Why Kubernetes Still Needs VMs — Broadcom on GPUs, Security & the K8s Reality Check

Why It Matters

Broadcom’s VM‑centric Kubernetes approach ensures enterprises can run AI and container workloads securely and cost‑effectively, while its open‑source contributions shape the future standards of cloud‑native infrastructure.

Key Takeaways

• Broadcom contributes Valera for backup, DR, and migration.
• VMware’s VCF remains core for Kubernetes runtime and services.
• Dynamic Resource Allocation bridges VM and Kubernetes GPU handling.
• IDC predicts 85% of containers will run in VMs by 2028.
• Broadcom ranks top‑five CNCF contributor, emphasizing open‑source security.

Summary

At CubeCon Plus CloudNative Con Europe, Broadcom’s Wu outlined the company’s strategy for keeping Kubernetes workloads tightly coupled with virtual machines. He highlighted Valera, an open‑source backup, disaster‑recovery and migration tool, and reaffirmed that VMware Cloud Foundation (VCF) remains the primary platform for delivering a conformant Kubernetes runtime, while Tanzoo focuses on developer‑centric experiences. Key insights included Broadcom’s push for Dynamic Resource Allocation (DRA) to expose GPU and AI resources through familiar VM‑level constructs, enabling smoother AI inference workloads on Kubernetes. The firm emphasized multi‑cluster, multi‑release support with 24‑month enterprise guarantees, and stressed the importance of integrating existing CI/CD pipelines rather than forcing a wholesale re‑tooling of customer environments. Notable remarks underscored the security depth—"six layers of security from hypervisor to container namespace"—and cited IDC’s forecast that roughly 85% of containers will continue running inside VMs by 2028. Wu also pointed out Broadcom’s standing as a top‑five CNCF contributor, citing projects like Contour, Harbor, and Cluster API as evidence of its open‑source commitment. The convergence of VMs and containers signals that enterprises can achieve higher resource utilization, cost savings, and consistent security postures while retaining flexibility to migrate workloads across any conformant Kubernetes distribution. Broadcom’s integrated stack and open‑source engagement aim to simplify complex, multi‑cloud deployments and accelerate time‑to‑production for AI‑driven applications.

Original Description

85% of containers will still run inside VMs by 2028 (IDC). The narrative that Kubernetes would kill virtual machines has been debunked — and Broadcom’s Weiguo He explains why.At KubeCon + CloudNativeCon Europe 2026 in Amsterdam, Weiguo sits down with Techstrong TV’s Mike Vizard to break down how VMware Cloud Foundation is integrating dynamic GPU resource allocation, why running containers on VMs gives you six layers of security, and what enterprises keep getting wrong about Kubernetes at scale. He also discusses Broadcom’s donation of Velero (backup, recovery, and migration for Kubernetes) to the CNCF, and reveals that VMware has been a top-five CNCF contributor for over a decade.Whether you’re running AI workloads, building an internal developer platform, or trying to get hardware costs under control — this conversation cuts through the hype with 18 minutes of real infrastructure talk.

0:00 Intro — KubeCon Amsterdam

0:30 Velero: Broadcom’s New CNCF Donation

1:30 VMware’s Kubernetes Heritage (etcd, Cluster API)2:15 VKS on VCF — Where Kubernetes Runs in VMware’s Stack

3:15 Tanzu vs. VCF: What Goes Where

4:00 Dynamic Resource Allocation (DRA) for GPUs

5:00 AI Inference Workloads on Kubernetes6:00 Meeting Customers Where They Are — Two Approaches

7:30 CI/CD Pipelines: Don’t Rip and Replace

8:30 VMs vs. Containers: Convergence, Not Competition

9:30 IDC: 85% of Containers Will Run in VMs by 202810:15 Six Layers of Security — Hypervisor to Namespace

11:00 Hardware Costs & Resource Consolidation

12:00 Distributed Workloads: Cloud, Data Center, Edge

13:00 VKS as a Conformant K8s Distribution

14:15 Multi-Cluster Management at Enterprise Scale

15:30 VMware: A Top-5 CNCF Contributor

16:30 Open Source Wishlist: GPU Drivers & Platform Engineering

17:30 What Enterprises Get Wrong About Kubernetes18:00 Closing

#Kubernetes #VMware #VCF #CloudNative #KubeCon #Broadcom #CNCF #Containers #VMs #GPUComputing #DRA #AIInfrastructure #PlatformEngineering #DevOps #Velero #EnterpriseKubernetes #MultiCluster #vSphere #TechstrongTV #OpenSource