Virtualized compute platforms have been a boon in allowing enterprises to more easily shift workloads between different cloud environments, but that shifting ease could also grant organizations greater flexibility as they face increasingly stringent data sovereignty laws, especially when it comes to growing use of AI in operations.

Virtualized compute platforms have allowed for the disaggregation of compute outcomes from compute resources, a motion that has also led to the disaggregation of those compute outcomes into their own parcels of code that can be plugged into different application environments. This model has fostered the rapid development of the cloud-native ecosystem, where platforms have been formed to fill in the management and control of this new composable environment.

As such, cloud-native platforms like containers and the subsequent Kubernetes orchestration platform are increasingly being used to construct and manage sandbox environments for application development. This allows development to take place somewhat off the grid, but also by their nature allows for the data to run those applications to be plucked from different cloud environments and then transferred easily into production via different cloud architectures.

This is great for the development of new services, but can cause a headache when it comes to data sovereignty rules that can require strict line-of-sight into where those composable data pieces come from. Even one piece of data from the wrong cloud environment could cause an application to run afoul of the myriad of sovereignty regulations.

Bola Rotibi, research director for software development at CCS Insights, noted this challenge was core to many discussions at the recent KubeCon Europe 2026 event, stating the “most useful line of the week was that sovereignty shouldn’t mean fragmented codebases. It builds sovereign deployments on shared global commons.”

Real geopolitics at play

But that remains a challenge, heightened by current geopolitical tensions, and one that is forcing organizations to more deeply consider their cloud and data-hosting arrangements.

Dan Ciruli, VP and GM for cloud native at Nutanix, explained during the vendor’s recent .NEXT event that this problem is outsized, depending on location.

“This is not so much true in the U.S. It is definitely true pretty much everywhere else,” Ciruli said. “That is causing people to think about moving workloads, especially if they're in one of the major cloud providers, moving them somewhere. Could be [on-premises], could be a [local] service provider.”

Nutanix, for its part, has tried to broaden the deployment architecture for its platforms toward solving this challenge.

“The big goal of ‘run anything anywhere’ is assuring our customers that should they need to put a workload someplace different, they'll have the ability to do that,” Ciruli explained later. “What are the factors that go into them? It might be cost, it might be just regulatory, or it might be risk. … Fundamentally, that's up to them to decide.”

That decision could be required to adhere to sovereignty rules and how those tie into broader regional requirements, something that is shaping how those regions tailor their cloud infrastructure laws.

Stephen Watt, VP and distinguished engineer at Red Hat's Office of the CTO, explained this to SDxCentral, pointing specifically to the oft-referenced European Union (EU) model in comparison to its opposite in China.

“China has plenty of access to land. Europe, not so much, only in certain places and there's legislative impediments there, so even if there is access to land, Europe, I would argue, is one of our global leaders in driving the sustainability agenda, and they've put all different kinds of laws in the EU around ensuring our planet's going to be in a better place for future generations, but those same laws make it tough to build new data centers in Europe,” Watt said. “And so, Europe, I think, is going to go into this challenge of, do they want to support the sustainability agenda or do they want to stay current with providing their citizen populaces with the AI infrastructure they need?”

The ‘sovereign paradox’

These questions are only expected to intensify as power demands increase. Watt noted that the most recent Nvidia Vera Rubin architecture provides for a lot of processing capabilities, but also requires significant power and cooling.

“So, the question is, can Europe deploy those?” Watt said. “And it turns out if you actually take a look at what's going on in Europe, there's this thing called the ‘sovereign paradox,’ where they don't actually have the capacity in country. Specific countries that are running into this are Germany, the Netherlands, Republic of Ireland, where they're struggling to find the capacity to find data centers that can run these new architectures. And what they're doing is they're giving up national sovereignty in view of regional sovereignty with EU, and so they're looking at the Nordics, especially Finland, and in Finland, and you're having essentially these giant [colocations] that actually house the sovereign infrastructure of multiple nations inside the EU.”

While this model could solve some of that EU-based sovereignty concern, it does open up what Watt said was “a big single point of failure.”

Some of that can be mitigated by cloud-native architectures, which could reduce the compute resources needed to run smaller, composable data sets.

“What's so interesting about Europe is they have these huge fleets of, effectively, cloud-native data centers or virtualization,” Watt said. “This is back to the time of the data center you built for your [virtual machines]. Can one of those things run fleets of CPUs? … If you can't deploy the latest GPUs, wouldn't it be great if you could choose the tens of thousands of CPUs that you already have in place and scale the amount of concurrent inference cores you have available that way?”

Watt did admit that while there could be a solution to the physical challenge, there remains a learning curve toward greater use of cloud-native platforms like Kubernetes. This is especially the case for enterprises that are coming from more traditional models that run applications like high-performance computing (HPC) or are using a simple Linux utility for resource management (Slurm) to manage applications running temporarily in HPC environments.

“Slurm isn't sort of built with these strong reliability and failover guarantees that Kubernetes was designed around, and so there's this, what I would just call an educational growth that's happening, of like, ‘hey, look, my expertise is around Slurm. It was doing this kind of thing in it.’ You still need that for sovereign AI. There will be people pretraining models. There will be people post-training models. But the inference piece, the Kubernetes piece, is new to this game,” Watt explained. “It's not easy for somebody new to wrap their heads around, and there needs to be a strong incentive to actually move over to that. And so, there's a learning path, the maturity path, where they basically start understanding why and how you use Kubernetes to support the inference workloads.”

AI equals aye, aye, aye

Watt’s inference reference is directly related to the growing challenge around AI-fueled application development, a rapidly evolving front that has conveniently found a home within containerized environments.

“AI is making it easier than ever to write applications, and it is increasing the number of people who can write applications. And what that means is we've seen a steady increase in the rate of new applications in my entire lifetime … and AI is going to make that increase fast,” Ciruli said. “For the last five years, containers have been kind of the de facto way you develop any new application. It’s to be deployed into the container. All the new applications are ready to go into containers, and all the AI applications are going to be new applications; they're going to be running in containers. All the new things are running in containers, and all the things that run in containers are running in Kubernetes.”

The Cloud Native Computing Foundation’s (CNCF) most recent Annual Cloud Native Survey showed 66% of respondents reported their organization was using Kubernetes to host generative AI workloads. This was deemed as illustrating “how Kubernetes has become the common denominator for cloud native scale, stability, and innovation, especially as organizations bring AI workloads into production environments.”

“Over the past decade, Kubernetes has become the foundation of modern infrastructure," CNCF executive director Jonathan Bryce noted in a statement. “Now, as AI and cloud native converge, we're entering a new chapter. Kubernetes isn't just scaling applications; it's becoming the platform for intelligent systems. This community has the expertise to shape how AI runs at scale, and we have a massive opportunity to build something open, powerful, and impactful for the next 10 years.”

However, the survey also clarified that only 7% of organizations were deploying AI models “daily” and 47% deploying “occasionally,” with 44% of respondents stating they did not yet run AI or machine learning workloads on Kubernetes, “underscoring the early stage of AI production maturity.”

That early-stage opportunity was also highlighted by a recent Spectro Cloud survey that found enterprises using Kubernetes are twice as likely to scale so-called “edge AI” applications and 20% more confident presenting return-on-investment plans to company leadership.

Ciruli agreed, noting that this AI angle remains nascent, but interest is surging.

“We are in a state right now that 95% of the conversations we have involve AI at some point, but I would say probably fewer than 5% of the workloads that are running today are actually using AI,” Ciruli said. “It is a place where we talk about it a lot, but most of these deployments are still in the future.”

Analysts have also linked this Kubernetes-orchestrated AI opportunity to broader use cases.

ABI Research senior research director Dimitris Mavrakis recently noted that operators need platforms that “combine Kubernetes automation with telco-grade security, observability, and lifecycle management.”

The analyst firm linked this shift to operators preparing their networks to better support what it termed the “next phase of the AI supercycle,” which it claims will be the distribution of AI inferencing across the network.

Looking ahead, CNCF said the next wave of Kubernetes-linked AI innovation “will hinge on resolving cultural adoption barriers, investing in platform engineering, and evolving security and observability standards.”

"Enterprises are aligning around Kubernetes because it has proven to be the most effective and reliable platform for deploying modern, production-grade systems at scale – including AI – and because of the ecosystem and community that support it," Hilary Carter, SVP of research at Linux Foundation Research, explained. "This year's data shows that the next phase of cloud-native evolution will be as much about people and platforms as it is about the tech itself. Organizations that invest in both will have a clear advantage."

This article first appeared in the SDxCentral Sovereignty Supplement.

To read the Supplement for free, simply register.