Self-Managed VMware on Google Cloud Debuts on Z3 Bare Metal

Google Cloud has announced the general availability of its self-managed VMware solution on Google Cloud across 22 initial cloud regions. The offering provides enterprise infrastructure teams with complete root administrative control over VMware Cloud Foundation (VCF) 9.x stacks running directly on dedicated Compute Engine bare-metal nodes. Operating as part of the Google Cloud Modernize portfolio, the release complements the existing Google Cloud VMware Engine (GCVE) managed service, establishing two distinct operational paths for running enterprise VMware workloads inside Google Cloud data centers.

Self-managed VCF 9.x arrives on dedicated Google Cloud bare metal

The primary architectural shift in the newly released deployment model centers on administrative authority and operational ownership. Where Google Cloud VMware Engine operates as a turnkey, Google-managed platform built on VMware Cloud Foundation 5.2—restricting customer access to role-based vCenter permissions—the self-managed VMware solution provides unfiltered root administrative privileges across ESXi hypervisors, vCenter, and NSX Manager. This level of access allows infrastructure engineering teams to preserve existing operational playbooks, deploy custom monitoring agents, and schedule maintenance windows according to internal policies rather than provider-enforced schedules.

A central driver behind the offering is Broadcom’s introduction of VMware Cloud Foundation 9.x, which features a consolidated control plane designed for unified stack deployment and automated lifecycle operations. Through this self-managed model, enterprises manage their hypervisor lifecycle, security patching, and configuration drift directly using native VCF Operations Management tooling, rather than relying on automated cloud-provider maintenance pipelines.

Hardware architecture, NVMe tiering, and storage integration

The self-managed offering is provisioned on dedicated Compute Engine Z3 bare-metal instances (specifically the z3-highmem-192-highlssd-metal instance shape). Each node is powered by 2.2 GHz 4th Generation Intel Xeon Scalable processors (Sapphire Rapids), providing 96 physical execution cores (192 hyper-threads), 1.5 TB of system RAM, 72 TB of local high-throughput NVMe flash storage, and 200 Gbps network interfaces. Dedicated physical isolation eliminates multi-tenant hypervisor sharing risks, addressing operational and compliance concerns regarding shared-infrastructure security boundaries such as those explored in cloud isolation bypass vulnerabilities.

Storage architecture relies primarily on VMware vSAN Express Storage Architecture (ESA), utilizing the high-performance local NVMe drives. Google claims this local storage architecture can reduce storage total cost of ownership (TCO) by up to 39% compared to traditional external storage arrays, though actual savings depend on individual workload IOPS requirements and data reduction ratios. Furthermore, the architecture supports NVMe memory tiering between physical DDR memory (Tier 0) and local NVMe media (Tier 1), which Google and Broadcom project can lower server memory TCO by up to 40% for memory-intensive workloads.

To support database and storage-heavy workloads without requiring additional compute node purchases, the platform supports Google Cloud NetApp Volumes as certified external datastores attached via NFS. This allows storage capacity to be scaled independently of compute nodes. At the network tier, the bare-metal nodes interface natively with Google Cloud VPCs, Andromeda virtual networking, and the Network Connectivity Center (NCC), enabling low-latency private connectivity to services like BigQuery and the Gemini platform.

Licensing portability and operational governance

The commercial structure of the offering aligns with Broadcom’s revised licensing framework, which emphasizes VMware Cloud Foundation license portability. Under the Broadcom and Google Cloud partnership, organizations with existing portable BYOL VCF subscriptions can transfer their entitlements directly to Google Cloud bare-metal nodes without purchasing duplicate software licenses. On the infrastructure side, organizations can apply Google Cloud Flexible Committed Use Discounts (Flex CUDs) directly to the underlying Z3 hardware instances to reduce compute spend.

Operating a self-managed hypervisor environment in the cloud requires clear division of operational responsibilities. While Google manages the underlying server chassis, power, cooling, and physical network fabric, customers retain responsibility for hypervisor patch cycles, guest OS management, backup orchestration, and disaster recovery. Google reports that VCF 9.x active live patching covers up to 80% of critical security updates without requiring host reboots, which can significantly reduce planned maintenance downtime. For enterprises seeking root administrative control without expanding internal platform operations teams, Google has established support frameworks with systems integration partners, including Accenture, Cognizant, HCL, Infosys, Tata Consultancy Services, and Wipro, to provide managed deployment and Day 2 operational services.

What happens next

The self-managed VMware solution is available immediately across 22 Google Cloud regions globally, with additional regional deployments scheduled across subsequent quarters. Infrastructure teams planning data center exits or hybrid migrations should first assess their operational requirements to decide between the managed GCVE platform and the self-managed bare-metal model.

Key immediate steps for platform leaders include verifying that existing Broadcom contracts carry portable VCF subscription entitlements, evaluating whether storage capacity demands necessitate NetApp Volumes external datastores, and assessing whether internal operations teams have the capacity to manage VCF 9.x lifecycle tooling independently or require third-party managed service provider support.