VMware Explore 2026 Wrap-Up: What’s Next for Private Cloud and Enterprise AI

Patrick and Daniel break down what VMware’s direction signals for the evolution of private cloud, with particular attention to AI infrastructure, agentic applications, security, governance, and enterprise economics. Drawing on their conversations with Ram Velaga, Paul Turner, and Purnima Padmanabhan, they discuss the announcements and themes that stood out at Explore and what they mean for organizations balancing AI ambition with requirements around control, cost, data sovereignty, security, and operational resilience.

They also look beyond the technology itself to the broader economics of enterprise AI. As organizations move toward more persistent and autonomous AI workloads, infrastructure decisions increasingly become strategic decisions about where workloads run, who controls the data and AI stack, how agents are governed, and how enterprises capture the resulting productivity gains.

Key Takeaways:

🔹 Private cloud is being repositioned for the AI era. The conversation explores how private infrastructure can evolve beyond traditional virtualization and become an important platform for enterprise AI workloads.

🔹 The challenge is no longer simply deploying AI—it is putting AI into production. Enterprises need infrastructure that can support production-scale AI while meeting requirements for security, governance, reliability, and operational control.

🔹 AI agents raise the infrastructure stakes. As applications become more autonomous and agentic, enterprises need stronger controls around identity, access, data, security, and workload governance.

🔹 Control and flexibility remain major private-cloud advantages. For organizations with sensitive data, regulatory requirements, or complex operational environments, infrastructure control can be as important as raw compute performance.

🔹 Enterprise AI economics matter. The discussion examines the trade-offs between cloud consumption, on-premises investment, utilization, infrastructure efficiency, and the potential ROI of bringing AI workloads into production.

🔹 VMware’s direction reflects a broader infrastructure shift. The private cloud conversation is increasingly connected to the full AI stack—from infrastructure and compute to applications, security, governance, and management.

🔹 The next phase of AI adoption will be measured by production outcomes. The key question for enterprises is increasingly whether AI infrastructure can translate model capabilities into measurable business value.

🔹 AI infrastructure decisions are becoming strategic decisions. Where enterprises deploy AI can affect cost, control, compliance, data sovereignty, security, and ultimately the pace at which AI can scale across the organization.

‍

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.

Daniel Newman:

We've been sold the wrong metrics. We've been sold the metric that the most tokens is the best thing. And it's like, that's only true if every token is equal. And so the most successful, accurate outcomes, tokens with the least amount of spend is the real goal.

‍

Patrick Moorhead:

Hey, welcome to Six Five Webcast. I'm Patrick Moorhead here with my bestie Daniel Newman. Everybody knows Daniel. He's everywhere on every channel out there. Hey, today we're going to be recapping VMware Explore 2026. Dan and I attended, watched the keynote, and interviewed their key executives, Ram and Paul, and Purnima. Daniel, how you doing, buddy? It's good to be here.

‍

Daniel Newman:

Yeah, it was a fast, furious few hours. We got in and any of these, the event just felt very Broadcom, right? Tighter, leaner, you know, they really executed on their promise, which was a smaller, more focused, delivery of their, of the offerings. And I think that's what we got. And so it was great to get there. Great to have the chance to sit down. Leadership team still looks very similar in most cases to the leadership that, uh, that we, we actually interviewed you and I on, on deal day. Um, having said that, they have a new, um, you know, a new president though, someone that worked on the infrastructure side. And, uh, it was great that we got to speak to, to Ron, uh, Valanga as well.

‍

Patrick Moorhead:

Yeah, absolutely. And we talked private AI, AI in general, security strategy, and it is amazing how many of those things kind of align with Signal65's PINNACLE, which launched, how it really reinforced using open models, multiple configurations, kind of spanning cloud and also on-prem. So let's dive in. I mean, I think the biggest piece, it was great for Ron to set out the overall strategy. He had really set the plate here. And, you know, we've been talking about private cloud for forever, right? And the gear is finally clicking in place.

‍

Daniel Newman:

You always need that sort of catalyst, right? And the big catalyst of AI is creating some really interesting considerations for enterprises. And I mean, this really is an enterprise show, right? I mean, there's certainly a utilization of VMware by hyperscalers, but where private cloud is meeting the AI era really is about control and cost. And Ron really reiterated that in our conversation. But, you know, companies are looking to have more control over their infrastructure, more control over their destiny, more control over their data, their governance, the security of their business. And then, of course, to your point about Pinnacle and what we're all trying to measure is we're trying to measure what does it cost to deliver the outcomes, you know, knowing that I think increasingly knowing that not all tokens are created equal. And so this is the moment for a platform like VMware to make its presence known. When a company, what was that data point you shared? Something like 20 cents per correct. Yeah, yeah.

‍

Patrick Moorhead:

So a buck 26 to do it in the cloud.

‍

Daniel Newman:

Six times more expensive.

‍

Patrick Moorhead:

So I mean, that can't do it on-prem.

‍

Daniel Newman:

That case is going to make it to the desk of many CFOs and CIOs that are running their AI like a business to consider how they want to build infrastructure, secure it, and deliver in the AI era.

‍

Patrick Moorhead:

Yeah, that's absolutely right. And even though we've heard of VMware, basically private AI, for a while, The biggest takeaways for me was the increased simplification. And even Paul talked about, hey, you know, we could have made your customers are saying, make it easy, give us the easy button. And ironically, the easy button wasn't like, hey, make it as easy as the cloud. No, make it push button; we can go in and out of models. Where you can go in and out of compute. And the company also extended its private AI capabilities to AMD. It used to be the NVIDIA show, with kind of a hat tip to, yeah, we can do the basics, but full VCF 9.1 with AMD, I thought, was a really good one. Daniel, we talked about a lot. It's not just about the compute. It's just not about the tokens. It's not just about the outcomes. You have to do that with the right data, the right control, and the right trust. And Purnima really came in strong with a really strong case. Of why they created what is called TrueSource, which is literally securing open source code, right? Which there are a ton of vulnerabilities in that very few companies, if other, are actually paying attention to that open source software supply chain.

‍

Daniel Newman:

Yeah, that was very interesting because the software supply chain is a massive vulnerability. It's where you actually let the types of code and snippets of code that will be accidentally intentionally put inside of an enterprise's org that can create all kinds of vulnerabilities. A lot of the companies in that space are smaller. And so it was very interesting to get her perspective on why Broadcom was sort of called in to be, you know, the, you know, they provide, was it open source spring there, the provider, heavily utilized, heavily consumed by the open source community. And they're now asking them to say, hey, how can you help us secure the software supply chain more globally and, you know, invest in this? And so, you know, up-leveling from just the pure VMware story, this is now a Broadcom-wide proposition, true source. And so, you know, Broadcom has many software assets, and they're looking at how to secure open source across all of those assets. And then of course, across all software in the industry, which is going to be very important because this is where the increased threat attack surface from AI is going to be potentially most exposed.

‍

Patrick Moorhead:

Yeah, and the other thing that it was good to hear, and not just because it, it, it absolutely confirms what our pinnacle testing said, is the cleaner the data, the less expensive and more accurate your results are going to be. And Tanzu has expanded greatly into data lakes. And it's not a full-up replacement for something like Cloudera, but it's a good start to have all of your data sandboxed with agents and the control layer that sits on top of that. And the agent control is called Agent Minder. And essentially, it adds another control layer around agent access, roles, security rules, and policy. And there's even- I'm sure they won't like me saying this- there's an agent store, an agent marketplace where you can have approved agents that people can pull from, and even agents that aren't approved that the team can test and see how they work. And that's all about this easy button to make it just easier than the cloud for enterprises to put this together on-prem.

‍

Daniel Newman:

Yeah, I think managing agents at scale is going to be one of the largest challenges that enterprises face in the coming periods. And having tooling that simplifies that and that considers all I mean, what happens when agents are given too much autonomy, and you know we can make jokes about civilizations, and we can make jokes about rogue, you know, hacking outside of sandboxes- but if enterprises want to fully maximize the potential of autonomous tools, which is what agents are. And again, I think the word for Numa used that was very salient was agency. Agents have agency. I know.

‍

Patrick Moorhead:

I was like, I did one of those, you know, you say something to your dog, and I'm like, gosh, that's incredible.

‍

Daniel Newman:

Yeah, and with that agency, they have the ability to form a group that can decide to hack or create and stir up trouble or perform excellence with the right steering. And that's the human in the loop. So I thought that was really good. I think another thing that certainly came up was the AI economics. We talked about that at the top here. We're very invested in this, but this is the topic right now. A great analyst had this idea of kind of token-maxing the token efficiency. It happened over a quarter. We're at the point now where the optionality is creating an opaqueness in terms of decision-making, because we've been sold the wrong metrics. We've been sold the metric that the most tokens is the best thing. And it's like, that's only true if every token is equal. And so the most successful, accurate outcomes, tokens, with the least amount of spend is the real goal. And that really cuts across the infrastructure. It cuts across the model itself. It cuts across understanding the workload. And it makes a difference even across that infrastructure where that infrastructure is located.

‍

Patrick Moorhead:

Yeah, that's exactly right. It is really good to see. And by the way, we talk about, hey, in a multi-month period, went from token maxing to token efficiency. But we just said that that's still the wrong metric. right? It needs to be a cost per known good outcome. And it's funny, every time I say it, I'm thinking, oh, that sounds like marketing jargon. But in reality, that's what people have to buy. And Daniel, we're trying to do our best to get that modality and put testing measures of merit out there to help enterprises be able to do this. And one of the things Paul had brought up, and even Ram was, how quickly the idea- like, he had a customer example where he talked about, hey, you know, I had a big financial institution that's like, I'm pretty much doing all of my AI in the cloud. And that's just the way it's where I get all the benefits of the cloud, with resources at my fingertips. And then this movement, based on a configuration of governance needs- I think you call it the alpha or the IP, right? How do you protect that from going up to the cloud? And by the way, don't be confused out there: enterprises that just because they don't have data retention on means they're not learning from the back-and-forth of your humans, but also the agents, in how to do that. And that's why nobody should be surprised that Anthropic stood up its own company to be able to go in and create drugs that our pharmaceutical companies don't want to do.

‍

Daniel Newman:

Yeah. Without being sinister or cynical, you've touched the buttons on the controls with a lot of these consumer-kind of facing tools. But what actually happens That's why enterprise software isn't gonna die, because there's a different rigor, a different, like, you know, people building in VMware, people building in ServiceNow, people building in these tools; there's a different commitment in your MSA and your agreement about what happens. And if it doesn't, like, most people don't even look at what they sign when they use these consumer tools. They have no idea where their data is going. So it's a great point you bring up.

‍

Patrick Moorhead:

So maybe we'll wrap this up. What should enterprises take away from this show? First of all, and we've been very consistent, AI infrastructure needs to be an architectural decision versus isolated projects yu put out in AI land. And the truth is that it spans the workstation by your desk. Your on-prem infrastructure, and your clouds, your neoclouds, to the ultimate cloud. That's one, right? And private AI, even though we've been talking about it for three years, all directions indicate that it's going to be the real deal.

‍

Daniel Newman:

Yeah, I mean, look, I sit in the position where we track this market very closely, and now we're tracking even closer what matters. And what matters is delivering the most efficient tokens that are delivering usable outputs. And by way usable and correct. We're trying to maximize the human inputs, and we're trying to scale the productivity. Then everything else becomes math. Math of, you know, of which infrastructure to run it on, which security softwares to use to secure it, how to govern it, where the infrastructure needs to be. These are all math, and these are all the decisions that enterprises need to make. You and I have talked about edge; we've talked about, you know, on-device; we talk about on-prem; we talk about hybrid. Look, the AI era is going to, it's going to manifest in its own way, but it's going to be a derivative of what we saw in the cloud era, not cloud era, cloud era. And we are going to see a hybrid ecosystem, and increasingly companies are going to be looking very acutely, knowing that you see that the growth- I saw that, you know, the open router token data. Yeah. Doubled in a month, 25 times growth in the year, just incredible growth rate. These are paid tokens, folks. This isn't the free stuff. It's exponential. And so enterprises are starting to see the bill. The bill is a combination, like I said, of the cost, but it's also having the controls over your governance, sovereignty, security, tooling, harness; all that stuff is going to dictate how effectively and efficiently your business implements and what kind of value it derives. So that's at least my read of what VMware was focused on. Every company has to do their own math. They have to make their own decision. They have to assess these different tools. But the case for VMware is compelling because we certainly are seeing indications that in many, many workloads that that on-prem use case does deliver that best value on the token.

‍

Patrick Moorhead:

That's exactly right, Daniel. Good stuff. I guess moving forward, I'm going to be looking, as we wrap this up, I want to see customers going deep. I want to see a lot of them from different areas out there showing how they scale. And to me, the theory, the strategy, the deliverables are right. Now it's time for Broadcom to scale that business. VMware.

‍

Daniel Newman:

Absolutely. Well, we'll keep eyes on it. We'll be watching it. And, you know, the week of this event, Broadcom even printed. We'll have to have a look at those earnings and see how software is doing. That's right. More to come. But thanks, everybody, for tuning in to this coverage of VMware Explore. Postmortem week of we were on the ground, and now we're back at home, but still doing the doing the deal, doing the deed, as they would like to say.

‍

Patrick Moorhead:

That's what I thought you were gonna say, Bestie.

‍

Daniel Newman:

Appreciate everybody tuning in. Subscribe, be part of our Six Five community. For this episode, we gotta say goodbye. See you all later.