Rethinking the Role of Servers, a Tech Field Day Delegate Roundtable

I'm Steven Foskett here at, uh, our Special Tech Field Day event with HP's ProLiant team in Houston. We spent the entire day learning about, uh, ProLiant uh, innovations in the latest, uh, generation of servers. We talked about power and cooling.

We talked about AI and, uh, vision. We talked about, uh, security, we talked about server management, we talked about edge computing. All of these topics, though, uh, you know, when, whenever this happens, we always get some, some ideas coming up about the industry as a whole.

And that's what we're gonna talk about on this tech Field Day roundtable discussion. So, let me kick this off. Um, I don't know who wants to be the first victim, but, um, what I would love to know is what did today tell you about the state of the industry in 2025, the state of the servers and compute industry in 2025?

Uh, We're, we're at a huge crop. You know, there's a lot of change going on. It's been building and building.

We're, we're not at the plateau yet, but it shows that the r and d that HP has done has made a difference. And they're being able to adopt their hardware to the evolving workload demands of use cases that we've expressed. So there's, there's, there's good there, and we need to continue to see how they do that and what makes good, makes it better for the customer.

Yeah. So, uh, I'm Jack Poller with Paradigm Technica, and I think that, uh, I, I agree with vg, but I think if we take it a step further, it's, um, really, you know, the premise originally was our DO servers matter. And I think HP shows that not only do servers matter, but what you're doing as from an HPE and a ProLiant is it's not just about the hardware, it's not just about a box, but it's about the entire ecosystem.

Um, lifecycle management, the particular use cases, what you're running on the server, where it is, if it's in the data center, if it's in the edge, all of those things actually matter. And there's a whole lot that's going into what used to be commodity and generic server development that is now no longer commodity, generic. There's a lot of specialization depending on the use case and where it's going, what it's doing.

Yep. Uh, yeah. Guy Coer with future and group Invisible Impact.

I, I, I think that's right, Jack. Um, one of the interesting developments in the industry is, uh, this sort of diversification of processing, um, where, uh, duties that were handled by a c by the CPU, I mean, we've had GPUs forever, but now we're seeing, um, you know, other processing units, which I won't name except for one, which is the asic, because that's what ILO is, right. And, uh, management duties, security duties that are really critically important.

Um, especially in what was really enlightening was, uh, uh, HP's use of, um, GreenLake and ILO for management at the Edge. So what's and for, for, for, uh, dealing with the problems of management at the edge where, um, things are, you know, remote, um, or in special operating environments or what have you, through the use of this piece of hardware, um, on each one of these individual boxes. And I think that, that this is an enabling factor for, um, more capabilities in the workloads under more conditions, closer to the data, closer to the use of it.

Well, my question is to the group is who's pushing the edge of innovation? Is it the software or is it the hardware? Yes.

And, you know, 'cause there's always a matter of catching up with each other 'cause you know, algorithms or software, but, you know, how does your hardware support or vice versa? Are you maximizing That? Well, I mean, the hardware is usually, you know, grown faster than the software.

Software will always, I the hermit crab analogy, right? It'll grow into its shell, but it, even within a hardware, I think that's now been, um, sort of segmented, right? You've got Nvidia who's really outstripping the industry.

They're growing faster than everybody else can handle in every segment, right? So now, you know, HPE and, and, and companies like HPE have to catch up, have to keep up with those guys, and you know, it, it's really straining the entire supply chain going all the way back, going from, you know, how, you know, from a heat perspective, from a power perspective, going back to the, uh, the data center side, um, the entire way, the software part, quite frankly, it's, I think it's the more fun problem to solve because you get to code and have outcomes. And there's certainly no shortage of software companies out there that are doing a lot of this stuff.

Um, probably too many choices, right? And it's one of the reasons that I think a lot of these pilots aren't getting to market. I think it's more than just hardware and software though, right?

So, uh, we're at this point where there's a lot of innovation going on with both, but we've also got processes where, uh, they want to automate all of the coding they want to add, they being the typical, they, right? There's this push to automate the coding, the development of software. There's a push to, um, automate the deployment of the hardware resources underneath of it.

And overall, this is just gonna be my typical answer with AI in general. It's a lot of hype too. I think we kind of saw that today.

The hype is saying, you should be able to do this. And of course, we saw slides that were HPE plus nvidia, which is great that they're partnering together with this go to market that's gonna make it so easy to do. AI will be all good.

What ai what use cases are you solving? I agree that it's a workload. Okay.

So let's figure out how to manage for this workload. Let's figure out how to build the, create the, the proper infrastructure for it. But what are we building?

We leaving out, when we talk about AI and the software hardware, we're leaving out the whole idea of how do you get the data ready for that? And right now that requires people that are not getting paid for the data, the work that they're doing, not really getting, getting paid. You have to do that on the back end of when you do the inferencing to clean everything up.

So AI is a business process. Mm-hmm. It is not, it's a workload that, that runs that business process.

But there's a whole lot that goes with it that doesn't, does that does more, that includes more than hardware and software. So you used the term that, and Glen used it. It was the first times that came up in our discussion.

HPE used it exactly once. And that is the outcome, right? Yeah.

We don't do AI to do ai. We don't buy a server because we're buying a server. It's all in, uh, it's all in order to get an outcome.

Yeah. Right? And that's part of what I'm driving at is AI is just another workload that's helping us achieve an outcome.

For a long time, the technological development was entirely software because we had more than enough compute to do what we needed to do. Right? Now, we've gotten to the point where the software is using enough compute, we need to do more on the hardware.

And so now hardware is now innovating in hardware again, because we need to, because the use cases are pushing us there to get the outcomes we want. My point in our previous discussion was that outside of certain use cases, outright performance is not the metric people care about. Right?

We've had a discussion earlier today about power, right? And how we used to buy, um, data center space by physical space. How many square feet do you need?

How many racks do you need? Now we buy it by how much power do you need? 'cause we're dense enough, we've solved the density problem, we haven't solved the power problem.

When you look at inferencing workloads, the performance metric to get the outcomes we need is not raw computing power, but it's things like latency or power consumption, because we're in an edge location with low power, right? So there's all these different factors that come into play where a hardware ecosystem makes sense. And that's where I think I heard from HPE that there, it's not just, I'm not just selling you a box, right?

I'm selling you more than that. Well, I think to to, to Gina's point though, right? Um, she's talking about this is a lot more than just the hardware.

It's more than the software, more than it's all the different pieces. And if you really look at an AI outcome, right? There's so many layers of it that have to be achieved, right?

And even in the different kinds of software and the hard and, and the things they have to run on, um, that, what I think we're seeing and, and I think it's necessary, and this goes to answer, I think your question is, you know, you know, why would I go to an HPE for this? Right? And I think what you're seeing is that you're seeing a clumping of these layers, right?

HP sells hardware. And so they're gonna have to work with those people adjacent to their la their layer in this stack. And they're gonna clump together and, and work toge and to work to deliver layers in clumps as you go up.

'cause if there's, like, let's say there's, you know, I'll make up a number 40 layers of, of different things as you get from all the way down from the data center all the way up to the, the, the client access layer seven, uh, of, of the, you know, of the last application that touch touches the client, right? And there's all these things in between, right? A customer's not gonna be able to buy and integrate every single one of those layers themselves.

And they're gonna have to purchase bundles of pre-integrated verified stacks of, of technology, including in this case HPE. It'll be Nvidia, it'll be some of the ISVs that maybe they've validated with and created certain configurations and t-shirt sizes of that can be consumed that might take 10 of those layers and squash 'em into one consumable. Then you've got those other layers on top.

And those other layers all include process too, by the way, that'll include that up there. 'cause that has to be included. And that'll be like things that global systems integrators, uh, and, and some of the consultants deliver and, and they'll have to clump together too, be, but I don't think they've done that yet.

And I think it's one of the problems with customers getting to out successful outcomes on this is 'cause they can buy the technology. I think the bottom layers of this have done a good job of delivering technology that can do this. It's everything that's on top of that that I think has not coalesced into something consumable.

And, and that can be deliverable to a client that they can just get, get to happiness. And so right now it's the wild west. And so nothing gets, nothing's getting completed.

Maybe what's happening here is that the injection of this new workload ai, um, has such a disruptive force or a disruptive effect, I should say, has such a disruptive effect that there's a, there's a fair amount of confusion about what to do. How do I upgrade? What do I upgrade?

Is it more on the hardware side? Is it on the tooling side? Is it in the abstraction layers?

Is it in my orchestration layer? Yes. I mean, yeah.

Well, yes, yes. And, and Jack, you know, your point is, is is just, it's just perfect about how there are many elements here and the, the, it's, there's so many, um, ways to define requirements outside of performance. It's Not, not as, not as a chat robot.

Yeah. Yeah. Well, you know, um, if we could just buy everything that we needed a hundred percent all the time, I suppose we could just buy up all the, all the capacity and performance we need it, but we're not in that world.

Yeah. And so I I, I think maybe, um, that, uh, you know, one of the nice things that we saw from HPE today is this multidimensional approach. There's a lot of different ways to skin the cat, and there's a lot of different layers at which it, at which they operate.

Um, it was interesting to hear that dichotomy from them about hardware versus software when a lot of what we saw today had to do with software. Yeah. And, and I, I wanna put ILO in that category too, to a certain degree.

'cause it's an asic it is essentially a hardware package of a piece of mm-hmm. Management software. It just has that advantage of being attached to the, to the node and knowing where the node goes.

Yeah, I agree. Uh, uh, with what you're saying. I, I think one thing that HP showed us today too, one of the best, um, sessions I thought there was, was a security session.

Mm-hmm. And it might be 'cause I'm a little biased, but I thought it was really good. And he talked a lot about, um, about the, the, the coming quantum requirements for if you're working with the federal government.

So it's, it's not an easy job to be an OEM hardware vendor. It takes a long, long time to pro, it takes almost a decade to go from conception to delivering a server. And, um, he was the, the gentleman was talking about how they pulled in the, the dates, the of requirement dates from the government from 23 to 2033 to 2033.

Yeah. 27. That's 27.

And that's like, that's huge. It's a big deal. And they're, they're working on that already.

And it, so you push that back down to the customers that's gonna impact customers. The whole quantum thing will impact customers at some point in time. Um, so you've got this hardware vendor trying to deliver all this stuff right now in collaboration with their software partners and their other vendors to provide solutions for AI on the top of underneath of them different regulations and things changing.

So I, I give it to 'em for that. I just thought it was really interesting, you know, referencing the conversation that Gina was just talking about, that, you know, and I, by the way, that security session was amazing, but they, they were, they were, they're hyper-focused on what's going on inside that box, which they need to be, but they seem to have forgotten a little bit about, you know, what's coming into that box that helps that box operate under optimal conditions. And it's just, and some of that data is just as important and to, um, so I, I look forward to seeing them like expanding that perimeter, but it just, that illustrates the complexity of the whole AI chain stack, as I was calling it before this multilayered stack is that everything's gonna depend on every other stack and everything's gonna have to communicate and talk to and secure all that stuff too.

Right? There's, we didn't really talk about security going all the way up that chain at the record level for g things like GDPR later. I think it's, this is a very, very complex, you know, string ball that they're gonna have to deal with.

Mm-hmm. Uh, that the customers, this is not for the meek. I don't I agree with you.

It's not for the meek. I don't think they were ignoring it, or, um, I wasn't a jar of a No, no, no, no. But I think as, as a designer, you have to put a fence around what you're designing.

You can build the whole thing. You can't build the whole world at once. Right.

And so we are talking, we were talking to the hardware team who was responsible for that particular thing. It is a software and hardware. It is.

And asic actually, what's really interesting is it is custom silicon. It is not, uh, a, an off the shelf part, which is a very important part of the security. And also we talked about on the back channel of the supply chain, right?

It's if it, if you're buying commercial product, you can't guarantee that that commercial product over time doesn't change. If it's yours, you can guarantee exactly what's in it at all times and deal with, um, any other, the supply chain issues, you can, if, if your silicon vendor goes away, you can move that, you can take your own design and move it to somebody else, right, right. Argument.

No, no, no. I'm, I'm looking forward to them looking past box. Well, and I, and I, but I, and I think, I think they did with, and that's part of what I asked, is I think they did that with the, um, is it calm and open view?

And comm is one view, one view, one view. Open view is the, uh, very old product. It's gray hairs.

That's, that's the, the right, that was the one that I did. So anyway, Go ahead. Yeah.

Well, today was a 35,000 foot view of an elephant, and that's what we saw. And, uh, with the hardware, with the ai, with the security, um, each company is gonna be different for what they do. And they talked about how they, that it was the case between, uh, between manufacturing and medical and, and anything like that.

But they're an all in one solution, and they gotta look at that whole view as an all-in-one solution. There's never going to be a, the only case study we saw was during the security we saw. Mm-hmm.

Um, whoever that was and his failed, uh, his failed attempt and uh, where he got was malware, whatever. And, uh, and so we saw a lot of stuff. I was pretty impressed with how, uh, their different, uh, solutions from, uh, big enterprise to small business to remote business or, or anything like that.

Excuse me. And then I was also impressed with the fact that they're looking at what ultimately is gonna happen to an enterprise, uh, in the, between the air cooling and the, uh, and the water cooling, because that's something that, you know, regular PCs are doing. And so enterprise needs to, to, uh, put that into the factor.

And so I think they're just creating an all in one solution so somebody can just turn around and say, this is what I'm gonna buy because they have this, this, and this. I just want to give HPE props, uh, since we were talking about, it's not just hardware and software and that sort of thing for the security approach. Um, it, it was, I mean, they talked about providence, you know, uh, you know, where, where a chain of custody essentially, um, for what, you know, Steven is installing in his server room, um, with an understanding of, uh, um, every potential vulnerability point and how that is being addressed.

So it's not just this piece of software on a chip, it's not just a management software. It's also how is it secured in the factory? How is it secured in transit?

How is secured in delivery? How is it secured and billed? That was impressive.

Yeah, I agree. Uh, I I would actually say more important than that. We, we, I had, we had a quick discussion about, um, uh, how would you distribute and could, uh, HPE uh, distribute boxes on your behalf in an edge case, right?

Yeah. Where you have a thousand locations, right? What really impressed me with that conversation was not so much his answer, but was the very first thing he said was, it depends on how secure you want to be.

And it is unusual for particularly a hardware vendor who's roots are really hardware to put security first. And I thought that that is really, really important. That the very first Security was the first of their design principles.

First, the first of their design principles. And it was the very first words out of his mouth was, well, how security you wanna be? Not sure I can be, I'll give, I'll wherever you want boxes, I'll give 'em to you, right?

Yeah. And so that's, that I think is the security first design. A lot of companies say that they do that, but this really made me feel that they're not just, it's not just a slogan.

It really is security first. Another thing that engenders trust and makes you feel sort of better about working with them was, um, that was just one example of a lot of trade-offs they talked about. Yeah.

So they were not, and they didn't use the old, the usual terminology of, you know, without compromise or what have you. Right? Um, or the inverse of that, uh, whatever that is.

They, they just were discussing, you know, real world utilizations of their solutions. And that was one example of such a trade off, which was to say, well, I mean the, I believe the point he made was, if you really need to be secure, then that's not the way you want to go. We're gonna want to ship it to a central point for view.

Um, but if that level of security is not needed, then yeah, we can ship direct. And then the trade off there is speed of deployment. There were a lot of tradeoffs implied in a lot of what they said.

And I, I think that was really nice. Yeah. And I, I think, you know, on the thing, the trade offs, right?

They were talking about, you know, we were talking about comm and delivering units and, and, and showing the management, right? And we asked a question about, um, you know, infrastructures code and the Terraform and, and Ansible stuff, right? Yes.

And, and the, you know, there really wasn't a, you know, it didn't, it doesn't look like they've really done that, but I'm now what, what they didn't say, and I'm wondering whether or not that they looked at it and made the decision willingly not to put that in because perhaps customers aren't using that as much as maybe we'd like them to or think they are. And so they, their customers haven't been asking them for that at, because you figure if they were asking them for it, they would've done it. And so it's an interesting, it was, I thought it was a very interesting answer that they were, you could tell they weren't very keen on, on getting too deep into the, into that question of, Hey, what about Ansible?

What about Terraform? 'cause they, after I looking online doing some search, there really is no effort that I could see from HP to, to, to provide, uh, you know, Ansible or Terraform adapters for their framework. And I was like, that was that.

Maybe that's it. Not, I'm not gonna blame them for that, because it's possible they don't see the customer need. And I, I thought it was a very interesting thing.

I will say this about security first, and this is how I'd approach a company. If I put everything in first, then the customer is basically going to be going through my system. And if anybody else wants to come in, they gotta play by my game.

And if, uh, if they don't, then they're just not gonna play at all. Well, this, it's been, it's been really interesting. And I think that the, and I, as I said at the beginning, I think the most interesting part of this is always sort of the big picture takeaway that we all get from these presentations.

'cause you think, oh, you know, I just heard a bunch of people presenting on servers or storage or whatever. Uh, no, no. This is a chance for us to really better understand our industry and the customer requirements and where things are going.

And to that point too, I would say that a lot of this stuff, now, I can't speak for HPE because, um, you know, I'm not sure what their, uh, what their corporate mission and goals and so on are. But I can tell you what the people that we talk to today are very customer driven in that they're basically trying to build, they're trying to build for their, their markets. And maybe in some cases, their market is not what we wish it was.

But I guarantee you that it's a big buyer because that's who they're building for here. You know, they're building for some very important, very big companies. And over 12 generations of ProLiant, they have refined this thing.

I mean, I was there for the first generation. I've seen almost every one of these systems, as Vung said, you know, you look at these systems, they're not making minor advances. They're making major advances with every generation.

I do think it's interesting that the major advance in Gen 12 is IO seven. I mean, I, I, I th I don't want to give short shrift to the rest of the product line, but IO seven is a big differentiator. And it's all to Jack's point, it's all about security.

Uh, you know, to sum up, uh, you know, the question is sort of why would somebody buy from an HPE instead of, uh, an ODM or, you know, kind of put things together themselves. I think it has to do with this real, like, truly integrated approach. Like literally from, you know, assembling boards all the way through the supply chain, all the way to your door.

Uh, you know, they'll build racks. They'll, they'll integrate them. And yes, you know, they, they do have a lot of partners that they work with.

Yes, they do need to think about the software that's running on this hardware. Uh, but you know, you can't forget that there's hardware underneath all that software. And this is a company who seems to really get that.

So I really appreciated that aspect of this presentation. So if you watched this and you enjoyed this, uh, checkout on YouTube, you'll find the HPE, uh, ProLiant Gen 12, uh, videos. Just look for Tech Field a look for HPE ProLiant.

com. And, uh, thank you very much for joining us.