New research: Virtual machines remain central to cloud native infrastructure as AI workloads evolve

The research arrives at a pivotal moment in the technology sector, where the rapid adoption of artificial intelligence—specifically agentic AI—is forcing architects to reconsider the fundamental security and isolation boundaries of their data centers. While the industry spent the last decade focusing heavily on container orchestration via Kubernetes, this report provides empirical evidence that the "container-first" mandate is shifting toward a more nuanced, hybrid approach where VMs remain the bedrock of production stability.

A Landscape of Coexistence

The survey data clarifies that the industry has moved past the binary choice between containers and virtual machines. Across the 492 organizations polled in September 2026, virtual machines still account for 53% of production applications. Perhaps more significantly, this figure is not a sign of stagnation but of stability; 71% of respondents indicated that their current VM-to-container ratio is expected to remain consistent over the next year.

This trend is most pronounced in large-scale enterprises. Organizations with more than 10,000 employees report that 64% of their workloads reside on VMs. This discrepancy compared to smaller firms (45% for those with fewer than 500 employees) suggests that as infrastructure complexity grows, the tried-and-true isolation and management paradigms of virtualization become increasingly valuable for operational continuity. The data confirms that virtualization is not a relic of the past but a foundational element that is being modernized to support future-facing applications.

The Rise of Agentic AI as a Driver for Virtualization

Perhaps the most notable finding in the report is the correlation between the emergence of agentic AI—autonomous systems capable of executing code and interacting with external tools—and the renewed demand for VM-based isolation. AI agents, by their nature, introduce a level of unpredictability that traditional container runtimes may struggle to contain without significant overhead.

When asked why their VM footprint was expanding, 34% of respondents specifically cited AI and agentic workloads. The reasoning is clear: security and resource management. The research highlights that 47% of those expanding their VM usage are doing so specifically for isolation requirements. By placing AI agents within secure, hardware-level VM boundaries, organizations can mitigate the risks associated with autonomous, potentially malicious, or error-prone code execution.

This is a shift in the perceived utility of VMs. Previously viewed as "heavy" compared to containers, VMs are now being repositioned as the "hard shell" required to safely execute complex, dynamic AI processes. Furthermore, 48% of respondents noted that improved hardware resource utilization was a key driver, proving that virtualization technology has evolved to become more efficient, making it a viable choice for high-performance AI environments.

Organizational Synergy: Bridging the Silos

Historically, the "VM team" and the "Kubernetes team" operated in separate silos, often characterized by different management tools, procurement cycles, and operational philosophies. The 2026 survey data suggests this wall is crumbling. According to the study, 61% of organizations now describe their VM and container administration teams as collaborative or fully integrated.

This organizational shift is crucial for the success of platform engineering initiatives. As infrastructure teams attempt to provide a unified "internal developer platform" (IDP), the ability to treat VMs and containers as part of a single resource pool is becoming a competitive advantage. The adoption of KubeVirt, a CNCF project that allows Kubernetes to manage VMs alongside containers, is a testament to this trend. While only 8% of respondents are currently utilizing KubeVirt, the architectural alignment it represents—managing everything through a single control plane—is becoming the industry’s North Star.

Historical Context and Evolution

To understand the significance of these findings, one must look at the trajectory of the cloud native movement. In the mid-2010s, the "microservices revolution" prioritized containers for their agility and rapid deployment capabilities. By 2020, Kubernetes had become the industry standard for orchestration. However, by 2023, security researchers and cloud architects began noting that "container breakout" vulnerabilities required more robust isolation for sensitive workloads.

The 2026 report serves as the maturation point for this cycle. The industry has reached a synthesis: keep the agility of containers for web services and microservices, but return to the security and isolation of VMs for data-intensive, autonomous, and high-risk AI workloads. This is not a "step back" but a "step forward" into a mature, hybrid infrastructure model.

Implications for the Future of Infrastructure

The implications of this research for CTOs and infrastructure architects are threefold:

1. Strategic Hybridization: Organizations should stop viewing VMs as legacy debt and start treating them as a strategic component of a modern, secure architecture, especially as they integrate AI agents.
2. Unified Operations: The move toward cross-functional teams is not just a cultural preference but a technical necessity. Platforms that cannot manage both VMs and containers under a unified policy engine will likely struggle to scale as AI workloads grow in complexity.
3. Isolation as a First-Class Citizen: As agentic AI becomes a standard component of enterprise software, the "isolation boundary" will become the most important design choice for engineers. The survey clearly indicates that the industry is trending toward multiple layers of isolation, including MicroVMs and hardware-level enforcement, to protect sensitive data.

Expert Perspectives and Industry Reactions

While the report is data-driven, the implications have resonated across the broader cloud native community. Industry analysts have noted that the findings mirror the recent push toward "platform engineering," where the goal is to abstract the underlying infrastructure (be it bare metal, VM, or container) from the developer.

"The findings from the 2026 Cloud Native Workloads Survey provide an essential reality check for the industry," noted a representative involved in the study’s release. "We have spent years debating which technology would ‘win,’ but the data shows that the winners are those who can effectively orchestrate the mix. The future isn’t about choosing one or the other; it’s about having the flexibility to run the right workload on the right foundation."

Broadcom’s involvement in the report further underscores the importance of the virtualization layer in modern hybrid cloud environments. By backing this research, the firm highlights its continued commitment to helping organizations leverage existing investments in virtualization while simultaneously scaling their containerized and AI-driven initiatives.

Conclusion and Looking Ahead

The 2026 Cloud Native Workloads Survey serves as an essential benchmark for the industry. It confirms that while the cloud native movement is fundamentally changing how applications are built, the underlying infrastructure is evolving to be more inclusive of established technologies.

As we look toward 2027 and beyond, the focus will likely shift from basic adoption to the optimization of these hybrid environments. Organizations that succeed will be those that view their infrastructure as a spectrum rather than a series of disconnected platforms. By embracing the strengths of both containers and virtual machines, businesses can build a resilient, scalable, and secure foundation capable of supporting the next generation of intelligent, autonomous applications.

For those interested in the granular data, the full report and its associated datasets, available on Zenodo via the Linux Foundation Research community, provide a wealth of information for further analysis. It is a mandatory read for any professional involved in the strategic direction of enterprise cloud infrastructure. The data is clear: the infrastructure of the future will be defined by its ability to integrate the best of the past with the demands of the present.