The pendulum swings: the current state and future of cloud repatriation

March 25, 2025
John White

In this episode of Sovereign Database Decoded, Vinay Joosery sits down with John White, COO of US Signal, to discuss the cloud repatriation movement, what repatriation actually means in today’s infrastructure landscape, what its key drivers are, its current stage and direction of travel, and what you should consider when deciding what, when and how to move back on-prem or in a private cloud.

Key insights

The repatriation movement shouldn’t come as a surprise

The pendulum swing back to on-prem shouldn’t come as surprise; because, the narrative of the cloud as Nirvana is far from the reality that it’s a tool to solve problems. In this case, it’s just a different way to build and buy infrastructure — so a reckoning in the shape of right-sizing was always in order.

⚡ Public cloud as a business solution, not a technical one

Many people think that cloud adoption surged due to it solving a technical problem. In fact, it surged as a business solution for hardware requisition and financial planning, initially making everyone from the developer to the CFO happier. However, it introduced unaccounted for technical complexity.

⚡ The reality of the cloud’s image as an easy button

The cloud’s image as an easy button wherein the provider takes care of everything is far different than its reality, sowing confusion and driving dissatisfaction; for instance, what the shared responsibility model actually entails. The cloud does not, can’t, and shouldn’t solve for everything.

Ops knowledge is not a repatriation blocker

The cloud operating model is driving strategy, architecture and skill development within enterprises. This, along with the proliferation of providers like US Signal offering alternative infrastructure options, is making repatriation an easier proposition for organizations.

⚡ Latency is the next driver of repatriation

People will continue to become less tolerant and AI-augmented machines intolerant of latency, meaning more distributed workloads, supported by local nodes materialized by data centers large, small and mini, owned, co-located, and rented.

⚡ Operations Anywhere is the future

Coined by John and team at US Signal, the Operations Anywhere concept will require service providers to ensure that customers can run their workloads anywhere, regardless of their infrastructure — workload ubiquity is closer than ever.

Episode highlights

💡 Getting our vocabulary straight [05:04 — 06:51]

Before diving into the repatriation movement and its drivers, Vinay starts out by asking John how he conceptualizes the space. John breaks it into three buckets, virtualization, private and public cloud, the former being enterprises taking, running, and virtualizing their own servers via solutions such as Nutanix and VmWare, the middle as hosters building services off of those and transforming those into OpEx, and the latter as larger, hyperscale projects that are a little different.

💡 The pendulum is swinging, so what [06:55 — 08:54]

John acknowledges that there is a pendulum swing back to on-prem and private cloud, but he doesn’t think it’s a surprise to anyone. The public cloud’s perception is developing into one as a tool like any other and there are also cost drivers expediting that. For instance, he mentions that estimates of wasted public cloud spend hover around 30 to 40% — that’s billions in real money and right-placing and right-sizing of workloads was only a matter of time.

💡 Cloud and on-prem ops merge [15:00 — 19:11]

Vinay questions the repatriation movement in terms of operational SLAs to which John speculates that the differences in expectations and realities between operations in the enterprise and the public cloud have mostly been resolved in the decade+ that public cloud has become a thing, stating, “Every application has to be run like a hyperscale application, essentially.”

💡 Should I stay or should I go… [19:57 — 24:09]

John says both; pick the right environment that’s going to help run your workload efficiently, performantly, and securely. Using an analogy he picked up from an analyst, train your puppy where it’s ok to wreck and take it home once you know it’ll be a good dog. Run the ROI and TCO analysis, understand your application’s performance tolerances, and decide.

💡 How to unite ops fiefdoms [30:34 — 32:04]

John acknowledges that repatriation and running workloads in multiple environments entails complexity and dispute. One way to overcome the zealots and true believers is to build consensus by creating something along the lines of a Cloud Center of Excellence — this will help get everyone on board and speaking the same language.

💡 Cloud’s still early innings [34:37 — 40:36]

Vinay and John discuss the fact that we’re still in the early stages of cloud’s life cycle, earlier still when you consider AI. John speculates that latency will be the new currency, driving data center development, including mini ones, as enterprises strive to get data closer to the consumer, human or machine, pushing hybrid models.

Here’s the full transcript:

Vinay Joosery: Hello and welcome to 2025’s third episode of Sovereign DBaaS Decoded. I’m Vinay Joosery, and this episode is brought to you by Severalnines. So today we’ll talk about repatriation, the movement of workloads from public cloud back to on-premises or hybrid environments. To help us unpack the topic, today’s guest, John White, COO at US Signal. Thanks for joining us, John.

John White: Yeah, thanks for having me. Appreciate it.

Vinay Joosery: So, first thing first, tell us a bit about yourself and what you do.

John White: Sure. So I’ve been in the infrastructure industry pretty much my whole life. Started a hosting company out of my bedroom when I was, like, 13 and just never really stopped. So I’ve done a few runs here; just really enjoy building the foundation and plumbing that people need to run their applications. And so at US Signal, I jumped on board about two years ago now after they got purchased by a new infrastructure fund. So it was a single owner operator. And then we moved into a PE backed, which is where I’ve really spent most of my life, is being PE backed. 

And, we’ve been really just, on a tear the last two years since that acquisition occurred. So as a daily, I’m COO and that’s kind of a broad title. But I have a great partnership with Dan Watts, the CEO, and the board; and, I’m here to help build new products, help figure out where kind of the tech industry, you know, where we need to kind of build and go from a tech industry. So I have engineering, delivery — when we sell new customers, they roll up through me. As well as that, support. So kind of the whole cradle to grave, from building the products to helping sell the products, from solution architecture team to engineering the products, delivery and support of the products. So, it’s an awesome job. 

Believe it or not, I actually don’t feel like I work much because this is exactly what I love to do. And we’re having a lot of fun doing it, and it’s crazy. For so many years, I told people I was in the data center industry. People had no idea what the heck I was talking about. But, with some of the latest and greatest with AI and then Trump getting up and talking about how much data centers are important to the United States, it’s actually been pretty in vogue, which is kind of rare in the tech industry.

Vinay Joosery: Yeah. Excellent, excellent. Well, welcome. It was nice to meet you in Madrid.

John White: Sure. At the CloudStack conference.

Vinay Joosery: Yeah. So I’m glad that we met there, looking forward to Milan as well.

John White: Yep yep yep.

Vinay Joosery: So today we’ll be talking about the building repatriation wave. You know, we’ll speculate a bit on the causes, we’ll consider the specific challenges to organizations wanting to join it and what the future of infrastructure environments look like — especially in the light of AI. 

So the past ten, 15 years, we could say has all been about cloud, you know, market size $330 billion. AWS, Azure, Google has about 65% market share. And last year the cloud services market grew, you know, 20%. So half of that growth was was due to gen AI. Now, interestingly, last year, the Barclays CIO survey suggested that more than 80% of enterprises were planning repatriation of workloads back on-prem, private cloud. Right. And this is similar to a 2023 IDC case study. Now, before getting into this on-prem versus public cloud discussion, how do you conceptualize the space?

John White: Yeah, that’s a pretty good question, I think. And it’s very nuanced. And I think that’s part of the reason why people are saying repatriation will be a thing or won’t be a thing. And so the way I look at it is we kind of have three buckets. We have virtualization, which is what enterprise has been doing for a long time. Some people labeled this cloud, you know, some people tried to make it a cloud thing or tried to call it a private cloud. 

And that that’s, you know, we kind of all know that. So that’s your Nutanix, your Hyper-V, obviously your VMware, running inside of the enterprise and then the hosting companies and even myself included in the past have built offerings on those that are OpEx offerings that would be considered private cloud. So, you know, virtualization, I consider more of like a CapEx based where, you know, an enterprise is going and just buying a set of servers, they’re operating, they’re running them. And then you take that and move it into private cloud. 

That’s usually by a hosting provider or somebody like me that’s providing that as a service. And then public cloud has kind of been noted as, you know, your big I guess I’d consider it maybe 4 or 5 now your AWS, your Azure, your GCP, your Oracle, your IBM. Those are bigger, larger hyperscale style projects that are a little bit different. And I think that’s, you know, really some of where we’re, we’re a little bit nuanced in kind of what repatriation, you know, means, because it’s, it’s it’s very kind of cloudy right now as far as, you know, what those actually workloads mean and what those mean to specific companies. 

Vinay: Right. So is the pendulum in a way swinging back to, you know, let’s say on prem,  and was it only a matter of time?

John White: Yeah. I mean, I think it, I think it, it is swinging back to on-prem, but I’m not sure that it’s that shocking to most people that have been in the industry. I mean, we as IT people and tech people always look for the tool to solve our problems. And public cloud was just one of those things. And people, you know, viewed it as this Nirvana place that, you know, really, it wasn’t that — it was just a different way to build infrastructure and buy infrastructure and that’s that’s it. 

And so, you know, I think what we’re seeing right now is just a right-sizing; of maybe, of a misstep of like, we sent everything into a public cloud and realized that, oh, that wasn’t, you know, really the best thing for our business and to conceptualize really the best thing for our business and to conceptualize that it’s I’m actually working on, a white paper right now. So I have this stat handy, you know, and I think we could all agree that and we read in the media, but, you know, depending on what analyst firm you look at, 30 to 40% of cloud spending is considered waste okay. Waste. 

And so that’s either maybe inefficient usage or, you know, utilizing trying to take our old monolithic application and running in public cloud, which doesn’t really work or people just over-running. But if you think about that worldwide public cloud spending for 2024 was 

600 billion. And if we′re saying 30 to 40 of 600 billion, 180 to $240 billion dollars that were maybe put in the wrong place. And so that’s where I think repatriation is going to go and clean up, is that I’m not saying public cloud is wrong or bad, but that 30 to 40%, I mean, that’s real money out there. And the people are spending that. It might just not be the best tool for them.

Vinay Joosery: Yeah. No, I mean, there are definitely expectations. And I think in a way, some of these expectations are unfulfilled. Right. So let’s, let’s go into, a little bit deeper into that. So, you know, when we talk about, you know, you mentioned this Nirvana, right? Huge expectations, lower costs go to the cloud and you lower cost, you know, why own when you can just rent, right? Improve security is another one. Improve SLA is, simplified ops, right. You don’t want that mess in your backyard. Just outsource it to somebody else, right? But this is not the case. I mean, first of all, you know what leads people to believe that it’s going to be less costly, in the public cloud, is it a matter of misjudging pricing?

John White: Well, so I should have said this in my intro, but I did an infrastructure company for 16 years, did a lot of consulting, and I spent a year at AWS. I got to sit in the belly of the beast and kind of get perspective there. I think, you know, we have to remember why public cloud was such a big thing in the beginning, and it wasn’t really a technology solution as much as it was a business solution. You know, most of the time, I mean, and remember, we don’t talk about it anymore, but Shadow IT was the biggest reason why people were jumping to the public cloud — they were sick of their departments saying, “It’s going to take you two months to get a server up and running.” And somebody was like, “That’s stupid. I could swipe a credit card on AWS and my servers up and running in, you know, a matter of minutes. Why do I need to wait?” And so the business created a problem on multiple fronts, even to think about it from the CFO perspective. 

I mean, how many times did, you know, an IT person say, oh, we need 100TB of storage, you know, for the next five years, only to use 20TB and the CFOs like, “What the heck? Why do we buy five x the amount we needed?” Well, we didn’t know if we might need it one day. And so we did it. And so we solved those two core business problems with public cloud, which made a huge impact in the industry and allowed us to move faster. And heck, we had all these new startups come; but again, those were all business problems we solved. When you look back on those business problems and it’s like, “Okay, yeah, we found that solution.” But your solution on the technology side was a different SLA than you were used to, different hardware than you were used to, different operating model. I mean, the whole pets versus cattle is still lost on a lot of people not realizing, “Hey, AWS and Azure was actually really built for this cattle thing, not pets.” 

And so it was a completely different technology underneath. But the business people, you know, got jaded with, you know, where, “Oh man, we can move faster as a business. We could reduce our, you know, capital expense.” And that’s really where the precipice of this whole thing started. And so I think really what we’re seeing right now is just a cleanup of that as a business now — we’ve learned how to deploy, and you know, launch virtual machines in a matter of seconds. We’ve learned how to, you know, clean up a lot of those challenges that we’ve had. We have all kinds of different operational models by Dell and Pure Storage and all those things that you can build them on a private, you know, private cloud on your premises. And that’s the shift that we’re seeing. Right this second.

Vinay Joosery: Yeah, yeah. Another expectation is, you know, that public cloud improves security, right. And data governance, I mean, you know, the logic is, being a provider like AWS, you know, you have all the money in the world, right, to hire the top security experts in the industry. So, but we all know that, security is, it’s a shared responsibility, right? So you can’t just wash your hands. How come that, you know, that false sense of, let’s say, confidence.

John White: Yeah. I mean, again, I think it’s an uneducated buyer. Right. I mean there are definitely elements that, you know, AWS and the shared responsibility model they take on as a security. I mean they’re saying, “Hey our facilities are gonna be protected. Our hypervisor is going to be protected.” Your, you know, your core EC2 instance is at the root, you know, from an infrastructure, their perspective, but what goes on inside the OS and the patching and the care and feeding that has never been in their vocabulary. I think it was an assumption that people made, you know, right, wrong or indifferent. 

I mean, I’ve heard it many times when I was working at AWS and when I wasn’t working at AWS. I mean, they thought it was going to be the easy button for all those things. And that’s again, just reality, of the technology, you know, really coming to light that I think we’re seeing and, you know, missed expectation from the enterprise. And how does it how do you get that expectation? Unfortunately, people had to learn the hard way that all this all of a sudden wasn’t secure. 

Now, AWS has done a fantastic job at making sure and helping customers. I mean, if you have a public S3 bucket or something like that, I think you have to like click through five warnings before are you sure you really want to do this? Then they have tools that they can go and make sure that, hey, anything that was made public, you know, make sure that it actually should be public. So I think there are a lot of positive security things that the hyperscalers brought that now are just becoming the norm. But yeah, there is no, you know, Nirvana magic bullet, kind of tech that’s out there in the public cloud that people are, you know, are getting. And I think that’s part of the realization and why maybe some people are repatriating as well.

Vinay Joosery: Yeah. Yeah. So one of the other expectations is, you know people want to wash their hands. Right. If you own you got to own it, you got to fix it. You got to do everything. Why don’t you just outsource the mess. Right. So, so when you move to public cloud, you know, you could say, “Hey, I want to simplify my ops. I want to prove my SLA is. Then I don’t have to have those teams of people around doing things,” right. But then the question is, if you look at cloud users, you know they still have ops teams to manage, you know, their fleet of, of VMs, you know, they still have, you know, people managing their security in, you know, in Amazon, they need people to, you know, even DBAs to, you know, to look after the RDS instances. 

So that’s kind of a, you know, first of all, you know, you’re you’re kind of getting rid of people because you think you’re going to go somewhere else and you’re going to outsource the problem, but you still have all these people that you need to bring in a way to make it work. And those people need to understand, right, that architecture. So you still have these ops teams, these SREs and all that stuff. So, what’s the status there? What do you see? I mean, you know, are SLAs being met or is it so that, you know, maybe it’s easier if it’s all on-prem because, you know, we as an industry, many, many, many enterprises, they have a long history of, you know, doing that. And, and they have that, let’s say, internalized, you know, knowledge internally. 

John White: Right now. I mean, it’s been a good decade. So I think we’ve made progress and a lot more people understanding public cloud. But in the beginning, I mean, that was a real problem. Operations in the enterprise were way different than operations in the public cloud. But we’ve made that gap. But I think it was really hard for a lot of people because they were sitting there and doing all the firefighting in the enterprise applications that they had, and then trying to firefight in the public cloud and not understanding it, knowing it is a completely different model. And when I first started getting into AWS and Azure, I really was thankful that I had a development background because writing infrastructure as code, as just a traditional, you know, enterprise IT guy was it would have been a challenge for me if I didn’t understand that it was a lot more code like and more development, you know, like, than it was anything that I’ve done and tweaking and tuning knobs of, you know, the days of running, you know, it servers all over the place. 

And I think that balance is starting to shift a little bit. And we have a lot of, you know, newer, younger people in the market now — the group grew up in cloud, where, you know, a decade ago that just wasn’t the case. And so I’m not sure that the operations side is as big as a gap that’s driving the repatriation, because I think we’ve kind of tightened that out. You know, a lot. I mean, there’s just so many, just people that are there certified in cloud anymore. And I think we’re, we’re even starting to move the enterprise to more of a, you know, a DevOps and, and SRE model, which makes a lot more sense, because if you think about it, I mean, 20 years ago, you know, we didn’t have the applications that needed to run 24×7 as much as we do today. 

Like right now, that’s a norm. I mean, my kids will never understand the concept that something only was supposed to happen at a certain time. We were on a trip a few months ago, for a few weeks, and my son was like, no, I want to watch this. It’s like, no, you have to watch what’s on TV. And so think about that concept in that mindset of like having to wait for something — that’s kind of been forced out of us. Now in an enterprise and a hyperscale. So it’s just the norm. You have to be running 24×7. There is no such thing as downtime anymore or scheduled, you know, change or an outage. So I think we’ve tightened up on that. I don’t know that that’s the big thing that’s pushing repatriation in my opinion.

Vinay Joosery: No, no. Yeah. I mean, you know, talking about linear TV versus streaming, I think, you know, the younger generation, I think they’ve…

John White: Yeah. That’s a difficult thing to understand how it was before and how it is now. And we were like that. I mean, think about it. 20 some years ago a whole website was down.

Vinay Joosery: Yeah. You just came back later.

John White: But like that concept for, for this younger generation is just not there. It’s, they just want it now. And so every application has to be run like a hyperscale application essentially.

Vinay Joosery: Yeah. Yeah. When we started off, you know, when I started off in the database industry, that was back in the Ericsson days, you know, big telco and they had they had five nines there, you know, you needed to have dial tone, you know, all the time. Right? Because even like, these 911 services depended on it. Right. But, and at that time, I remember selling databases outside of telecoms, you know, you would have like, you know, downtime.

John White: Yeah. Yeah. Downtime.

Vinay Joosery: We mean, that’s unplanned downtime because we have planned downtime, right? You have four hours every month where you actually upgrade stuff, but nowadays you’re expected to upgrade stuff while things are running in production. So. Yeah. But so coming back a little bit, you know, looking at all the issues we just discussed, perhaps the, the TCO, you know, considerations, they are perhaps the most glaring, you know, because so comparing an on-prem hybrid CapEx kind of, you know, model versus a cloud OpEx model, when does one make sense over the other?

John White: I heard an analyst say this phrase, I don’t know, probably a good six years ago that, public cloud is, if you’re going into the analogy, I’m butchering it already. But the analogy of of getting a dog, receiving a dog, you want to train your dog in the public cloud because if it makes a mess, you can throw it away and it’s good to go. But once you get to your steady state and the dog is not a puppy anymore and it’s, you know, a good boy, it’s, it’s better to run it in the private cloud. And to take that analogy, to like a company like Dropbox, who was very famous in saying this. And then, you know, 37signals was out there, talking about this as well. Yeah, you could do some training in the public cloud, but at the end of the day, it’s better for your business to capitalize it, own the hardware and own the full stack, own the update cycle, kind of own everything. 

And I don’t think that that’s the, you know, again, Nirvana that we’re all looking for. But I think it’s just a balance between the two. And that’s why I’ve been bullish of the concept multi-cloud for, you know, pretty much my life is that, you know, you got to pick the right tool that’s going to, you know, go and and run your business, and do it efficiently and secure it, you know, with the top security and you know, your, your consumers, both your, you know, your customers as well as then the people internally need to be happy with it. And so you look at like big storage right now, if you know that you’re going to run a petabyte of storage, it’s going to be, you know, and you’re going to run that petabyte for the next five years, it’s going to be better for your business to capitalize that and go and put it in your data center or a shared data center or something like that. And, you know, that makes sense. If you’re going to go and run an application that’s going to be heavy. Burstable. And it it’s, it’s all built on microservices. And so you can have scaling engines and everything behind it. It’s probably better fit inside of the public cloud until you figure out exactly what you need from a steady state standpoint. 

And I think that’s a lot of the the work that we’re trying to help our customers, you know, go through right now is help them figure out which cloud is actually the best for their application, and that’s a story that everybody should be thinking about and they should be doing ROI and TCO analysis. But if you’re going to, you know, the general rule of thumb — if you’re going to keep something for a few years, yeah, you probably want to figure out how to reduce that expense. And then, you know, put it into more of a steady state environment and go, but if you’re going to burst…yeah, I mean, I’ve worked with a lot of customers over the years that, you know, had to deal with Black Friday or deal with back to school shopping, you know, big advertisements that would come out that would just push a ton of traffic, perfect for the public cloud. 

You know, and as you as you look at, like AI, you know, I think that’s the one that’s, that’s kind of busting everybody up right now is like, oh, what do I do? Should I go to a public cloud environment? Should I go and build my own? It’s probably a combination of both. Go figure out and test the different models that are out there and the different tools that are out there. Keep your data private, you know, run some sort of scripted data out there in the public cloud. But then when you figure out how to, you know, actually, you know, own it and you know, you know what you need, then use it on your private cloud data. That’s probably the best of both worlds. Make sure you have the security protection, because I think there’s a lot of unknowns out there right now. 

And, you know, as people start to think about latency in the future, I think that’s going to be a big challenge that people are going to have, that you’re going to probably want to have, you know, maybe AI close to you where your data is actually being generated. If you’re a manufacturing or healthcare company like that. So you’re going to have to go, you know, all these different things are going to be thrown at you as a, as an IT leader. And so if you’re not doing the ROI and TCO analysis right now, I think you’re going to run into the same issues that we saw with public cloud ten years ago.

Vinay Joosery: Yeah, yeah. And so and you know, how is on-prem or hybrid more cost predictable?

John White: I think it’s only cost predictable if your workload’s predictable and you know that, “Hey I’m going to grow this data set at X percentage year over year,” then you can go and you know, and if you have cash, you can go and buy the hardware devices. You need to run that application. If you have no idea what’s going to happen, and you’re not sure…yeah, you moved to Public Cloud, but the problem with that and what happens there is just like it’s unpredictable, you know, just like anything unpredictable, is your expense could get out of control. 

And that’s like one thing you can’t really walk back. And if you have users just consuming whatever, but you don’t have the money to pay for it, that could be a challenge for them. So sometimes putting people in a box and saying, “Hey, we bought X amount of hardware, we have X amount of storage, this is what you get to play with.” And that’s it. I mean, that might help some customers get rid of that unpredictability that, could blow out their budget in one month.

Vinay Joosery: Yeah, yeah. I mean, that’s the you know, I guess convenience is what public cloud is, is famous for. But definitely, it’s like giving a credit card to all your employees in a way.

John White: Yeah, before, like, giving my kids a credit card at the, the concession stand, you know, like, they’ll go nuts. I mean, they’ll buy every ice cream and popcorn and nachos that they can see. But if I go and say, “Hey, look, these are your nachos, and that’s it,” they’ll be happy with it and they’ll make do. But I think that’s the balance you have to strike because you don’t want to get Shadow IT going again, which has people, you know, swiping a credit card and expensing it because they’re frustrated. There’s just something in the middle. And you know back to your point of the pendulum shift. And I think yeah might be because I think we went too far public cloud. And we really need to end up in the middle.

Vinay Joosery: Yeah. Yeah. So beyond dollars and cents right. What value can enterprises, you know, realize in running their own infra? I mean are we talking, you know, reduced latency, better performance for data heavy workloads? I think you mentioned something earlier about maybe, you know, bringing the AI to where your data is?

John White: I think. Yeah, all those are definitely really big topics that I’m talking about right now — the latency is probably one of the largest. So US Signal just kind of give it background on us. And we provide fiber data centers. And then one step adjacencies like cloud and security and those types of things. And so we are working heavy right now, we’re building, we have about 700 miles of fiber being built right now. And a lot of that has to do with making sure that we are ready for inference, which is going to be focused heavy on latency. 

And if you think about it, I mean, when the internet was created, I mean, we were really, you know, again, we talked, you know, just about availability. But as users, I mean, we, we deal with 100 and 200 milliseconds of latency, no issues. But when you get machines talking to each other, they don’t want to wait. They want to make a decision and move. But if you look at some of the automation happening, especially in the manufacturing space or logistics spaces, like you have to have access to that data, you have to have insights coming from that data real time. You can’t take that application and move it into a public cloud and say, “Hey, you got 20 milliseconds of roundtrip latency to deal with.” You know, the machines aren’t going to want to deal with it. It’s going to compound and create issues for us, maybe even safety issues for us in the future. 

And that’s where I think we’re going to see a new renaissance of private cloud occurring, just because of the latency applications that you’re going to need when you start moving everything into more of this inference stage of AI that we’re going heading to over the next few years.

Vinay Joosery: Yeah. So now let’s look a little bit at you know let’s say let’s say some of the challenges right when it comes to doing the repatriation. I mean even if there is a strong, you know, motivation, let’s say to, you know, to repatriate some of your workloads, I mean, you know, some companies that aren’t ready for the day to day tasks of running on-prem. Right. How do you help bridge that gap?

John White: Well, I don’t know that everything has to be on-prem. And I again, as technologists, I think we always create these divides of it has to be either or. You, I mean, I think, you know, one of the services we offer is co-location. And so a customer can come in, use our data center for the power, the cooling and the internet, and it’s close to where they want it to be. And they can capitalize their hardware. Or they could lease their hardware from us, or they could, you know, buy it in a host, a service. I think there’s all that intermediary that happens in between. 

So if you have public cloud over here and you have on-prem, you know, customer owns the data center, you have private cloud people like, like your US Signal, that are infrastructure providers that are kind of sitting in the middle. And as an IT organization, you should be thinking about how you’re doing all three, because there is still stuff that you’re going to want to keep on-prem or you want to own. I mean, we still see a lot of phone systems out there in the wild that people are running on-prem. They need to be close to their consumers. And so I mean, that works. And then you still have your on-prem where you need to have virtual servers to be there for development or whatever you’re doing. And so that makes sense. 

And then this private cloud infrastructure provider like you’re a Signal, I mean, we sit in the middle to kind of provide the best of both worlds. And then you have your public cloud providers out there and you should be thinking about multi. You should be thinking about hybrid. One of the strategies we launched last year was a term called Operations Anywhere. And it’s a strategy for us to say as a provider of services that we will provide ubiquitous service no matter what the infrastructure is. So if you want to buy, you know, backups or data recovery or data protection or, or disaster recovery from US Signal, you can do that on-prem. You could do that inside of our four walls, obviously, and you can do that in the public cloud. And I think that’s what a lot of people need to adopt as their strategy is that, “Hey, we need to then make sure that we’re using the correct infrastructure, the correct location, and then we’ll run our applications ubiquitously no matter where it lives.”

Vinay Joosery: Right. Right. Do you see any, you know, common practical hurdles that organizations face, right, when moving, you know, workloads into like a multi-cloud or a, you know, hybrid model?

John White: I think it’s going to add some more complexity to it. And you’re always going to have people that, you know, they’re going to be zealots, they’re going to say, “No, this is the best way to do it.” And then you’re going to get, you know, a group of people that are like, you know, “I’m, I’m good with anything.” And so you’re going to have those battles of fiefdoms inside of an organization. And I think, you know, the sooner you can build, you know, one of the things I built at AWS was like Cloud Center of Excellence, which is getting basically everybody to agree on how you should adopt cloud. That was the big value that I saw out of it. 

So as soon as a CIO or CTO or whoever’s kind of leading the charge can help build that and get people to be more open, I think that’s going to help everybody. And I think when you can pair it together with the CFO conversation, as far as making sure that we’re keeping our finances tight to where it allows us to innovate and the CEO conversation or this chief strategy, operations conversation that says, “Hey, we need to do better for our customers.” Well, you can bring all those people into the mix to kind of have their say. You’re going to get to more of a flexible, more pliable technology person, which is maybe an oxymoron, but something that I think people need to do to have success in the future.

Vinay Joosery: Yeah, yeah. So one more concern is, you know, when you see all these, you know, Amazon or, you know, Google cloud conferences, they just added 20 more services or 30 more services managed, automated, you know. I mean, how do organizations overcome that? Because when you go to a large hyperscaler, you have this menu of, you know, all these services you could use. And then when you consider, you know, something else, it’s, it’s quite hard because then it means you, you probably have to do a lot of this. You know, yourself.

John White: Yeah. It’s always the conversation of, do you want to buy into an ecosystem? If you want to buy into an ecosystem completely and you’re just generating a new application and it makes sense for you to use more serverless, style applications. Yeah, it’s great thing to do it. But just like you said, you got to have somebody or multiple people full time just understanding what’s being released and how you can use it. I think, you know, whether it’s a good secret or not. Most of AWS, I would say, you know, most AWS users are using EC2 and that’s it. Yeah. You have some people dabbling in EKS or ECS or Beanstalk, but that’s like the 20 or 30% of the spend that’s occurring. Most everything else is still EC2 instances underneath the covers. 

And that’s, you know, that’s the easy button. And so I think there’s going to be some right-sizing of all of those services eventually. And I don’t think that’s going to happen because I’m not sure AWS and Azure and GCP and everybody, they even know what services they need. You know, I think they build some things a lot of times that come out because they’re using them internally, whether it’s Microsoft building something for O365 or LinkedIn or, you know, some of their bigger customers, and same with AWS. And then you see GCP, actually GCP probably has the best lifecycle management of all of the services that come out that, you know, they get knocked on all the time. But yeah, they release a product and they have very little uptick. 

It’s like, why should they go and spend all this money on R&D and operations and management? Let’s just kill it and move on. And I think, you know, we’ve seen a lot of that happen. And I think that’s where, you know, we just have to have a maturity, in the market still. But it’s crazy to think how much time flies. I mean, again, we’re so what hyperscale we want to put a label on in 2012. I mean, we’re 13 years into this cycle. Think about end user-computing. I mean, end-user computing. We’re in 30 some years of 30, 40 years of running end-user computing. And it’s still, you know, the wild, wild West. 

And so we’re still very early innings in cloud and AI and all these other things that, you know, a typical life cycle. We just ​​haven’t seen it happen yet. And I do expect there to be a lot of right-sizing and the public cloud with all these other, you know, periphery services that are going on at the core. Most customers just need to run a virtual machine today, or they’re going to build something off of a virtual machine stack. And that’s really what they’re doing really well already. People just need to understand that, “Hey, that’s okay, that I’m just running virtual machines.”

Vinay Joosery: You mentioned AI, you know, a topic that you can’t really get from. I mean, we’re in the midst of a, you know, all this AI boom, right? So many of these workloads, they need specialized hardware, right? Like GPUs. How does that factor into this, you know, let’s say repatriation decision?

John White: It’s, I think if I had those answers, I have some ideas. I’ll give you my ideas. But if I had all the definite answers, I think I’d be a pretty rich guy. Because I would know what stocks to buy and who to back. I, you know, my my, my ideas and concepts on this are very similar to, like, public cloud in the early days. Like, you’ve got to go and play and once you play, then you go and you build your, your, your permanent structure, your foundation. And so I think people right now are definitely all-in on creating LLMs to, you know, help them answer questions. And it’s crazy. 

I mean, think about search engines. I mean, we had search engines forever and Google was the, you know, the key search engine forever. But I’m finding I’m asking questions and looking for resources out of my LLMs more than I am a search engine now. But then you can see how they have it interacting together. When you do a Google search, you also get Gemini responses. Because it was a disruptor. And I think we’re going to see that over and over and over again. I do believe that companies are taking more precautions this time, and they’re starting to buy more hardware to then make sure that they don’t lose control of what I can provide them. 

Now, I think, you know, we’re going to get some disruption in the processing space. Obviously. I mean, that’s going to be, you know, that’s always going to continue to happen. I mean, the whole, you know, DeepSeek thing, whether you know, what side of your fence you’re on; whether you believe or not, the numbers, I mean, what did happen is there were models that were trained off of other models. And I think that’s going to continue to occur. I mean, you get the snowball rolling down the hill. You don’t keep throwing new snowballs out there. 

You build off of that single snowball. And I think that’s going to be, you know, where we focus, you know, a lot of efforts as IT folks in the future is looking at what the people are building and then making it bigger and better. And if you can privatize that with, specific hardware…I mean, look at all the new processor companies that are being, you know, released right now. There’s a lot of new players that are coming in to, to form, I mean, CoreWeave launched their IPO. You have companies like Groq, g.r.o.q., that have processors that are only available in a serviceable model. You have, you know, small people building SLMs that can run on, you know, my Mac, for instance. And so we’re going to see that continual evolution, where people I think now are going to make sure that they’re more secure, which is going to probably lead to a resurgence of private, which is going to make more sense for our repatriation long-term then.

Vinay Joosery: So let’s say five years down the road, right. How do you see workload placement, you know, evolving, I guess, you know, cloud is still big, on-prem has always been big. Where do we see the balance in the next five years?

John White: It’s that latency. Latency is going to drive. Latency I think is going to drive everything. Latency is going to become the new currency for the future — you’re going to have to have reduced latency, to make sure that your application can give you the insights in a timely manner. If you think about this from like a B2C basis, you know, you as a consumer, you know, running and asking AI for an insight at your house. If it starts to wait for, you know, seconds or minutes, you’re going to get frustrated or you’re going to move on. And so you’re going to have to have whatever AI tool be local to where you’re actually located. 

So hanging off of fiber that’s maybe coming into your house and into these small little mini data centers, and if you think about it, it’s not really a new concept. I mean, how if you’ve ever deep dived into how Netflix works, they have these distribution pops all over the world. They get closer to the end consumer so that when you click, “I want to watch this video,” which is going to be Drive to Survive for me soon. The new season came out. So they know that “Hey, John wants to watch that. And this is local to John and I’m gonna make sure I stream from it somewhere local.” 

We’re going to see a huge push in that where Netflix has all these mini data centers all over the world, every application, every, you know, big CDN is going to follow suit. And so we’re going to see distributed data centers occurring at a pretty rapid rate. And I think that’s where you’re seeing the hyperscalers build all these new data centers. But if you think about it, I mean, like AWS, I mean, they had a few different regions inside of the United States, and now they’re talking about building data centers more, you know, more places. And that’s just going to be a huge drive for us in the future. So whether you call that private or whether you call that, you know, local, that’s going to be a big push for us and that’s going to be, you know…just I think the number one thing that drives applications to be in a multi- or hybrid state is because you can’t just live in one, single ecosystem anymore.

Vinay Joosery: Yeah. Yeah. So I take a note here. Physics, speed of light. That’s going to be you can’t you can’t change it. You can’t beat physics. We’ve tried a million times, but, at the end of the day, I mean, that that speed of light is there.

John White: Yeah.

Vinay Joosery: All right, all right, well, time to wrap up. So, John, you know, this has been an enlightening discussion on repatriation. From the reasons, you know, driving it to the blockers and the solutions. You know, we’ve spoken about TCO, we’ve spoken about AI’s influence. So thank you for joining us. And that’s it for today, folks. Thank you all for listening.

John White: Yeah, thanks for having me.

Guest-at-a-Glance

Name: John White
What he does: Chief Operating Officer
Website: US Signal
Noteworthy: John White is the Chief Operation Officer at US Signal. He is an accomplished leader with a proven track record of driving organizational growth, operational excellence, technology transformation, and innovation. Previously, John was the Chief Innovation Officer at Effectual and Expedient, where he spearheaded significant transformations, optimized technology, and developed strategic partnerships that resulted in increased profitability and customer satisfaction.
You can find John White on LinkedIn