Why Unified Infrastructure Matters More in the Age of Agentic AI
I hear this in almost every executive briefing: "We run thousands of VMs. We also need a modern Kubernetes experience for containers and VMs. Do we really need a second platform for that?"
No. You don't.
VMware Cloud Foundation runs VMs and containers on one platform that includes Kubernetes to orchestrate both. One operational model. One policy framework. One team. Shipping today. Broadcom leads the industry in making this work at scale. And unified infrastructure matters even more as agentic AI demands the orchestration of VMs, containers, and GPU workloads together.
The Hidden Cost of Two Platforms
Most organizations that adopted Kubernetes early stood up a dedicated container platform next to the existing VM estate. It made sense, but the operational reality has been expensive.
Two platforms. Two skill sets. Two security postures. Two licensing models. Two support contracts. Two capacity plans that never talk to each other. The CNCF's own survey data confirms the friction: 77 percent of Kubernetes practitioners report ongoing issues with cluster management and deployment. That is not a technology problem. It is a platform-sprawl problem.
Look at utilization and it gets worse. Siloed infrastructure means stranded capacity on both sides. VMs over here, containers over there, neither team able to share when demand spikes. Fragmentation drives up CapEx and OpEx at the same time.
What a Unified Platform Actually Looks Like
VMware Cloud Foundation (VCF) 9.0 eliminates that divide. At its core is VMware vSphere Kubernetes Service (VKS) - a CNCF-certified Kubernetes runtime built directly into the platform your infrastructure teams already operate.
VKS is not a bolt-on. Cloud admins provision Kubernetes clusters through the same console they use for virtual machines. Platform engineers get a fully conformant Kubernetes API with self-service access, GitOps integration, and multi-cluster lifecycle management - without waiting on IT tickets or standing up separate infrastructure.
The result: VMs and containers share compute, storage, and networking with fault and security isolation from the data center to the application layer. Policies apply uniformly. Security controls with six layers of fault and cyber-threat isolation that span both workload types. One operational model, end to end.
VMs on Kubernetes: The Best of Both Worlds
VCF doesn't just allow you to provision and manage Certified Kubernetes™ clusters with VKS. It allows you to use the same Declarative Kubernetes API to provision and manage VM's.
VM Service supports virtual machine-based workloads, ensuring traditional applications continue to run seamlessly. VMs are managed under the same vSphere Namespaces as Kubernetes clusters with unified provisioning, consistent policies and one operational model.
For workloads that aren't ready to containerize - including legacy applications and vendor software with VM-only support - VM Service provides a migration path that doesn't force an all-or-nothing decision. Platform teams can provision and run traditional VM workloads within the same operational model they use for containers and they can be managed by the platform administrator or the end user. Modernize infrastructure first. Refactor applications later, at your own pace.
The Numbers Tell the Story
The performance data is clear. Broadcom commissioned two studies from Principled Technologies to benchmark VCF against Red Hat OpenShift on bare metal.
In one study, Principled Technologies measured infrastructure platform efficiency. VCF delivers 5.6 times better pod density and 4.9 times faster average pod readiness compared to OpenShift.
In a second study, Principled Technologies measured workload performance. VKS delivers up to 73 percent higher throughput, 78 percent lower latency, and 80 percent more OLTP transaction performance than OpenShift.
Our internal TCO analysis shows that the better container/VM density translates into infrastructure cost savings which lowers OPEX and delivers a 46 percent lower total cost of ownership than OpenShift.
Higher density. Fewer servers. Lower hardware, licensing, power, and cooling costs. That is math any CFO will appreciate.
When “Open” Isn’t Open Enough: A Real-World Test
One of our partners, MomentumAI, who is working with a U.S. national security agency running a mission-critical system across roughly 20,000 CPUs - faced exactly this challenge. Their environment was a sprawl of VMs and bare metal, accumulated over 15 to 20 years across a dozen subsystems owned by different contractor teams. Deployments required four to five hours of scheduled downtime. Scaling meant adding more VMs and hoping for the best. Observability was effectively nonexistent. When performance degraded, the default response was "add more CPU."
They evaluated OpenShift and VKS side by side. OpenShift's opinionated model made it difficult to run several of their older workloads without significant rework. VKS offered the flexibility they needed on top of their existing vSphere infrastructure.
The outcomes were decisive:
- Compute footprint reduced 40–70 percent per subsystem through containerization and data-driven rightsizing - recapturing thousands of CPUs of stranded capacity.
- Deployment windows cut from 4–5 hours to approximately 15 seconds of container restarts for key services.
- Build-and-deploy cycles compressed from 10 hours to 30 minutes in the worst case, with many services dropping from 5 minutes to 30 seconds.
- New environment provisioning dropped from an all-day manual effort to 15–20 minutes using Git-driven, repeatable cluster definitions.
One subsystem had been allocated 400 CPUs per environment. Actual observed usage: 18 CPUs average, 25 at peak. That kind of overprovisioning is invisible without the right platform telemetry - and it was happening everywhere.
This was not a greenfield experiment. It was a production transformation inside an air-gapped, high-security environment. The same platform that ran their VMs now runs their containers, with consistent governance and security across both.
AI Without the Infrastructure Tax
AI is driving the next wave of enterprise workload growth. The conventional response is to stand up another platform - with GPU clusters, container platforms, and VMs all in separate places.
That is the fragmentation trap you want to escape.
VCF treats AI workloads and hardware as first-class citizens on the same unified platform. GPU scheduling, distributed training pipelines, model serving infrastructure - all managed and secured by the same operational model that runs your VMs and containers. No separate platform and no duplicated overhead.
Why This Matters to the CIO
Infrastructure decisions are business decisions. Running two platforms where one can do it all not only drags on agility, it drains budget, and blocks the talent efficiency every organization needs.
Why is VCF with VKS a better solution?:
Skills gap. Kubernetes expertise can be in short supply. Hiring takes quarters, not weeks, and salary expectations reflect that scarcity. Your existing vSphere administrators can manage Kubernetes clusters without retraining. VKS extends the tools and workflows they already know.
Compliance and security. VCF enables customers to meet data-sovereignty and regulatory requirements with FIPS support and end-to-end encryption. Containers run inside VMs, providing hardware-level isolation that bare-metal Kubernetes deployments cannot match. Six layers of multi-tenancy isolation — from datacenter to namespace — protect against both faults and cyber threats.
Ecosystem strength. Broadcom ranks among the top three long-term contributors to the Kubernetes community. VKS is CNCF-certified, which means choice and flexibility. Validated partnerships with F5, Kong, Tigera, and Cosmonic extend platform capabilities for networking, API management, and security. Your teams choose the tools that fit - not the ones a platform vendor mandates.
AI workloads. AI should not force you into another silo. VCF integrates AI and ML workloads with GPU scheduling, lifecycle management, and native support for distributed training pipelines. Your infrastructure team avoids yet another platform to manage with one unified foundation for VMs, containers, and AI.
Application Modernization at Enterprise Speed
Application modernization in real enterprises is gradual, pragmatic, and constrained by business priorities - not vendor timelines.
VCF with VKS gives you that flexibility. Modernize infrastructure first - consolidate onto a single platform, establish consistent security and governance, and eliminate operational silos. Then modernize applications at the pace your business can sustain. Workloads that benefit from containerization move when it makes sense. Legacy apps stay as VMs. New AI initiatives launch on the same platform.
That's pragmatic engineering at scale. This means one workload at a time, on one platform, with one team.
The Bottom Line
You already run VMware. The platform you have can also be the Kubernetes platform you need.
One platform for VMs and containers. Lower TCO. Stronger security. Faster delivery. A team that manages both without doubling headcount.
That is what VCF with VKS delivers. And we lead the industry in doing it.
See what a unified platform looks like for your organization: Explore VMware vSphere Kubernetes Service.