Research Finder
Find by Keyword
Does VMware Cloud Foundation 9.1 Reposition Private Cloud as the AI Control Plane?
Broadcom focuses on infrastructure efficiency, governed developer workflows, and emerging AI observability to redefine VCF’s role in production environments
05/05/2026
Key Highlights
- VMware Cloud Foundation 9.1 emphasizes extracting more usable capacity from existing infrastructure through memory tiering and cluster-wide efficiency controls
- Developer velocity is framed through repeatable, governed environments rather than open-ended infrastructure access
- AI-specific telemetry begins to surface alongside traditional metrics, signaling a focus on managing inference behavior
- Native Kubernetes services and object storage expand VCF into a broader application platform footprint
- The release aligns with enterprise demand for controlled environments to run production AI workloads
The News
Broadcom previewed VMware Cloud Foundation 9.1 with updates across infrastructure, application delivery, AI operations, and recovery; this note focuses on the capabilities that define how the platform operates and coordinates resources in production. The release includes memory tiering using NVMe, cluster-wide deduplication and compression, native Kubernetes services, object storage on vSAN, and new observability capabilities for AI workloads. Broadcom also cited early momentum for VCF 9.0, including approximately 2,000 customer deployments and over 19 million allocated cores, positioning the platform as widely adopted across enterprise environments. For more details, read the official Broadcom press release.
Analyst Take
Enterprise infrastructure teams are balancing a different set of priorities and constraints than they were 12-18 months ago. As more operational functions consolidate inside VMware Cloud Foundation, the platform begins to govern infrastructure decisions.
The reality for most organizations is that AI workloads are moving into production, where cost, data locality, and operational control become immediate concerns. Hardware pricing remains elevated across memory, GPUs, and networking, while power availability and data center capacity now shape scaling decisions. Meanwhile, platform teams continue to push toward Kubernetes as a default interface, even as most environments remain rooted in virtualized infrastructure. The result is friction: move faster without losing control, adopt new tooling without duplicating cost, and support new workloads without destabilizing existing systems.
According to the HyperFRAME Research Lens (1H 2026) survey, enterprise IT and data leaders reported that only 23% of AI/ML projects reach production and meet their original ROI objectives, reinforcing the need for systems that prioritize execution discipline over experimentation.
In addition, finance now enters these decisions earlier. AI workloads introduce consumption patterns that are difficult to forecast, especially when tied to accelerators and memory. Most organizations are not rebuilding from scratch. Existing estates, teams, and processes remain in place, which limits how far operating models can evolve in a single step. For many organizations, the practical path is to extend existing resources and processes.
VMware Cloud Foundation 9.1 aligns with that reality. Broadcom is positioning VCF as the system that coordinates how infrastructure is consumed, which places it directly in competition with Kubernetes-centric control layers such as Red Hat OpenShift and emerging platform engineering stacks.
We see that the integration of AI-specific telemetry, such as token usage and agent activity, directly into VCF 9.1 enables Broadcom to pivot the sales conversation from virtualization density to AI unit economics, providing CFOs with the granular cost-per-inference metrics they currently lack. By embedding these capabilities into the existing vSphere estate, Broadcom freezes the competitive landscape, making the leap to a pure-play Kubernetes stack like OpenShift appear unnecessarily risky and expensive for the 77% of AI projects currently failing to reach production.
This governed execution model compresses the sales cycle by enabling infrastructure teams to repurpose their existing skills and hardware for AI workloads, bypassing the typical 12-to-18-month talent re-skilling bottleneck. By optimizing memory and storage at the platform level, Broadcom creates a distinct competitive moat where VCF becomes a hedge against volatile GPU and DRAM pricing, offering a software-defined discount on AI scaling.
From our viewpoint, the shift toward blueprint-based, self-service catalogs also positions VCF as a direct competitor to platform engineering startups, as it offers a pre-integrated IDP-in-a-box that aligns with the enterprise preference for governance over raw developer flexibility. This strategy transforms VMware from a hypervisor into a strategic AI control plane, forcing competitors to prove their value in environments where operational discipline and capital preservation have replaced experimental agility.
Infrastructure Efficiency, Productivity, and Performance
In VCF 9.1, efficiency, productivity, and performance are treated as parts of the same system. Memory tiering extends DRAM with NVMe, relocating inactive memory while keeping active data in place. Storage follows the same logic, with cluster-wide deduplication and compression operating across the environment to reclaim usable capacity. Visibility, recommendation, and action form a continuous loop, allowing infrastructure teams to move away from peak usage assumptions and operate closer to actual limits.
That loop changes how resources are consumed. Infrastructure teams can increase density without immediate hardware expansion, while platform teams encounter fewer constraints when deploying workloads. Finance sees a slower expansion curve as existing assets carry more load.
The same consolidation extends across the platform. Compute, storage, networking, and application services work within a shared model, with object storage pulled into the same boundary. Fewer systems require coordination, with fewer transitions between teams.
From a developer perspective, VCF 9.1 can increase velocity by capturing a working state and making it reusable as a catalog action. VM groups define dependencies, namespaces preserve configuration, and validated environments are published as blueprints. Platform teams gain speed within defined boundaries and infrastructure teams can avoid repeated configuration work. This model contrasts with Kubernetes-native approaches that emphasize developer autonomy, suggesting Broadcom is prioritizing repeatability and governance over flexibility in enterprise environments.
Execution models expand without forcing a single abstraction. Lightweight container workloads can run without a full Kubernetes cluster using VMware Kubernetes Service (VKS) integrated directly into VCF, while Kubernetes services remain available for more complex applications. Environments can also be extended post-deployment with containers or full clusters without introducing a separate control plane. Runtime becomes a decision tied to workload characteristics rather than a default applied everywhere.
AI observability begins to extend the platform into workload behavior through VCF Operations telemetry enhancements. Metrics such as token usage, agent activity, MCP servers, and model performance appear alongside CPU and memory utilization, supported by visibility into GPU usage, topology-aware scheduling, and autoscaling behavior. Infrastructure teams gain more context when diagnosing issues. Platform teams can start to connect application behavior with system behavior.
Networking and edge extend the same operating model outward. Standards-based EVPN integration with Arista, Cisco, and SONiC enables coordination between physical and virtual networking environments without requiring replacement of existing infrastructure. Edge deployment follows a similar pattern through zero-touch provisioning integrated with vCenter, where systems can boot, connect, receive configuration, and self-integrate into the VCF environment.
Across these areas, manual decision points are reduced through VM groups, blueprint-based provisioning, policy-based resource controls, and catalog-driven deployment workflows. Templates define how environments are constructed, sequencing controls manage dependencies, and policy enforcement governs resource usage. Behavior becomes more consistent as the system applies these patterns across deployments. In our view, these capabilities reflect a shift toward treating infrastructure as a governed execution layer.
Production conditions still determine whether these gains hold. Variability, contention, and changing demand patterns introduce pressure that does not appear in controlled scenarios. Predictable behavior under load and visibility into tradeoffs remain critical, particularly across memory tiering, cluster-wide data reduction, and capacity modeling workflows within VCF Operations.
What Was Announced
VMware Cloud Foundation 9.1 introduces enhancements across compute, storage, networking, application delivery, and AI workload support. The release builds on VCF 9.0 with improvements spanning resource management, deployment workflows, and operational visibility.
Memory tiering uses NVMe as a secondary memory layer beneath DRAM. In the demo, VCF Operations identified memory constraints and presented configuration options, including device selection per host, mirrored NVMe devices for resiliency, encryption enablement, and policy-based tier sizing. The system modeled host impact before applying changes. Broadcom indicated memory tiering CPU overhead in VCF 9.0 was under 10% (approximately 5% for most workloads), with VCF 9.1 reducing CPU utilization by 12%.
Storage efficiency is extended through cluster-wide deduplication and compression within vSAN. The demo enabled deduplication across the cluster while maintaining encryption, then updated the capacity summary to show reclaimed storage measured in terabytes.
VCF Operations introduces capacity modeling and planning workflows. Dashboards display resource utilization, projected capacity, and estimated cost impact. In the demonstration, the platform evaluated support for additional AI workloads, identified constraints, applied storage and memory changes, and recalculated available capacity after optimization.
Application deployment enhancements include VM groups and blueprint-based provisioning. VM groups allow administrators to define sets of virtual machines with ordered startup and shutdown behavior. The demo captured an application namespace as a blueprint, validated network and storage mappings, stored images in a project library, scoped access to specific users, and published the blueprint to a self-service catalog. Users could request and provision environments directly, with cost tracking associated to business units.
Modern workload support includes native VMware Kubernetes Service (VKS) integration and object storage delivered through vSAN. VKS is deployed as part of VCF, and environments can be extended with containers or Kubernetes clusters after initial provisioning. AI workload support includes GPU slice selection, autoscaling capabilities, topology-aware scheduling, and DirectPath I/O support for AMD GPUs. The platform also introduces telemetry for AI workloads, including metrics such as token usage, agent activity, MCP servers, and model behavior.
Networking updates include alignment with standards-based control models and integration with Arista, Cisco, and SONiC through EVPN-based approaches. These capabilities support coordination between physical and virtual network environments. Edge deployment capabilities include zero-touch provisioning. Bare metal systems can boot, connect to vCenter, receive configuration, and integrate into the VCF environment without manual setup at the remote location.
Additional operational visibility enhancements include topology views, capacity dashboards, pre-deployment cost estimates, and business-unit level cost tracking for provisioned environments.
Looking Ahead
Infrastructure platforms are now evaluated on how effectively they manage complexity. AI workloads amplify pressures around cost, coordination, and governance, requiring systems that operate predictably under sustained demand. Broadcom noted that the company’s Private Cloud Outlook survey indicates that 56% of organizations plan to run inference workloads in private cloud environments, driven by sovereignty, security, cost control, and regulatory requirements.
Differentiation is moving toward coordination and control. Hardware performance remains necessary, but the ability to orchestrate resources, maintain utilization, and provide visibility into workload behavior carries more weight as environments scale.
Adoption will reflect how organizations balance consolidation with flexibility. Expanding platform scope simplifies operations while increasing dependency on a single system. Integration depth and ecosystem compatibility will influence how that tradeoff is managed.
Transparency will determine trust. As automation increases, organizations will expect clear visibility into how decisions are made and how they can be adjusted. In our opinion, VMware Cloud Foundation 9.1 moves VMware toward a more central role in coordinating enterprise infrastructure. If Broadcom can sustain this direction, VMware’s position may evolve from a virtualization layer to a control point within enterprise AI infrastructures. This is particularly relevant in environments that prioritize governance and cost discipline over full cloud-native transformation. Its impact will depend on how consistently the platform performs as workloads scale and variability increases.
Don Gentile | Analyst-in-Residence -- Storage & Data Resiliency
Don Gentile brings three decades of experience turning complex enterprise technologies into clear, differentiated narratives that drive competitive relevance and market leadership. He has helped shape iconic infrastructure platforms including IBM z16 and z17 mainframes, HPE ProLiant servers, and HPE GreenLake — guiding strategies that connect technology innovation with customer needs and fast-moving market dynamics.
His current focus spans flash storage, storage area networking, hyperconverged infrastructure (HCI), software-defined storage (SDS), hybrid cloud storage, Ceph/open source, cyber resiliency, and emerging models for integrating AI workloads across storage and compute. By applying deep knowledge of infrastructure technologies with proven skills in positioning, content strategy, and thought leadership, Don helps vendors sharpen their story, differentiate their offerings, and achieve stronger competitive standing across business, media, and technical audiences.
Ron Westfall | VP and Practice Leader for Infrastructure and Networking
Ron Westfall is a prominent analyst figure in technology and business transformation. Recognized as a Top 20 Analyst by AR Insights and a Tech Target contributor, his insights are featured in major media such as CNBC, Schwab Network, and NMG Media.
His expertise covers transformative fields such as Hybrid Cloud, AI Networking, Security Infrastructure, Edge Cloud Computing, Wireline/Wireless Connectivity, and 5G-IoT. Ron bridges the gap between C-suite strategic goals and the practical needs of end users and partners, driving technology ROI for leading organizations.