Public cloud vs. on-prem: Summit on where each workload belongs

Speaker A: Summit's mission is to keep the technology that businesses depend on running reliably and securely without the burden of managing it themselves. We do it with a team that knows your environment and is there when it matters most. Hey, everybody. Welcome back to the newstack makers. My name is Alex Wilhelm. Um, I'm a journalist here at tns. Now, the cloud, you may have heard of it, it's a big deal today. It has been for a very long time. I'm old enough to recall the early days of the cloud when it was a very simple service with simple at times, falling prices. How things have changed in the ensuing decades. Now, startups today I talked to, a lot of them are still pretty much default cloud when they're started, though AI workloads may be changing that a little bit. But for larger companies that went cloud native, or were told they had to, the bills and complexity are piling up. So what I've done is I've seconded Byron Dill, a solutions engineer over at Summit, to come talk us through what where the public cloud falls short, where private clouds fit in, and what the modern company needs to understand as they choose their IT footprint. Byron, welcome to the show.

Speaker B: Happy to be here.

Speaker A: Now, I want to start with basically how we got to today, because as I kind of mentioned, if you go back to the early days of aws, handful of services, and I recall when they were constantly lowering prices for things like S3 and all that, clearly we're way down the pike. But, but, uh, what are some of the key things that have happened maybe in the cloud in the last 10 years? And I'm thinking specifically about your time at Magnatar Capital, a investing firm that I know quite well, and you were there for a while leading their it. So you really had a good look into how this progression happened kind of in the field.

Speaker B: Absolutely. You know, the, the pieces that we saw when I had my time at Magnatar were, you know, um, kind of making sure we could forecast things on a longer term. Uh, for the it, we saw our teams kind of shrinking, our budgets shrinking. Um, you know, we were, we were trying to be, uh, responsible with the dollars that we were, uh, using to support the it, because every dollar that we were using there wasn't going back to our clients. Um, you know, the big forces that we saw were the, the, the cost stability and projected projecting costs into the future. Um, we also saw some changes in how we had our platforms deployed, but we were still doing very much things in house.

Speaker A: Uh, so at Magnetar, you were a Private cloud, first company, yes.

Speaker B: And we ran everything on Prem, so we had a pretty decent team. Um, but over the years, the demands on that team kept increasing and we really were not able to keep up with a lot of that.

Speaker A: So do you think that it would have been smarter for Magnetar to move to the public cloud and keep their existing team size, or do you think they made the right choice in keeping everything on Prem? Even if they were, let's just be honest, under investing in human talent at

Speaker B: that point, uh, I think we probably should have done a little better. Look at the hybridity in there. Um, you know, there was a lot of inertia. We had built out our own computer rooms. We had invested in, you know, crack units and batteries and all sorts of things that, that needed to get, you know, some kind of return on that investment that we had put in. Um, so we did kind of start to shrink our footprint. So we optimized as best we could using the pieces that we needed to leverage. We went from some 30, 40 racks down to 10 by the time I had moved, uh, on, um, that had come in the place of putting things onto virtual machines, uh, instead of having pizza boxes everywhere, using the larger Cisco ucs, uh, environments to really pile on, um, while the actual physical footprint shrunk. The demands on the team were expanding though.

Speaker A: Yeah, I presume that Magnetar, much like every other company that I talked to, was doing a lot more compute, uh, creating and preserving a lot more data, doing a lot more work with it. Was there anything unique at the company or was it just kind of what we've seen in general in the enterprise, which is more data, more compute, more demands, and less patience, I would say, for downtime.

Speaker B: I think it's very similar. I mean, there's no secret sauce there. I think, uh, there are some things that we did that were different, but the underlying, what we had deployed, if you walked into one of those computer rooms, would look very much like everybody else's computer room. Um, you know, the one thing that we had that is maybe different is, you know, the demands of our end users, the traders. Uh, that's where you get into that. You know, there's no downtime, there's no acceptable, especially during trading hours. Um, in fact, most of the time we had rules. You can't even go into the data center during trading hours. We didn't want people in there even like opening a cab or looking around, um, just because, uh, and not that we had had issues. It was just our rule set.

Speaker A: No, you, I, I Bet you didn't want the people to breathe on the servers during trading time because if you're doing high frequency trading, and I'm not saying the Magnitar was, but if you were, then milliseconds could cost you millions of dollars. So I, I appreciate this context because when I think about the private cloud, I think about this kind of setup, uh, a little bit of inertia at the organizational level. This is the way we've done it, this is the way we're going to do it. A bit of maybe some cost fallacy. We already have these rooms, we have these machines and then teams that are a little bit overrun and busy trying to keep up with increasing demands with probably a limited budget. But there's also another element to this which is just security. And I know that everyone listening to this has been kept up on Anthropic's Mythos model and how everything's going to get insecure. But I'm curious if you think that the on prem private cloud system or setup is more secure for the average company than using the public cloud. Because on one hand if it's in your house, you can air gap it, etc. But if it's in the public cloud, you have access to all of Amazon's research team. So where does the balance kind of come down on the security side of this debate?

Speaker B: Uh, you know, the security comes in where you got a lot of folks that still have a very server hugger kind of mentality. They, they want to be able to touch a truck. You know, my time there, we saw uh, compliance teams from outside investors come in and you know, not just want to look at the servers and make sure things were actually there, but actually started asking questions specific to what's your business continuity program? What happens if, how are you going to handle this scenario? So the end user of uh, most of our companies, you know, not just Magnetar, but most of the companies I work with are very much, uh, when it comes to security. They're not just checking a box out there. They're not just saying, oh, do you have SOC 2? Uh, summit? And we'll just use that. They want to build their own policies, procedures and controls on top of that. Um, so the users have become clients, I shouldn't say users, it's a very Tron esque, uh, statement of me. Um, but our clients are becoming much more savvy to their needs and the risk profiles out there when it comes to security. Uh, okay. You know, they feel like, hey, well I'm on a Shared infrastructure if I'm in the cloud. So that's, you know, is that inherently insecure? And I tell clients, no, it's not. It's really making sure you have the controls and you have the right design to take advantage of what's available.

Speaker A: So if we're thinking about having kind of the right workload in the right place, where do you think companies should think about putting stuff on the public cloud versus on a private environment? Because I think a lot of people, like every startup that I talk to is just default cloud they incorporate. It's usually a Delaware C, an AWS account and an address. Right. Like that's kind of like the startup stack. Um, but where should companies, as they get larger, begin to think about moving stuff off the cloud? Where do you see that having the most impact in terms of control? And I think also security?

Speaker B: Absolutely. That model is perfectly viable. I think what it does with startups and clients that are just starting off, they need to focus on their product, they need to focus on what they do, not the infrastructure. And I think those, those, um, hybrid cloud or the, the public clouds do a great job in providing that for a client.

Speaker A: Yeah.

Speaker B: Um, as that client grows, as the workloads become more stable, you move out of a prototype or a fast, um, prototyping phase and you get to some stability. That's when I feel that the thought needs to be looked at. Is this workload now right for the cloud? And it's not necessarily a cost, but yes, there's cost factors in there. But there's also, like you said, there's security. Hey, this is, this is not a stable workload. And security doesn't just mean you're bad actors or you're, uh, someone that might infiltrate or do something that you don't want done to your infrastructure. But it's the security of knowing it's running in the right place. It's got the right backing, it's got the right people to help you manage it. So then you're off to focus on that next wave of product or the next release. Um, and security also comes into, uh, we see a lot of clients that, you know, they have a developer team that just, uh, didn't understand the economics behind it and ran something. They find out at the end of the month. Someone ran 20 GPU servers for 730 hours that month, and here's the bill.

Speaker A: Yeah, and then that person gets to discover what the job market looks like, because what a fun time to look for one.

Speaker B: Um, he hasn't just, you know, You've got your firewalls and you've got your endpoint protection and you're analyzing logs. It's very much security is knowing your environment and what's going on inside of it.

Speaker A: Your point about iteration and how companies, once they stop essentially rebuilding their product every three months, end up with a need for more stability. I presume that that kind of encompasses nearly every single major enterprise software company, or just enterprise period, that I can think of, because startups are a pretty small chunk of the market. So it sounds like when we think about being public cloud first, it really only applies to younger companies. And as they all get larger, they should explore more of a hybrid solution. Is that what you're shooting for there?

Speaker B: Uh, I think it's, yeah, as the workloads mature, I think it's not necessarily the size of the company, but as the workloads mature. Um, and you don't need that super flexibility. You can, you can really sit down and say, what does this workload need? We can help design that specifically, because when you're in a hyperscaler, you're really, the Lego blocks that you have to build your solution and run your application are very set. Whereas when you come talk to myself here at Summit, we're going to ask you, what does your workload need? And I think that's kind of shocking to clients. Sometimes they want to talk technical, they want to talk servers, they want to talk licensing, and it's really getting them to take a couple steps back. Now you get to reassess. Let's do a workload readiness assessment. You've heard that from back in the 2010s to move to the cloud. Let's not rush back out of the cloud. I think is very much a mistake that folks are making. Hey, the cloud's too expensive. Let's get out. And that ends, uh, up the same problems. But you just kick the can down the, uh, street, maybe three to five years, you're going to run into the same problem.

Speaker A: So don't just follow the herd. Don't go wildly to one extreme than the other. Think about the actual workloads in question. Uh, can you go a little deeper for me on the idea of, uh, major cloud services being kind of Legos or building blocks that snap together in kind of rigid ways versus what you guys do with customers that come to you with their own workloads and workflows and how you can better fit, uh, compute and storage to those. Because I'm not quite sure how that actually works in practice. If that Makes sense. Byron.

Speaker B: Yeah. Uh, the biggest thing is you can go to your console, your favorite hyperscaler console, and there are a lot of what I call nerd knobs and levers that you can pull and you can build your solution on. And it can be very overwhelming for a client. And then you get a lot of tinkering, a lot of, well, hey, can we make this do that? Um, and then also realizing those feature sets are changing and unless you're going to be a major spender in that level, you're going to, you're going to be forced to, uh, adapt your workloads to feature changes and updates that are made.

Speaker A: Um, how often does that happen? Because I watch every single major developer event, I read the blogs, and I am blown away by the sheer amount of stuff that gets announced. And I'm always a little bit torn between, on one hand, great, lots of new stuff, you know, toys for developers, go forth and have fun. And on the other hand, how the hell do they keep up with all this? It seems like a punishing amount of new, which is good. But if you're running these, you know, stable, more mature workflows, do you, do you care that.

Speaker B: That's kind of the thing is, is you may care on the newer workloads, on the newer things coming down, your AI workloads, your uh, LLMs and, you know, creating better chatbots and assistants for your, for your employees. Sure. But do you need someone to really, um, you know, monitor that application that's been, you know, a moneymaker and just running solid for, you know, three to five years? Or can that get shifted to somewhere else so you can focus on those things? And that's where I, you won't find me saying it's all on prem or it's all in the cloud. There's always a hybrid moment for everybody depending on the workload, not just the, uh, how long you've been doing this? Yeah, you can wait a long time for infrastructure to get spun up and. Long time, I mean, weeks. Whereas like you said, you can go to your hyperscaler, plug in a credit card and boom, you could be off and running in a moment just to test a hypothesis or, or test your code, uh, in a place, uh, that then you can find out, is this even worth it?

Speaker A: Your point there implies, ah, ample available hyperscaler capacity. And one thing that has been talked about ad nauseam for the last 24, 36 months has been kind of a general global compute crunch, usually through the lens of AI loads. In particular, has that Same compute crunch impacted non AI workloads to a similar degree. As in our company is still able to get as much AWS Azure as they want whenever they want it. Or is that now more of a if you can get it if it's still available type thing?

Speaker B: It's a challenge. Um, so I think that it is permeating the entire environment. So even for us working with our partners, um, we're seeing the crunch. It's either um, pay through the nose or you're still going to pay a lot and you might get it a little later. Um, and we're not seeing that ease up for quite a while. Uh, now Summit's well positioned, we've got plenty of hardware, we've got some solutions where clients may not be able to scale with big iron, you know, deep. So you have, you know, your, you know, hundreds of cores per machine and terabytes of RAM per machine. Because of the crunch right now you may have to spread a little more horizontal using, you know, servers that are a little, you know, less powerful. But we can get the GPUs in there, we can get this. The, the RAM footprint in totality can be what you want. Um, the challenge there is, was the application written to do that. We have to match the infrastructure to the application or the demand that it needs to have. So while, yes, I live and breathe, you can't give me that until when or that stick of RAM costs how much and it won't be here until when. Um, it's pretty wild, but there's a little bit of uh, past is prologue. During, uh, Covid, there was a big, big crunch as all the factories started to shut down. But that was a very much of a blip that came through the market. This is something that talking to our suppliers, they're saying we're sold out through 2027 or longer.

Speaker A: In the case of some memory prices. If you don't know what Byron's talking about, just look up some charts for RAM prices, especially higher, uh, end high bandwidth memory. The price decision has gone absolutely through the roof. So I'm curious how this impacts Summit because you guys have data centers around the world. I think you have more of a US footprint, but definitely some global as well. Are you guys able to get the parts that you need to keep expanding that to meet customer demand, or are you going through the same crisis, frankly that we're seeing everywhere else, which is we can't get transformers, we can't get walls, we can't get workers, we can't get, welders we can't get. H Vac, we can't get. I mean it seems to be top to bottom kind of a mess.

Speaker B: Yeah, we're, we're dealing with that. I, I'm not going to say we're not because I think everybody listening would probably know, uh, that that's me up here lying, uh, through my teeth. But yeah, sure. The thing that we do have is we've got a large bare metal offerings. We have plenty of servers that we can leverage. Um, we've built a global IP network and we've had it in place for decades. Um, we've got great partnerships with most of the providers that allow us to scale quickly. Um, even when it comes to power. We don't necessarily tell a client. Well, we don't tell a client. No, we just try to say, well, what does the ramp look like? You know, hey, you need, you need 5 megawatts. That's a very difficult ask right now. So again it's very much. What are you trying to do? I, I doubt you're going to on day two, light up five megawatts worth of power. Let's talk about what your journey looks like. And part of my job and any solution engineer out there that's listening is being creative is finding a path forward that gets you to where you need to be. Not necessarily a, hey, let me just go pull this off the shelf and give it to you and we're done. Even the hyperscalers are starting to deal with this. There's, there's, there's a crunch that's going on throughout the entire uh, landscape that even your aws, Google and Amazon, uh, are going to deal with.

Speaker A: Yeah. And then the best part is after eight years, all the parts appreciate to zero and we get to do it all over again. It's a really fun time or it's a great time to be a memory maker or a GPU manufacturer or just TSMC M writ large. Now you guys offer private uh, clouds to companies which you and I were talking about before we hit record doesn't um, mean that you go into their building, into their basement and install racks for them, but you more provide um, their own machines even if they're co, located in one of your data centers. Yeah.

Speaker B: Yes. So we'll provide a stack that is dedicated to a client and that can be typically. It's a private cloud in my vision is dedicated compute. So you have your servers, your ram, um, running a platform. And the great thing here at Summit is we got multiple platforms we can run on those from VMware, Hyper V, Proxmox and Kubernetes. So we can find multiple paths to solve. In fact, some of our clients have a mix of all of those and they still have a hyperscaler footprint.

Speaker A: Oh, that sounds very. Not complex and not wasteful. That sounds great.

Speaker B: Yeah, well there's a trade off, there's complexity, whereas I can maybe increase complexity. But we're bringing the entire Summit team, we're bringing hundreds of engineers here to support you. So it's not a, we'll build it and it's all yours by, you know, we'll send you the invoice each month.

Speaker A: I was going to ask about that because when we were talking about Magnitude Capital, your time there and the team that didn't seem to scale to the increasing workload. I've worked in a lot of companies. That seems to be kind of the way things go. And if you're going to bring compute, you know, from the cloud, where it is by definition managed by someone else's dedicated team, to your own, you know, are we going to end up in the same situation where you have once again an understaffed human team running more machines? But it sounds like Summit actually steps in and ensures that companies don't wind up, uh, over their skis.

Speaker B: Yes, absolutely. And we work with the teams we want to partner. And that's what I was saying earlier is a challenge. When you go in and talk to clients, there's very much a, um, concern for some of these teams especially that have operationalized around a specific platform, be it AWS, VMware, um, Kubernetes.

Speaker A: Sure.

Speaker B: How this is going to work, where does the racy change from them? Because they're, they're worried about their job and yeah, uh, yeah, yeah, we come in and, and that's one of my, one of my former um, colleagues, uh, Paul Armanakis, I'm going to name drop for him. Uh, you know, one of the things he said was when I first started here at Summit is our best clients, where we have the most success are where we are partners with the client, not just providing services or vendor and they'll call us when they need something. So that partnership is critical and I think that's what a lot of clients are seeking. You know, you can be a very anonymous face in the crowd, um, for some very important apps when you're on the hyperscaler versus when you work with a managed service provider, especially Summit, you get on first name basis with a lot of the engineers that are helping you um, so you get to know people, you get to know them, um, on a level other than just, hey, well, I need this app restarted or I need that server kicked.

Speaker A: Okay, so that's the second time, I think during our chat you've gently said that customer support from hyperscalers may not be particularly good unless you're a simply massive customer. Is that what you're getting at, Byron?

Speaker B: I think it is. There's plenty of managers. I don't want to dog on everybody too much, but there's plenty of managed service providers and we partner with them that shout out they are an AWS MSP or Azure msp and they focus on that and they do a wonderful job there. But yes, they exist because there's just so large a footprint of different companies. Yeah, unless you're willing to pay the money, it's hard to get that kind of support. It's hard to get a first name basis or a I want to talk to X engineer because they know my environment. When I was at Magnetar, I had, when I was a client of, uh, back in the day, server centrals became deft and now Summit. I knew the engineers that knew my environment and I was willing to wait an hour or so to talk to, you know, Jonathan Warren, who's still with Summit, uh, because he understood what I was trying to do. There was a lot of just that knowledge and not necessarily that it wasn't documented, it wasn't tribal knowledge. It was just, I knew Jonathan and I could speak a language that was different than, you know, finding another engineer or something that someone that was unfamiliar with it.

Speaker A: I mean, I'm a big AI bull, personally, but I also kind of like being able to call somebody and get some help because I have found that humans are a little bit easier to talk to than, uh, chatbots, you know, just generally speaking. So that all tracks with me. Okay. So companies were told for a long time, you got to go to the cloud or you're falling behind or you got to go to the cloud or you're not modern. They get there, they move their workloads over there. It gets expensive. Expensive. I speak English. It gets more expensive as time goes on. Was that kind of a frog boiling type thing? Like maybe they didn't quite notice how the prices were stacking up, but people complained about their cloud bills. Has been, it feels like pandemic for the last 10 years.

Speaker B: Absolutely. It's just that slow burn and then eventually someone turns around and realizes, you know, the trend of, uh. Well, we used to spend $2 million a year and now we're spending $5 million a year. And you know, it's there, there's a panic that'll set in for that. And it's one of those, you know, maybe the people that, that set down the everything needs to move to the cloud edict at a company has since moved on and you have different leadership that is, uh, that is not necessarily anti cloud or anti, you know, on prem, but says, well, are we efficient? Are we actually running things in the right place? And I've dealt with a lot of decision makers, uh, that start those calls with us. Um, and then we get, you know, kind of, we bring in some more of the technical people to figure out the nuts and bolts. But absolutely, it's a slow burn. Um, I find a lot of times though, it happens slowly over time, but quickly when they come to us, when you have a change in leadership or there's an audit that went on, or there's a merger and acquisition phase that the company's going into that triggers the conversations that I personally love to have with clients.

Speaker A: So we've talked about a number of ways that companies that are heavy public cloud users run into issues. Cost, inertia, blah, blah, blah. What part does just data sovereignty play into this? Because it has seemed to be a pretty chaotic couple years geopolitically. And I'm hearing a lot of, you know, European companies say we want to buy European, we want to keep all of our data here. So do companies come to Summit looking to build their or secure their own private cloud to help resolve those issues?

Speaker B: They do. And we've been able to build, um, our infrastructure in Europe, uk, apac, um, to meet those needs. Um, every data center that we're in started with a specific client that needed to be in that location. Ah. So we, we will build a bespoke infrastructure. We will make the strategic decision to go into a certain market, um, not just for that one client because there's a, there's quite a bit of build cost for us when we go in.

Speaker A: Well, I mean, I think everyone's really aware of how much it costs to build a highly performant data center. I mean, even before everything got three times as expensive, it wasn't cheap. So, yeah, that, that tracks.

Speaker B: So, yeah, we, we end up, you know, with clients that have multiple locations. You know, hey, they need all their data resident to Canada. Um, you know, they need to be in a very specific. And I see that changing a bit. And uh, I think this is what you're getting at, is it's not just, I need my data in Europe, it's no, I need my data in Germany or I need it in the uk. Like they're getting hyper focused on where that data needs to be. And I think some of that is, you know, regulatory. There's different laws passing, there's a geopolitical climate right now, um, that clients are. And if the underlying everything of all of our conversations is de risking, they're looking to mitigate risk even if that costs a little bit more. They want to understand their risk profile when it comes to how they're deploying their critical assets.

Speaker A: Okay, so let's say that I'm a large enterprise and I've run into these problems that we've just gone over and I think, okay, it's time for a change. We have to at least start looking at other options. So if I came to summit and I said, look, we want to look into getting our own private cloud to handle a chunk of our workloads and how much would you recommend that a new client actually take off the cloud on average? I know every case is different, but I'm kind of curious, like, what portion of compute do people tend to repatriate to their own private cloud once you guys get it up and running for them?

Speaker B: Uh, it's hard to put a number on that. I, um, will say that typically there's effort in moving the workloads. So you have to have a conversation about that. You have to, have to say, you have to say, hey, there's costs involved in the movement and if what you're going to save does not justify that cost of the move, we need to talk about, like, is the move worth it? Got it. Usually we can make it worth of it because it's not just a dollars and cents. It's a risk, it's a skill set, it's a change in strategy that we're supporting. Um, for me, you know my rules and I'm not going to speak for everybody, but mine is, you know, if I can save you 40%, then it usually becomes worth it too. Um, repatriate a workload.

Speaker A: 40% cost, essentially.

Speaker B: Cost, yes.

Speaker A: That's, that's a lot more than I thought you were going to say. How, uh, is the public cloud that expensive?

Speaker B: Well, I think people underestimate how much inertia the cloud would have. How much inertia that platform, wherever they are, not just in the cloud. Um, you know, we've got clients moving from, you know, uh, uh, ah, vm, um, specific platform to kubernetes there's still a, hey, we can't gloss over the fact that everybody's going to spend months working on this and there is a cost to those months of working on it. Um, so starting at 40%, I think everybody's comfortable. You get a nice round number that people continue to have conversations with you because it's guiding everybody. We have to do a, ah, discovery. We really have to understand the environment. Um, and like I said before, we found clients. It's like that workload is perfect where it sits. That is the best solution for this, even if it may cost. And then we talk about risks, we talk about backup. What's your doctor strategy? Um, if it's that critical to your workload, um, that's its own tech talk, uh, for doctor and backup strategy. So I won't dive into there. But um, again I'll go back to the risk. It's not just a move, it's a modernization of what you're doing. And I think a lot of people treat it as a we just need to get off the cloud. Well, hold up. 10 years ago, 15 years ago we said we need to get on the cloud, right? That obviously didn't work because we're here today. Let's figure out the best strategy and align to that before we move one bit.

Speaker A: So let's say I'm um, the same company from the same example and I have my own team on site that helps manage all my public cloud commits that I'm currently using. Can I keep that team if I move partially to the private cloud? Do they have skills that transfer one to one or do I need to bring in new people to help manage my, my own private cloud and then also keep the people that are managing my public cloud commits as well?

Speaker B: In many cases you keep the, you keep your people, I mean you can optimize. But again that goes back to the. We're trying not to, I'm not looking here to put anybody out of work. Um, you had mentioned it also that when you start doing this you increase complexity. Yeah, well, complexity requires more smart people to help manage that complexity piece of it. So there's a little bit of a trade off. There's that, you know, hey, I know how much this workload is going to cost me for three to five years. I know how it's connected. Here's the people that are going to be working on it because the ultimate experts at how the IT serves your company is your employees. And I will never, as a solution engineer, be an expert in what Your company does. And the thing we get to say to our clients is, you get to focus on your workloads and what IT and what the stack does for your company to grow, not just the IT as a boat anchor, you have to have all these people just making sure servers are blinking and fans are running. That's not a value driver for 99% of our clients. They need it to do something else for them. So, um, turning around and saying, hey, guess what? You know, rather than just, you know, doing the same thing in AWS or, uh, VMware Hyper V every single day that you've been doing it, wouldn't it be more fun to build new things for your company? Wouldn't it be fun to kind of push the envelope of it, not just for your company, but for your skill set?

Speaker A: Uh, focus on what you're good at and don't focus on what you're not good at. I mean, there's a reason why when the major tech companies in the Bay Area were putting together bus routes from San Francisco and back down to their offices, they didn't become bus companies. You know, it's not their core competency. So from that perspective, then, Byron, it feels like Summit is offering, therefore, uh, the benefits of private cloud with the help that you might need to make sure that it works smoothly. So you can kind of have your cake and eat it too, as long as you have enough workloads that are private cloud applicable.

Speaker B: Yes. Yeah.

Speaker A: Okay.

Speaker B: And there's plenty of clients where we do that. We do that assessment with them, and that discovery is as much for us as it is for the client. Many times. So I like to talk about that digital therapist, um, getting them to understand what they're doing and how they're doing it. Um, we've done paid engagements for that. But, you know, as a solution engineer, my job is to ask questions and be curious. And many m times those questions uncover a client asking themselves, you know, further questions about, what does this look like?

Speaker A: I've long thought that solutions engineers are kind of like the investigative journalists of the technology world. Is that a. Is that a fair analogy, or am I just saying that in my head and it makes sense?

Speaker B: I think that's true. I just wish there were a school you could go to to specifically become a solution. Uh, I always find it interesting when I talk to New Sesame. How did you get into. How did you become an se? Because our paths are never the same. Uh, and it's very unique. So you have to have.

Speaker A: What's the key driver? Because I feel like when it comes to unsung heroes, this is part of the. You guys help keep the technology world running, but no one wakes up and goes to college and says, I'm gonna go and be an se. You know, Exactly. How do people end up in your seat?

Speaker B: I think the curiosity, the wanting to do different things and challenge yourself. Not that that's dogging on anybody else.

Speaker A: No, no.

Speaker B: But, um, you know, for me, my journey was I came through multiple different, you know, verticals. I started at healthcare. Uh, I'm sorry, I started at education, went to pharmaceutical, entertainment, healthcare, financial, and, you know, at some point, you know, you kind of. And I hope other people would feel the same way. There's only so many times you can keep upgrading the same stack and you want new challenges. Um, I will say at an MSP here, especially here at Summit, there's been no shortage of new things coming along. Not, uh, just from the overall landscape now with AI, but how clients need to use it. Um, that's always. It's fun. It's fun to talk to new people, it's fun to sit down and find out what keeps you up at night is my favorite question. Because you can ask a CFO that, you can ask a third shift data, uh, center person that, and they will have an opinion on it. And that usually uncovers a pain point that you can say, okay, hey, I've dealt with this. Or I've had clients that have dealt with this. Let's talk.

Speaker A: Yeah, I was trying to avoid having our chat be too AI heavy because I feel like whenever you turn on the news, a podcast, even look up in the sky, it's all about AI. But what is the AI case for building out a private cloud? Because everyone's freaking out about token costs. Everyone wants their employees to have access to lots of AI, but they also don't want to pay for it at, uh, list prices. So our companies now, in the last, I don't know, six, nine months, coming to Summit and saying, hey, can you guys build us a private cloud that's more GPU than CPU heavy? Because we specifically want to bring our AI workloads in house.

Speaker B: Absolutely.

Speaker A: Oh, okay.

Speaker B: We were working on a client for renewal just the past three months. And, you know, right. You know, right as we're trying to get signature done, it was, oh, well, hey, we do all this AI stuff and it's costing us a fortune. Can you guys build us, uh, GPU capable servers? The answer is absolutely. Um, let's talk about the workload. What do you need, what models are you running, what is this in support of, what is the driver and what is your return expectation? So we talk to clients about that and build that out. Um, in fact, years ago, before this big fad, we had clients, you can go read, um, about Basecamp and DHH and the things that he's talked about with working with Summit and how much it saves. Um, a different client we brought in and we deployed racks and racks of GPUs. And you know, the ROI on those versus running it in the cloud were enormous. But their workloads were also stable and mature enough that that made sense. And so they bifurcated, they ran just like we're talking about with your hyperscaler versus private cloud. Running those stable workloads on an on prem or dedicated instance of GPUs and then keeping their new models and their rapid prototyping in the cloud where you can fast react is how they ended up doing it. So it's not just your private cloud versus aws. We're seeing that in the AI space, specifically. GPU space specifically.

Speaker A: So going back to the top, when we're talking about startups being cloud first and then when their workloads become more mature, maybe more private, it sounds like for even companies that have some mature and some fast iteration, having a foot in the cloud where you can access things that are absolutely brand new very quickly and have a lot of flexibility is good. And then move all your more mature stuff to your private cloud. You could really, I mean, I hate to say it again, but have your cake and eat it too. It sounds like the best possible arrangement because you can save money and have everything that's cutting edge at once.

Speaker B: Yeah, diversification. I mean, we talk about it in your, in your management portfolio, your investment portfolio, it's the same thing. Um, don't have all your eggs in stocks or bonds. Well, don't have all your eggs in hyperscaler or on prem.

Speaker A: It's joke's on you. I have all of mine in Solana and SpaceX stocks. So I'm either going to be worth $0 in five years or $10 trillion.

Speaker B: Byron. Yeah. Well, don't look at the markets today.

Speaker A: Just in case anyone was thinking. I'm serious. That's not true. I'm a diversified index fund guy. No worries. Um, all right, one last question before I let you go. What, uh, percentage of companies today that have a public cloud bill of over a million a year is, do you think are probably good fits for having some private cloud in their life?

Speaker B: I, uh, won't deal in absolutes, but I'll say all of them. I mean, there's an opportunity for everybody, um, especially if you're spending over a million dollars, um, per year, even in hyperscalers or on any platform. It's the. What is the inertia? There's always opportunity to, um, investigate your workloads with someone that's got the experience doing it, like Summit, and figure out where does that fit. And even at the end of the day, you find out you are in the right place, it's a valuable exercise to run.

Speaker A: Yeah, it's like, again, back to your point of that it's therapy for your it. I mean, frankly, we could all use a bit of rummage and a reordering. Well, you heard it here first, folks. Uh, only Summit deals and absolutes. And if you want to m learn more about what Byron's talking about, you can go to summithq.com roi simple URL summithq.com roi to learn more. Byron, thank you for sitting down with me and talking me through all of this. I am very glad that I just get to use the cloud instead of having to build it. So thank you for all that you do.

Speaker B: No problem. Thanks for having me.