In the rapidly shifting landscape of enterprise virtualization, the strategies for moving away from traditional hypervisors have become a top priority for IT leaders. As standard tools like the Virtual Disk Development Kit (VDDK) become less accessible, organizations are forced to find more efficient, automated pathways to the cloud. This interview features an expert in cloud migration and enterprise virtualization who explores the technical hurdles of modern VMware exits and the rise of next-generation cloning engines. We discuss the transition from traditional agentless tools to high-speed network exports, the role of Large Infrastructure Models in automating complex firewall policies, and why the current trend is moving toward AI-native services and sovereign cloud resiliency.
With the VMware Virtual Disk Development Kit no longer serving as a standard entitlement, how does the shift to using Network File Copy exports fundamentally change the mechanics and speed of a large-scale migration?
The loss of easy access to those C libraries has forced a significant evolution in how we handle virtual disk files. By moving to a proprietary disk copy engine that utilizes Network File Copy (NFC) through standard APIs on the ESXi hypervisor, we actually bypass many of the bottlenecks associated with older agentless tools. While this method performs a full export rather than a delta sync, the performance gains are striking, with measured improvements in restore and image creation speeds reaching up to 2x over the old VDDK-based methods. We are also seeing a massive reduction in the actual volume of data being moved; because we focus on snapshots that cover only the data on the disk rather than its total provisioned size, the total data transfer is often cut by 5x. This makes the “full export” approach much more viable for enterprises that need to move quickly without losing any core functionality or metadata during the stream.
The complexity of translating networking and security policies often stalls migration projects, so how are new technologies like Large Infrastructure Models being used to bridge this gap?
Translating complex environments is no longer just about moving data; it is about replicating the entire infrastructure logic in a way the target cloud understands. Since launching the Large Infrastructure Model technology in March 2026, we have been able to automate the generation and validation of Terraform scripts across diverse multicloud deployments. This system scans the source environment—including the virtual machines, the underlying networks, and specific NSX groups—to deliver a comprehensive firewall plan that maps original rules to the destination’s native security protocols. For an organization, this means they can choose between a self-service model that provides a plan in minutes or a more intensive three-day engineer-led process for highly complex setups. By remediating drivers and converting images into specific formats like those used by OCI or AWS before the move, we ensure the workload is ready to run the moment it hits the new environment.
As organizations look for new homes for their workloads, what are the primary drivers behind the choice between staying on a hypervisor like Nutanix versus moving directly to native public cloud services?
While some organizations find comfort in moving to another hypervisor, there is a clear and growing trend of moving straight to native public cloud services like AWS and Microsoft Azure. A primary motivator for this shift is the desire to integrate more deeply with AI-native services that simply aren’t as robust or accessible in a traditional on-premises environment. We see critical workloads increasingly gravitating toward these major players, while Google Cloud is often selected for more specific, non-critical applications. This isn’t just about finding a cheaper place to host a VM; it’s about gaining the architectural flexibility to leverage modern data tools and specialized compute resources. The goal for most modern enterprises is to stop managing the hypervisor layer entirely and start focusing on the value of the applications themselves.
In the current climate, why are many enterprises shifting their perspective on migration tools, viewing them more as permanent resiliency assets rather than one-time move-and-forget solutions?
The conversation has moved beyond a simple “one-way trip” to the cloud and is now firmly centered on long-term infrastructure control and data sovereignty. Many companies we work with now view these migration engines as resiliency tools that allow them to perform full backup and restore operations across different regions or entirely different cloud providers. This capability is vital for meeting strict sovereign cloud requirements, where being locked into a single ecosystem is seen as a major business risk. We are now seeing batches of 1,000 plus virtual machines being processed simultaneously, not just for migration, but to ensure that the entire infrastructure can be “hopped” to a new environment if the current one no longer meets cost or compliance needs. This shift toward “cloud cloning” gives the power back to the enterprise, making infrastructure truly portable and resilient to vendor changes.
What is your forecast for the future of multi-cloud portability?
I expect that within the next few years, the distinction between “migration” and “standard operations” will almost entirely disappear as infrastructure becomes completely fluid. We are heading toward a reality where Large Infrastructure Models will allow workloads to drift between clouds automatically based on real-time cost, carbon footprint, or AI-service availability. The technical barriers, like disk format conversions and firewall mapping, will become invisible background processes, making the underlying hypervisor or cloud provider a secondary concern to the logic of the application. Ultimately, organizations will maintain a “live” version of their infrastructure that can be redeployed anywhere in the world in a matter of hours, turning what used to be a multi-month project into a routine operational task.
