The Major Decision Hiding Behind Every Hypervisor Renewal

A wave of licensing and ownership changes across the virtualization market is forcing enterprises to revisit infrastructure decisions on a timeline they didn't choose, and the toughest question isn't which hypervisor to pick next, it's what operating model sits underneath it.

Patrick Moorhead, Founder, CEO, and Chief Analyst at Moor Insights & Strategy, and Daniel Newman, CEO and Chief Analyst at Futurum, are joined by Varma Kunaparaju, SVP and GM of Cloud Platform and OpsRamp at HPE, to kick off the HPE Virtualization series, unpacking what a private cloud and AI factory operating model actually needs to deliver on real enterprise choice.

For CXOs, that shift means treating modernization as a phased, multi-year operating model decision rather than a one-time migration project. Planning requires understanding the current infrastructure estate first, organizing the data that feeds AI-native workloads, then extending governance so private and frontier models can be routed safely. It also means drawing a deliberate line between what infrastructure operations can act on autonomously and what still requires human approval before a disruptive change goes through.

Key Takeaways:

🔹 Virtualization licensing changes are forcing a decision enterprises didn't plan to make yet, but the real question sits one layer above the hypervisor. Choice across infrastructure and cloud creates value only when the operating experience underneath stays simple, governed, and consistent.

🔹 Modernization works best as a phased operating model, not a rip-and-replace migration. Kunaparaju outlined a sequence: assess the current infrastructure estate, organize the data feeding AI-native workloads, then layer in provisioning, full-stack observability, and data protection.

🔹 AI-native applications are forcing a new cloud operating model that spans edge, data center, and public cloud. Workloads increasingly combine local models and local data with routing to frontier intelligence, requiring infrastructure that behaves consistently across all those environments.

🔹 The line between automation and human-in-the-loop is shifting operation by operation, not all at once. Infrastructure signals that can be remediated autonomously and safely are moving out of the loop, while disruptive changes still require human approval, mirroring how AI harnesses already ask permission before acting.

🔹 Cost optimization now spans both infrastructure footprint and token economics. CxOs are balancing on-prem and cloud footprint decisions against the cost of routing workloads between open-weight and frontier models, especially as frontier-level intelligence keeps showing up in open weights within weeks rather than months.

Enterprises that treat this as an operating model decision, not just a licensing renewal, will de-risk their AI factory buildout and move workloads into production faster.

Learn how enterprises are reducing virtualization complexity and cost with HPE CloudOps Software. 

Watch the full video at sixfivemedia.com, and subscribe to our YouTube channel so you never miss an episode.

Explore the full series at www.sixfivemedia.com/our-series/hpe-virtualization-series.

Disclaimer: Six Five Media is for information and entertainment purposes only. Over the course of this video, we may discuss companies that are publicly traded, and we may reference their equity share prices. Nothing discussed during this webcast should be considered investment advice or a recommendation to buy or sell any security. We are not investment advisors, and you should not rely on this content as financial advice. Six Five Media collaborates with technology companies and industry leaders to produce research-driven interviews and multimedia programming for enterprise technology audiences.

Varma Kunaparaju:

The way to kind of think about is not ripping and completely replacing, but you kind of take the approach of doing a complete understanding and the visibility of the end business users and business application needs and taking those business application needs and looking at the entire estate of infrastructure and applications and take it in phases.

‍

Patrick Moorhead: 

Hey everybody, welcome to the Six Five podcast. I'm Patrick Moorhead. I'm joined by my partner in crime and bestie, Daniel Newman. How you doing my bestie? Hey, always great to be with you.

‍

Daniel Newman: 

Love the virtual webcast. Always love talking tech, Pat. I have a funny feeling we're talking a little bit, I don't know, AI today, but probably some other things. And by the way, You know, a lot's been going on lately. There's a lot going on in tech.

‍

Patrick Moorhead: 

I mean, there really is. And, you know, some things that, hey, we've been kicking around like three years ago on how do we enable enterprises to scale, right? Seem to be similar conversations. We're talking known good output. We're talking about governance and around the data. And we're talking about costs and open and small models when combined with frontier. And as part of that, I think we also always agree is to get to make all these amazing changes inside of the enterprise, you really do have to modernize what you do, setting up your, we'll call it, you know, an AI factory operations. You cannot do that on today's software and today's infrastructure. So it does require an upgrade. So I would like to welcome a many-time guest back to The Six Five. Varma, great to see you, my friend.

‍

Varma Kunaparaju: 

Thank you. Great to kind of reconnect here and thanks for having me here.

‍

Daniel Newman: 

Yeah, Varma's what a three, four, maybe five time. By the way, so many great conversations. A builder that built OpsRamp and then was acquired in by HPE and has been doing a lot helping lead their software business. Pat and I as builders love talking to others that are doing the thing. You know, I think, Varma, you sort of heard us kick off. You know, we'll talk about enterprise. You know, there's so much going on there, right? You know, in the AI world, so many people think about, oh, I just subscribe to Claude and start prompting things. But that's nothing like what an enterprise actually looks like, right? You know, they're having to think about cost. They have to think about control. They have to think about choice, how their private, you know, infrastructure looks like, the complexity of software and licensing. You know, I guess you know, with everything that's happening and the change accelerating, these decisions are being forced to be made faster. I guess your view from being out there talking to the customers, like how should CXOs be thinking about, you know, and evaluating their operating models right now with all the change that they're trying to endure?

‍

Varma Kunaparaju: 

Again, you know, it's a very, very interesting time of IT in the enterprise today. You know, in the good old times before the AI era, the CXOs and the VP of infrastructure and the CIOs and the CIOs all focused on one thing, running the IT infrastructure. But today, that game completely changed with AI and the data that is driving the AI and the compliance, the governance. and the ability to leverage enterprise data for agentic enterprise needs today is an order of magnitude different than four years ago, let alone, I'm not talking about a decade, just three, four years ago. So, when enterprises and the CIOs are looking at moving their organizations to take advantage of agentic and AI, and AI-first applications, it is not just infrastructure operations anymore. It is just agentic enterprise operating model that they are really looking at. What applications, what data that those applications are potentially going to take advantage of AI, and how to make that entire modernization of application stacks and how they consume the LLMs and the harnesses to leverage both internal private data, powering the open weights models, and at the same time, potentially leveraging frontier and frontier intelligence coming from external. So there is a lot that is going on from both infrastructure operations, as well as the data that is contributing to the applications. that they need to modernize for the business units.

‍

Patrick Moorhead: 

Totally get it. That's what we're hearing as well. It is interesting, Varma, you know, there really is nothing new. I mean, we can call it something different, but if I look at the last 35 or 40 years of IT, really nothing gets replaced. We're just adding on different layers and we might lean in and there's higher levels of investment for something, but for the most part, we still have a lot of the same technology that, quite frankly, we used 40 years ago. So when you're talking about modernization, there are different ways to pace this. But I think unlike the past, people are more looking at, hey, how do I not get fragmented here? So if you look at the different teams who are trying to modernize, you might have an AI team who's like, go, go, go, board of directors saying, I want my outcomes now. How do we do this? Their response is typically, hey, I'm going to come up with some sort of very proprietary, can't be shared type of thing. And then you might have somebody who might be on the VMware shot clock of three years that's like, oh my gosh, I've got this. I've got a hit, I have to figure out what we're gonna do here. And then there's people who might be like, hey, I can't run this application you want. It's not necessarily AI, but maybe it's gotta go from virtualization to bare metal. So all of these different things going on. So how should a tech leader sequence that work without just creating an even bigger mess?

‍

Varma Kunaparaju: 

It's a great question, but you know if you look at any time when a decision maker is forced to do some choices and decisions, it's always opens up an opportunity to look at the problem differently right now it's just. not tripping and replacing or migrating from one place to another place. But whenever you are asked and whenever you are forced to do something like that, it gives an opportunity to look at the problem differently. And I think, you know, in the context of what's going on in the virtualization, when enterprises are looking at Just to give an example, that gives an opportunity to modernize the applications because most AI applications are cloud-native and most cloud-native applications require a different runtime of choice that they need to look. So, this whole opportunity that is presented to an enterprise saying that, you know, I need to look at what my new modern application and how that needs to run, that gives an opportunity for them to look at their runtime look at their opportunity and how I operate that runtime and the application workload, because the application workload is no longer just in one single place. It's probably a combination of what they're running in with, like we touched upon in the previous conversation around local models and local data contributing to those applications with external frontier. So how do I create a new modern operating model, and we call it cloud operating model, where it is multi-vendor, it is multi-cloud, and the infrastructure that is participating into that environment is from edge to cloud. It's not just central in your data center, it's not central in public cloud. So, if you look at that as state, it is an opportunity for every IT decision maker to look at this problem in a slightly different way by creating an operating model and bringing those silos and giving an opportunity to make those silos to really work together as part of the modernization.

‍

Daniel Newman: 

Yeah, by the way, I don't know how you guys feel. I feel like everything we're doing now is sort of a repeat of the cloud era for the AI era. It's just faster, meaning, you know, the amount of time it's like, you know, we just kind of, I like to say all gas, no break. It's just like kind of all gas, no break right now. It's like you went through these same sort of phases of bringing teams together, fragmenting the businesses versus the IT teams versus the, you know, obviously the, the numbers people and the investments people, of course, security would come sprinkling in at some point, and you'd have to get them involved. And we're just doing it now. But it's like hyperspeed, it just the decisions are being made so much quicker. But Varma, I mean, look, these types of migrations, these types of transformation, real risk, You know, and the real risk sometimes create friction, slows down the process. So we're talking about hyperspeed, but at the same time, things get slowed down, outcomes get, you know, slowed from realizing their potential. The proforma never comes to fruition. What do you think is the lowest risk? Meaning that when you're moving to this private cloud operating model, usually it's budget-driven, security-driven, governance-driven, sovereignty in some cases driven, but how do you do this and how do you genuinely mitigate as much risk as possible in that process?

‍

Varma Kunaparaju: 

Yeah, I think, Dan, you know, the way to kind of think about is not gripping and completely replacing but you kind of take the approach of a phased, you know, multi phased approach you know first of all doing a complete understanding and the visibility of The end business users and business application needs and taking those business application needs and looking at the, the entire state of infrastructure and applications and take it in phases and the first phase potentially. is to really get an understanding of the current infrastructure and the state of the current infrastructure and what immediately needs to be modernized. Maybe some of them potentially require application modernization. Some of that potentially requires the workloads and the data that is feeding to those workloads can potentially make in AI and AI native application to really leverage. So that really requires organizing the data in a phase two. And in a phase three, how does an actual enterprise leverage that security governance that you said, right? Where internal data is feeding the local models to give the intelligence that you are not asking those questions and queries and prompts to go out to the to the outside the enterprise right from a compliance and governance perspective. So my view here is looking at this in phases and putting a framework. a operating model framework from day minus zero, where I assess and understand my entire infrastructure, day zero, day one, in terms of putting a runtime and provisioning and orchestrating that runtime, whatever an application developer needs for being able to do their cloud operating model, almost like operating in a public cloud, but you're operating in a hybrid infrastructure. That's the day one provisioning and orchestration. Then full stack observability, where you understand the entire application footprint and the infrastructure availability from both infrastructure point of view and application point of view, and looking at all the signals And observability is no longer just application availability observability anymore. It is also contributing to the security side because those signals, and those, those data elements that you gathered from metrics logs and traces will also give some visibility into AI. and AI native needs and compliance and governance. So that is day two. So now if you kind of combine all of that, along with the protection that you need to do, data protection and resilience, that's where you kind of really bring the full cycle to create an autonomous enterprise for IT operations and making applications delivered to the lines of business in a way that they're all agentic in the future. So that's the modernization that is needed for bringing an end-to-end solution. In our terminology, we call that cloud operating model, and that's where we put our effort to create a cloud ops suite.

‍

Patrick Moorhead: 

Yeah, so you're getting everything done, you're lowering your risk along the way. What I want to shift to is autonomous. I've seen a lot of marketing messaging over the years and branding that talks about the autonomous enterprise. And to some case, there are some things that we have farmed out particularly inside of security, right? Which is, hey, all these signals come in, how do you parse the ones that really matter? How do you prioritize? And then act in a way, shut down a certain connection or a certain user to protect the enterprise. But I'm curious though, where is the line between this automation and human-in-the-loop? Is it as simple as saying, okay, we're going to test some stuff that's typically been human-in-the-loop, or over time, we're going to compare what the human did versus what the machine thinks it should do, and then we're going to push this lever. How does that look in practice?

‍

Varma Kunaparaju: 

Pat, if you look at in the last quarter or so, there is a lot of talk around RSI and how the models are getting more and more RSI-influenced. The same thing is happening in the world of what an enterprise's journey in terms of making agent-driven, but it at the same time human in the loop, right? It is not a total binary. There are operations. Take an example of infrastructure operations. There's a lot of infrastructure operations when the signal clearly specifies that this operation can be autonomously and intelligently remediated without a human in the loop, those things automatically are fully qualified to shift from a human in the loop to not human in the loop. But at the same time, there are decisions that one may have to really make sure that there is a human in the loop to approve or to validate before a disruptive change that an agent wants to take. And you see that in the harnesses asking end user a question, hey, I am supposed to do, I'm planning to do this, do you want me to kind of go and execute? Same thing is happening. So now what frontier companies did with respect to harnesses, imagine that applying those harnesses to every business application, including infrastructure operations. That's where, in my mind, the autonomous operations or autonomous agents being able to help a workflow in a business application. In my view, where we are heading to is harnesses are not just for developers. harnesses are going to take place visibly or invisibly for every business application and every infrastructure operations one needs to do.

‍

Daniel Newman: 

In the end though, right, dollars and cents comes down to the money. You've been around this part of building the observability was also people could understand what their systems were doing and probably in many ways it was then how to understand what they should and shouldn't be spending on. Well, You know, when you're going through modernization, it's often a licensing decision. It's often, you know, very financially driven. And anyone that sort of tracks IT, they know that, you know, the CIOs, CTOs, heads of infrastructure, they have to build business cases, Varma. They have to build, you know, they have to make it make sense to people that don't know and don't understand fully how the tech works and how competition works in different platforms and comparatively. What do you recommend for CXOs right now that are trying to build the business case? for that modernization, that augmentation that you talked about to a private cloud operating model.

‍

Varma Kunaparaju: 

So Dan, I'm going to address it in two buckets. One bucket is, or one category is, the infrastructure side for a CXO, right? And then the second category is the applications and the costs associated in delivering those applications in the context of AI to end business users. So if you look at both of them, on the infrastructure side, there is optimizations that every enterprise is looking on their cloud footprint, and their on-prem footprint, and making sure that footprint of not one against the other, but how they balance that. Because in the context of data and things like that, that's one piece. And the second piece is with AI, the token costs and the amount of money that an enterprise needs to do. Both cost and security come in at the same time, where You know, can I optimize the cost that is involved in making some of the tokens that I need to deliver? You know, four months ago, frontier models is today's open weights. And I believe today's frontier model intelligence will come into open weights, maybe, you know, a quarter, a quarter and a half later.

‍

Daniel Newman: 

Sometimes, by the way, it's days. We actually test this. It's actually, you know, so you're right. I mean, it used to be months and now it's days to weeks.

‍

Varma Kunaparaju: 

Yeah. And in some cases, some of those tokens, not only for cost reasons, but also from a data privacy and governance reasons, you may have to optimize. That's why some of the private clouds for AI and AI needs is where enterprises are looking at, how do I make my applications potentially run in a cost-effective manner inside my data centers or inside where my data resides today, and then complement that with what needs to be done with frontier intelligence when LLM routing will take place on either to go to open rates or to the frontier model. So the cost optimization is twofold, infrastructure cost optimization, and then ultimately resulting in overall AI agentic journey of an enterprise's application and the cost to deliver those agentic needs for a business, how do I cost optimize? So both these spectrums are taking at their own speed and velocity. And in both cases, there are well and reasonably proven. When an enterprise is taking from a proof of concept to a production application, The production application needs to be taken into consideration in terms of optimizing the cost as well. Because proof of concept you may do in a small capacity, but when you are applying for a large enterprise workflow or a business need, there's a lot of cost involved. And that's where the optimization from both modernizing, but also taking the journey to make an application delivered with security and governance in a cost optimal manner is where the CXOs are spending their times today.

‍

Daniel Newman: 

Pat, it sounds like we should be building a comprehensive benchmark with an ROI calculator that all the CIOs can use.

‍

Patrick Moorhead: 

That's an amazing idea, Daniel.

‍

Varma Kunaparaju: 

That's a great idea. You know, hey, you know, if you run this workload, you know, with this footprint, what would it cost like, as opposed to this footprint?

‍

Daniel Newman: 

I'm thinking about it. We'll have to take this conversation offline. But that's, there is, again, because it all comes down to numbers. In the end, like, you know, obviously, presuming the product, the quality, you know, the security and governance, all those things are available. In the token era, the math has become incredibly measurable. It's more than ever before that people can measure outcomes and it's really exciting times. Varma, this is a great conversation and we should keep having it. No doubt that much like Cloud Arrow 1, we are going to increasingly see people start things and test things in public APIs and increasingly recognize the value of hybrid type of infrastructure, private infrastructure. Software matters a lot. And like you just said, the economics matter, the tokens matter. Congratulations on all the progress. Let's have you back again soon. And thanks for joining us here on The Six Five today.

‍

Varma Kunaparaju: 

Thank you. And I think it's a great, strong case for enterprises to look in their journey to modernize in a low risk, more consistent and give a good cloud operating model and AI in the context of AI. And thanks for having me and look forward to continuing our conversations.

‍

Daniel Newman: 

Yeah, we'll see you soon Varmon. Thanks everybody else for being part of this Six Five virtual webcast. Great to hear a little bit about, you know, modernization, private workloads, private cloud, operating models, all of those things. Varmon is always a great guest. Subscribe, be part of our community. Check out all the content here with HPE. And then of course, all the content across our Six Five Media network. For this one though, we got to say goodbye. We'll see you all later.

‍