The empire strikes back – My bad – the Private Cloud is back. For years, we’ve been building private clouds in the image of VMware Cloud Foundation: standardized, software-defined, and (ideally) automated. But anyone who’s run more than a single environment knows the limits—one SDDC Manager per instance, NSX pain points, no clean multi-instance control, and rigid storage design for the management domain.
With VCF 9.0, things are shifting. And it’s not just a product version bump—it’s a conceptual realignment.
Let’s walk through what’s new, what’s important, and why this might finally make VCF a serious candidate for scaled-out, AI-ready, and ops-friendly private clouds.
Being part of the Broadcom Knight program is both an honor and a responsibility. It’s a recognition of individuals who have demonstrated strength in both the technical and sales aspects of VMware’s evolving portfolio—something I’ve had the chance to prove in many projects over the years.
As a Knight, I’ve had the unique opportunity to gain early access to VMware Cloud Foundation 9.0, including deep technical sessions and direct interactions with the engineering and product teams. This gave me valuable insight into the architectural shifts, the rationale behind certain design decisions, and the operational vision that’s now embedded into the product.
Many of the concepts outlined in this blog—like the fleet-based model, automation layer, and updated storage flexibility—are things I’ve seen evolve firsthand during the pre-GA phase. It’s been a great experience to learn, challenge, and contribute feedback along the way.
Modular by Design: From Monolith to Fleet
One of the most fundamental changes: VCF 9.0 introduces a new modular architecture for large-scale deployments. Instead of each environment operating in isolation, VMware now defines three layers:
- A VCF Private Cloud is the top-level construct, representing the overall environment.
- Within that, you can run one or more VCF Fleets—each managed centrally.
- Each fleet contains one or more VCF Instances, which is the traditional combination of management and workload domains.
This allows organizations to scale horizontally, while maintaining centralized governance, policy enforcement, lifecycle management, and automation.
Importantly, VCF Operations now takes over the central control role from the legacy SDDC Manager. It becomes the fleet-level management engine—covering what used to be local to each instance. Lifecycle, observability, compliance, and visibility are no longer trapped in the four walls of a single deployment.
VCF Automation: The Control Plane We’ve Been Waiting For
VCF Automation complements Operations by exposing a fleet-wide provisioning and consumption layer. Think blueprints, templates, service catalogs—and most importantly: an interface not just for infrastructure admins, but also for platform teams, developers, and anyone consuming infrastructure as a service.
It finally answers the question: “How do I consume my cloud like a cloud?” – The big question for our service provider customers and friend – Will it replace Cloud Director? …. Not yet … while multi-tenancy is backed in we will still stick with the Cloud Director communicated end of support – Oct 2027. But let’s face it – VCF Automation will be the future for all kind of self-service offerings – also for Service Providers
No More vSAN Mandate: Fibre Channel and NFS for the Management Domain
In previous VCF releases, the management domain was tightly coupled to vSAN. While vSAN ESA has its benefits, many environments—especially in the enterprise and service provider space some customers have existing investments made in SAN or NAS arrays for core infrastructure services.
With VCF 9.0, you can now use Fibre Channel or NFS as the principal storage for the management domain (No iSCSI). This change aligns VCF with reality: where storage decisions are driven by operational efficiency, not feature enforcement. Besides that keep in mind that vVols are still supported but deprecated.
This also makes VCF more accessible to customers who weren’t ready (or willing) to refactor their storage strategy just to adopt the platform.
Performance Matters: Memory Tiering, vSAN Global Deduplication, and GPU Mobility
Let’s talk performance.
VCF 9.0 introduces memory tiering using NVMe. This allows lower-cost memory to supplement DRAM, without degrading workload density or performance. Especially relevant as AI workloads and memory-hungry apps enter the private cloud. Selling 2TB memory by only having 1.5TB in hardware -> Sounds like a business for me :P
For storage, vSAN ESA now includes global deduplication, meaning dedup isn’t limited to a single disk group or host. The impact? Reduced footprint, more predictable scaling, and improved storage economics—especially in mixed workload clusters.
And for those investing in AI/ML infrastructure: GPU vMotion is now up to 6× faster. That’s the kind of improvement that actually enables maintenance without downtime in production training environments or dynamic resource optimization in inference clusters.
NSX: Finally Simpler, Finally Integrated
Let’s be honest—NSX has always been powerful, but in many cases, not easy.
With VCF 9.0, NSX finally feels like a first-class citizen of the cloud platform:
- NSX lifecycle is aligned with vSphere and VCF upgrades
- A single NSX Manager setup is now supported for smaller or edge deployments
- VPC Model is simpliying how virtual networks can be consumed.
- NSX is now integrated directly in vCenter, simplifying visibility and operations
Licensing & Costs
VCF Operations is now responsilbe for all the licensing and will connect directly to your broadcom account / subscription. No more keys – No more perptual licenses. This is where the industry has moved to.
From a pricing point nothing has changed. 350$ per physical Core list price includes the pure VCF feature set (with 1 TiB of vSAN per comitted core). Other features can be extended as add-ons like Microsegmentation/AVI Load Balanacer/Private AI. While many say that sounds expensive the TCO improvements in hardware investments can be reduced tremendously if you size it right.
And P.S. If I listen to the talks on the street about the price of competitors like Nutanix / Red Hat OpenShift – VCF is equally expensive by offering more value.
Conclusion: VCF is Growing Up
This release of VMware Cloud Foundation signals more than product maturity—it reflects architectural maturity.
With modular VCF Fleets, central control via VCF Operations and Automation, storage flexibility, and real performance improvements, we’re finally getting a foundation that scales with the real-world complexity of modern data centers.
If you’re running multiple VCF instances, planning a private AI infrastructure, or trying to regain cloud-like control without giving away sovereignty or margin to hyperscalers—VCF 9.0 might be the platform you’ve been waiting for.
And if you’re wondering how to align your current architecture with the new fleet model, or whether Fibre Channel makes sense for your management domain—reach out. Always happy to dive deep.
Ready for VCF 9?
Since IT infrastructure is critical besides all simplicity made with VCF 9 the correct architecture and methodology to bring VCF 9 into production is critical. If you want to get challenged if you, your hardware, your organization and processed are ready for VCF 9 – Drop me a massage. W
