns, the neoclouds, the hyperscaler space. That's really what we want to look at is kind of how are they targeting those customer segments and how well are they executing on their relatively unique strategies.

Speaker B: Right. And I love that you use the word personality because I think there are times when at tbr, we actually do kind of assign personalities to companies, whether we like it or not. I mean, we try and be very analytical and, and thoughtful about it. But you also can't help but sort of personalize the companies that you're covering and the strategic choices that they're making. So when you think about the four companies you mentioned and you think about where they're current ramp is, do you see, you're saying the strategies are diverging. Do you see a real strong divergence coming in their performance with respect to ramp in the next few years?

Speaker C: Definitely, um, for sure. So I did some back of the envelope calculations where we'll be putting out some ramp reports on the companies we've mentioned pretty soon here. But just looking at the calculations, they kind of aligned with exactly what I would have expected, which is Dell, Dell's ramp performing very well, obviously revenue growing very quickly as Angela said earlier, and then also having reached that tipping point where that scaling is completely offsetting gross margin dilution and resulting in operating margin expansion. So we're seeing Dell performing really well. And then on the other, I guess maybe on the other end of that spectrum you see a company like Supermicro who's also growing very quickly but who has a much different kind of operating expense profile. I think that's due to the, the way that we're seeing these OEMs shift in their target customer base where Super Micro might have been more into low margin, high volume, box moving if you want to call it that. What we're seeing is maybe Dell shift a little more in that direction right now just because that's where the opportunity is. And as a result Supermicro hasn't, hasn't realized those same kind of scaling benefits just because of where they were and where their, their OPEX structure was previously. I think going further down the line, something that I'm really interested to see and I think it'll be something that we see maybe in the uh, in five or ten years when we're looking at these ramp scores is how well does Dell's strategy age and how, and how durable is that, is that strategy? Obviously right now with revenue growing so quickly, the effects of scaling on operating margin are really noticeable. I think we get to a point where the initial ah, infrastructure build out is done and maybe those really large scale NEO cloud deals start to slow uh, down And I think that's when it'll be really interesting to see how they're kind of a more aggressive strategy. Targeting, targeting scale balances against a company's like HP's strategy of maintaining margin. I think maybe the, the last thing I would say on that is it's important to note that while while Dell's going after these deals with NEO clouds right now, that's really where there's a lot of infrastructure demand and build out that at the end of the day. Dell wants the enterprise opportunity just as much as HPE does. And I think it's a situation where HP is more in a position to leverage their networking business and kind of stay in this limbo waiting, waiting for enterprise AI adoption to really accelerate and take advantage of that. Whereas Dell is capturing these immediate opportunity while also positioning itself for the enterprise AI opportunity. But that's where the higher margins are. All OEMs want that. So I think it'll be really interesting to see if, if Dell's strategy of scale and close alignment with Nvidia translates to them winning an enterprise AI.

Speaker B: Right.

Speaker C: Or if HP's strategy of really leaning into networking and a more integrated solution is, is what is the winning strategy in the end.

Speaker B: That's fascinating because I think everyone is expecting enterprise adoption to be growth for everyone, that it's going to happen, it's going to happen at a scale that's so massive that everybody's going to benefit from it. But you're right which way you're leaning in terms of where you are now and especially with HPE leaning more into the networking, um, pre fasting. One other question for you guys. There's four companies you're talking about. That's the peer set.

Speaker A: That's.

Speaker B: Are there other companies within your broader coverage that ramp would apply to or is it more because this particular set is diverging in such interesting ways on their strategy that you wanted to focus on these for?

Speaker A: I think the peer set will be expanded over time. The main reason we're starting here is because in the infrastructure side, AI server has really been where the most, I guess aggressive growth has been over the last couple of years. But storage for example, is another area where things are starting to pick up. Uh, it probably will never reach the hundreds of percent year to year growth but, but it is. That's another example of area where companies like EverPure or NetApp for example will also come into the fold. And the analysis, um, and as Ben alluded to, um, on the enterprise side there's also a lot of emphasis on network modernization as well. So I think that there's opportunity for that peer set to expand in a couple of different directions.

Speaker B: In the services side we have the human Intensity Reduction Index which is helpful because services is a people business. So reducing the intensity of the human element in it is a great way to measure what's happening. But we had to have peer sets based on similarity in their business model. So you can't compare a McKinsey to a Wipro which doesn't make sense with an EverPure and a NetApp. Would you still be using Ramp and comparing them to Dell and hpe, or will you have a new metric that looks at the disruption that is inherent in those businesses that's caused by AI and what that means going forward?

Speaker C: I think we'll have to look at them in those groups of peer sets, kind of similar to what you're saying you do with the uh, kind of of different peer sets in services. And that's just because of the different impact on margins and AI and AI hardware in particular across the different segments of the infrastructure stack. So particularly in server, that's where we see gross margins really being diluted quite a bit with accelerated systems. I think less so is the margin dilution in storage and in networking. So we'll have to look at those companies in groups of storage vendors, compute server vendors and then also networking vendors. I think another thing to note is that while HPE is included in this initial peer set, we'll be looking specifically at the company's AI server and systems business. We have those numbers modeled and are available in some of the products that we've put out more recently, like the AI infrastructure market landscape and the infrastructure benchmarks and market forecasts. We'll be looking at that segment in particular so we can score HP a company with a more diverse portfolio in the same way that we score a company like a Super Micro who is, who is much more server centric.

Speaker A: Right.

Speaker B: And Ramp will be included in sort of all of the company specific reports going forward as well.

Speaker C: Is that the idea?

Speaker A: We will be looking at it from obviously from a benchmarking aggregate perspective, but the plan is to also have individual vendor breakdowns as well as part of that.

Speaker B: So last question, because you brought up there's such a divergence in these strategies in a year from now, are we going to be talking about the great success of one of them or we can just talk about the, you know, absolute collapse of another one? Or uh, do you think. I guess I don't want to put you on the spot of picking winners and losers a year from now. What I'm really asking is do you think the divergence that you're seeing now will accelerate over the next year or will, will it continue at a pace where you see the differences, but they're not so dramatic that, that you're picking winners and losers?

Speaker C: I think the divergence remains. I don't know if it accelerates materially. I think it remains there through at least the end of 2027 and I think it really has to do with AI infrastructure build out driven by the NEO clouds. And then I think if we roll it back another layer, if I think about AI infrast and the replacement cycle of refresh and bringing in new systems and making those investments, I think that really ties to back to token economics. So I think it really matters what a company like Nvidia does, how fast they can bring increased tokens per watt rates with new platforms. Because I think the faster they do that, the more incentive is there for a company like a core weave or a nebius to invest in the, in the newest systems. So I think there's a few dynamics at play, obviously kind of an Nvidia driven market right now, but uh, we'll see kind of what some of these other companies have on the horizon. Of course there's a lot of AI chip startups right now going to market with some, some integrated systems that look at different ways of reducing token costs and then AMD still, still kind of.

Speaker B: They're still there.

Speaker A: Yeah.

Speaker B: Uh, yep. So Angela, agree, disagree.

Speaker A: I agree. And I think where we're at today, when demand still outpaces supply, it's only winners, right? Uh, uh, there's a hierarchy within that, but everyone gets a trophy for today. But I think what Ben's alluding to on when the AI buildouts potentially level off and that's when we'll kind of see who has the best enterprise and sovereign strategies that carry them for an even longer period of time.

Speaker B: Excellent. Well, you guys mentioned tokenization, token economics, sovereignty, and everybody gets a trophy. So we'll come back to this, uh, in six months or so, nine months or so, and check in on where ramp is and uh, who still gets a trophy and how much token economics and sovereignty have changed all that. So excellent Ben, Angela, thank you so much.

Speaker A: Thanks Patrick.

Speaker C: Thank you.

Speaker B: We'll be taking a break over the summer and we'll be back with season six later in 2026. Don't forget to send us your key intelligence questions on business strategy, ecosystems and management consulting through the form in the show notes below. Visit tbri.com to learn how we help to tech companies large and small. Answer these questions with the research, data and analysis that my guests bring to this conversation every week. Once again, I'm your host, Patrick Heffernan, principal analyst at tbr. Thanks for joining us and see you next week.