Building AI Where Enterprise Data Lives: The Next Evolution of Cloud Infrastructure with Oracle

The next phase of enterprise AI won’t be won by choosing the best model. It will be won by bringing the right model to the right data, without moving that data, weakening governance, or locking it into a single cloud.

That’s the harder challenge Karan Batta, SVP of Oracle Cloud Infrastructure, says CIOs are confronting now.

At The Six Five Summit: AI Unleashed 2026, Matt Kimball and David Nicholson asked Batta what that requires in practice. His answer is distributed cloud.

Across more than 50 dedicated regions and Oracle Alloy deployments, OCI runs the same code under a unified governance layer. That allows organizations to bring AI to their data wherever it resides rather than forcing everything into one centralized environment.

Batta frames Oracle’s multicloud expansion across Azure, Google Cloud, and AWS the same way. Multicloud is no longer simply a hedge against lock-in. It is a deliberate strategy that gives customers the flexibility to run the right model in the right cloud through a single control plane.

His test for whether an infrastructure bet will hold up is straightforward: Can it remain flexible as models and chips evolve, stay open rather than proprietary, and become nearly invisible to the people using it?

Key Insights:

🔹 Bring AI to the data, not the other way around. Oracle’s distributed cloud spans more than 50 regions and Alloy deployments, using a unified governance and security layer to run AI wherever enterprise data resides.

🔹 Multicloud has shifted from risk mitigation to strategic advantage. Enterprises once adopted multiple clouds primarily to avoid lock-in. Now they deliberately match workloads to each cloud’s strengths, with Oracle serving as a connective data layer across them.

🔹 The value of distributed cloud varies by industry. Healthcare organizations prioritize patient privacy and low latency; financial institutions need regulatory compliance; manufacturers require AI close to factory operations; and governments must preserve data sovereignty.

🔹 Durable infrastructure must pass four tests. It should adapt as models and chips evolve, favor open frameworks over proprietary stacks, improve long-term token economics, and remain largely invisible to the end user.

Batta's read reframes infrastructure spend as a decision meant to outlast whichever model or chip generation happens to be winning this quarter.

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

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.

David Nicholson:
Hi everyone, and welcome to the Six Five Summit, AI Unleashed 2026. For this AI Infrastructure Spotlight, we'll be exploring how enterprise AI infrastructure is evolving as organizations look to securely bring AI closer to their business data, wherever that data resides. I'm here with my colleague, Matt Kimball. And joining us is Karan Bhatta, Senior Vice President of Oracle Cloud Infrastructure. Karan, welcome to 6.5. Nice to be here. Thank you so much.

Matt Kimball: 

So this is going to be a fun conversation, Karam. We're going to kind of run you through the ringer here. I hope you're ready.

Karan Batta: 

Let's go. I mean, I work at Oracle, so it's not too dissimilar.

Matt Kimball: 

It's called a normal half hour at Oracle. Yep. So let's start out kind of setting the stage, OK? AI, we've been talking about it for a while, right? Early days was models, models, models. We're really starting to see kind of moving from those kind of discussing foundational models and training and experimentation into production, right? We're really starting to see that. And as we watch enterprise organizations move from, you know, kind of proof of concept and pilots and really start to try to scale, one of the things we're seeing is that infrastructure has really come back into focus. You know, we've moved away from talking about this model versus that model. From your perspective, what's what's changed in like how enterprises should be really thinking about AI infrastructure over the last year or so?

Karan Batta: 

Yeah, I mean, look, I think, you know, the overall kind of thought process I would say is like, you know, it's gone from how do I build an AI model to essentially how do I operationalize AI across my entire organization, right? That's the kind of the key message that I would sort of talk about. You know, a year ago, you know, a lot of the customers were like, well, which model should I use? What GPU should I buy? How fast can I train? You know, there's still sort of important questions, but not for the enterprise customer, right? Like today's, today, I think I would say CIOs are asking very different set of questions. How do I connect my AI to my enterprise data? How do I run governance on it? How do I keep my sensitive information to myself within the bounds of my data center or my boundaries? How do I scale AI across thousands of employees and business processes? Actually, Oracle is going through that significant change ourselves. And then how do I control costs over time? That's actually the recent conversation a lot of times is, great, we're using AI, but costs are ballooning. How do I control that? Is it through inferencing? Is it through building my own model? How do I think about the cost aspect of it? This is why AI infrastructure has become really strategic. It's no longer a standard workload, so to speak. It's really embedded into, I would say, all of our applications, especially working at Oracle. We're not just an infrastructure company. We're an applications company as well. So it's embedded into ERP, HCM, supply chain, health care, financial services, manufacturing. So basically, every enterprise workflow that you can think of. And infrastructure is just much more than compute. It's more than GPU compute anyway. It's everything from security, identity, databases, massive data sets. So essentially, we're seeing the enterprises go through the motions and be a bit more pragmatic about they don't want 10 different AI platforms. They want the data to come. Rather than data coming to AI, they want AI to come to the data. So there's lots of change going on as they really figure out How do I embed AI into all of my business processes, people, and function? That's essentially the big change. There's no answer yet, but I think they're going through that sea change now.

David Nicholson: 

Something you alluded to just a minute ago, one of the big challenges being governance. There was a time when mission-critical data for businesses existed in Oracle databases, snug and secure in our own data centers. And part of the cloud revolution was Oracle giving people the ability to have mission-critical data in a lot of different places. So now we're in the AI era, and people are trying to juggle innovation, innovation speed with safety, security, and governance. How does that change what people expect from Oracle or from any cloud or on-prem infrastructure provider?

Karan Batta: 

Yeah, I mean, I'll start with kind of the last statement I just made, which was, you know, most customers increasingly want the AI to come to their data, not the other way around, right? And what that means is that, you know, most enterprise data is not just sitting in one cloud, right? It's, it's on premises, it's in an Oracle database, maybe it's in some of the other cloud providers, maybe it's in some Neo clouds. And, you know, some must actually remain inside a sovereign environment, right? Like with with AI becoming critical infrastructure for a lot of governments, right? If you go outside the US, the geopolitical space, it's governed, right? It's regulated. So, you know, AI kind of changes all those assumptions, right? And what customers are really telling us now is they want to keep the data where it lives, right? They want to apply AI consistently across all their environments, not, you know, snowflake environments. They want to maintain a single security model and governance framework as well at the same time. For us, from an Oracle standpoint, this is essentially what distributed cloud has become. It's really become a significant differentiator for us. We've got now probably over 50 to 70 dedicated regions and alloys that are deployed globally. And it's actually worked out really well for us because it's the same code, it's the same set of services, but the deployment model is slightly different depending on the needs that you have. whether it's public cloud, whether it's a dedicated region, a customer's facility, whether it's cloud a customer. If you've got your data sitting on an extra data box, as an example, right next to your AI, whether it's sovereign cloud, partner clouds with Alloy, they expect the same level of services across all their different data pieces. And they want to apply across all of that. And then because you're getting a single cloud platform with OCI, you're getting single governance layer and identity and security across all those pieces. Not to mention, I talked about Oracle databases a little bit earlier. We have multi-cloud, obviously, with our databases spanning across all of the different cloud providers, including OCI. This also expands to the data that's sitting in those other cloud providers as well. So they don't want different strategies for AI across every environment. They actually want one single, consistent operating model for all of it.

Matt Kimball: 

And is that, you know, when you talk about that, Karan, and kind of, and it's funny, because I step back, and I think, you know, multi cloud has always been important for me as an enterprise CIO, right? And there are a lot of reasons for that. There's operational consistency, or not consistency, but uptime, right? Resilience, it's also, you know, I go to OCI for this, I go to AWS for that, and you know, because there are specializations that sometimes, or services sometimes, one cloud has that's just fantastic. It seems to me that has become even more strategic and more important as AI starts to play in the enterprise equation, right? Because of that kind of, not just like agility, but operational consistency, right? So when you think about how multi-cloud continues to evolve over time, especially as we start to see enterprise AI, what would you say, like, if an enterprise is looking two years out in enterprise CIO, how would they, how do you think they would define success? Like, what does that look like? Is it that just a, it's a single, you know, kind of operating plane for them or control plane for them? And is it that, you know, universal view of all their data? Is there more to it than that? Can you kind of add a little bit more color on that?

Karan Batta: 

Yeah, look, I mean, you know, I would say, you know, multicloud used to be more of a vendor lock-in strategy in the past, right? It's no longer that at all. It's actually about combining the best of the technologies depending on wherever they live, right? Like I would say, like five years ago, when we started our relationship with Azure, as an example, a lot of the multi-cloud conversations were largely defensive. Today, they're offensive, right? Customers are intentionally, I would say, choosing different, different clouds because each, you know, brings us unique strengths, right? And then with AI, I think it only accelerates this trend, right? Because, you know, different models live in different clouds. Everybody's kind of taking their own bet on what cloud and what model is going to going to be sort of the key. You know, different enterprise applications are in different clouds, databases are in different clouds as well. And that's one of the reasons why we have a multicloud platform for databases. And then specialized AI services sort of continue to emerge in different verticals as well, right? We're obviously investing a lot in our applications from a from an AI perspective so success isn't going to be running identical infrastructure everywhere it's going to be making those environments work really seamlessly across all the different environments, whether it's on prem, or whatnot right so It's one of the big reasons why we invested five, six years ago in multi-cloud. We started with Azure. We started with the network connectivity, right? We wanted to make sure that network connectivity is there because if you can't move data for cheap, nothing's really going to work, right? And with AI, the data is sort of the key part of it. So we started with Azure. We then expanded to Google Cloud. We then expanded to AWS. And now what we've become is the central core data hub, almost, I would call it, where you have network connectivity across all clouds back and forth. And you're able to move that data free of charge across different clouds using multi-cloud. And then if your data is sitting in different clouds, it still sits on a singular platform, which is Oracle Database. And you can move that data around. You can use whatever models you want wherever you want. And then you can, you know, maybe they want to use, you know, some verticalized service and Google, but maybe they want to run some core infrastructure and OCI, but then maybe they're running their data on prem, which is also why kind of going back to the last question about distributed cloud is the same API at the end of the day. So you're able to build a site control plane right on top, which sort of is fungible across many, many, many different environments at the end of the day. So look, I think. You know, customers aren't asking for multi-cloud because they love the complexity that it comes with, but they're asking for it because they want the freedom to choose the best of the best irrelevant to where it lives.

David Nicholson: 

You know, some people like to collect cloud merit badges though, you know, wear them on their jumpsuits, like an F1 driver. You know, whenever we're having conversations about infrastructure, especially now, but really always, real business outcomes are always tied to these kinds of decisions. And today, I think you can almost cynically boil this down to the question of how much money can you save me so I can spend that money on tokens? And so if you're making the pitch for distributed cloud, can you nail that down to a single greatest benefit of distributed cloud? Or you've talked a lot about it already, but we're in the elevator, we're riding up, we only have two more floors to go. What's the greatest advantage to distributed cloud?

Karan Batta: 

Look, I think distributed cloud is really ultimately about business outcomes, right? Not deployment models like I talked about, right? So, and it really depends on the kind of customer, right, that's using it, right? Because each different customer will measure success slightly differently. I would say when it comes to distributed cloud, different industries have different constraints. As an example, with healthcare, it's going to be patient privacy. They have to have the patient records and things. It's going to be data residency, clinical AI, being close to the hospitals because of latency. For financial services, it's something completely different, which is regulatory compliance, low latency because if they're doing high-frequency trading or you know, sensitive financial data, as an example, with manufacturing, it's going to be things like factory automation, it's going to be AI near the production environments or predict the maintenance. And then obviously with governments, it's, you know, things like sovereignty, national security, and so on. So it's every different industry has slightly different sort of benefits and success criteria. I think the thing that's key is that the data has got to live within the bounds that you've described. It's going to be right next to your core. infrastructure at the end of the day that's already pre-existing, and you have to have control over it. Those are sort of the three or four metrics that determine success.

Matt Kimball: 

Hey, one last question for you, Karan, and I promise I'll make it quick, but we saved the hardest one for last. We're in an interesting time, right? We're at a point where innovation in this industry, the tech industry, has kind of hit breakneck speed, right? That's partly thanks to AI, but new silicon is coming out on an annual cadence, new models are updating frequently, new frameworks, it's just everything is moving so fast. And as I talk with enterprise CIOs and IT leaders, it's really difficult to say, okay, here's the bet I'm going to lay down today. and I want to make sure I'm making the right bet because this thing is going to stick with me for years to come, right? So you're sitting back, you know, you spend a lot of time thinking about these things for OCI. So as these enterprise leaders are kind of struggling with these decisions, you know, do you have like principles or kind of like here are a few guiding posts for you to kind of think about or set out as you're making these strategic decisions, you know, that kind of brings you to what's next?

Karan Batta: 

You know, at sort of the highest levels is probably a few things, kind of like I would say tailwinds right so I think the key, like as an example the first factor I would say is designed for flexibility. Right? Like foundational models will continue to evolve and rapidly change. And so it's really hard to pick a winner or, you know, hardware generations are going to continue to evolve, right? Everybody's initially it was GPUs from NVIDIA, then it was AMD, and then everybody else is coming out with custom ASICs, right? Infrastructure, you know, today should outlast sort of the today's model choices as well, right? So the fungibility and the flexibility is, I think, really, really key. for these customers. The second thing I would say is building around openness. Customers, I would say they don't want proprietary AI stacks. So they want open models, open frameworks, APIs. They want multi-cloud architectures. Our philosophy has been consistently customer choice, which is one of the reasons why we put the ability for customers to have data in other clouds on our stack. And then optimize for me a long term economics, right, you know, inference is going to be the biggest operational expense for a lot of these customers. So thinking about, you know, how to optimize token cost, infrastructure, efficiency, GPU utilization, network efficiency, operational simplicity, right, what kind of people you need. So I think those are kind of the key things and then I think at the highest level keeping infrastructure. I would say invisible. It's going to be key right because the best infrastructure always sort of disappears under the covers, you know, developers or scientists or data analysts shouldn't be thinking about infrastructure business users that are the token user shouldn't be thinking about infrastructure. They should just simply trust that the AI is going to work securely, performantly, and it's going to scale economically. So I think that's kind of the key set of principles I would think about.

David Nicholson: 

Well, Karan, thanks for joining us for this AI infrastructure spotlight. To our viewers, don't forget to hit subscribe, follow us on social media, and check out all of our 6.5 Summit content at 65media.com slash summit. We'll see you next time.

‍