This article is based on a 2019 ZDNet interview with VMware SVP and GM for Cloud Native Applications Paul Fazzone, and reflects the state of VMware’s cloud strategy at that time. Some product names and partnerships have evolved since publication.
The Multi-Cloud Problem Nobody Has Fully Solved
Multi-cloud — the use of multiple cloud computing and storage services from different vendors — has been discussed in enterprise architecture circles for years. Yet widespread adoption has lagged behind the conversation. The reason is straightforward: before a company can operate across multiple clouds, it must first master one. That first cloud alone brings its own security, compliance, and operational overhead. Repeating that process for a second cloud is a significant additional body of work.
This gap between ambition and execution is precisely the space VMware has positioned itself to fill — offering a uniform operational layer so that enterprises do not have to learn the individual nuances of each public cloud from scratch.
VMware’s Kubernetes Thesis: The New Normalisation Layer
VMware’s core argument, articulated by Paul Fazzone, is that Kubernetes will become for cloud what server virtualisation became for on-premises hardware roughly fifteen years earlier. In the early x86 era, server vendors introduced hardware capabilities that, intentionally or not, created lock-in. Virtualisation broke that dependency by abstracting workloads away from the specific hardware underneath.
VMware believes Kubernetes will play an analogous role in cloud infrastructure. Organisations that operationalise their applications in a Kubernetes environment gain the ability to deploy them consistently across any public or private cloud — without being tightly coupled to a single provider’s proprietary APIs or tooling.
For context: Kubernetes is an open-source system for automating the deployment, scaling, and management of containerised applications. It has become the dominant container orchestration platform across the industry.
VMware’s Kubernetes Product Portfolio (as of 2019)
At the time of this analysis, VMware offered three distinct approaches to running Kubernetes in production environments:
- VMware Essential PKS — an entry-level Kubernetes distribution aimed at teams getting started with container orchestration.
- VMware Enterprise PKS — a production-grade Kubernetes platform with enterprise security, multi-tenancy, and lifecycle management features.
- VMware Cloud PKS — a hosted, SaaS-style Kubernetes service designed to reduce operational overhead for teams that do not want to manage control plane infrastructure.
Note: VMware has since rebranded and restructured parts of its Kubernetes portfolio under the Tanzu brand. Readers should verify current product naming directly with VMware documentation.
Strategic Acquisitions: Heptio and Bitnami
VMware backed its Kubernetes strategy with targeted acquisitions:
- Heptio (acquired late 2018) — Founded in 2016 by Joe Beda and Craig McLuckie, two of the original Kubernetes creators from Google, Heptio brought deep upstream Kubernetes expertise and enterprise tooling into VMware’s portfolio.
- Bitnami (acquired May 2019) — Bitnami specialises in building and packaging open-source software for deployment across clouds, including Kubernetes Helm charts, virtual machine images, and containers. The acquisition was intended to give VMware customers a consistent, cloud-agnostic way to deploy application packages regardless of the underlying infrastructure.
Together, these acquisitions signal that VMware’s multi-cloud ambition extends beyond infrastructure — it wants to standardise the application packaging and deployment layer as well.
Cloud Partnerships: Not Betting Against the Hyperscalers
A Kubernetes-centric strategy only works if customers can actually run workloads on a choice of clouds. VMware has forged partnerships with major hyperscalers to make this practical:
- Amazon Web Services — VMware and AWS formed a partnership in 2016, producing offerings such as VMware Cloud on AWS. AWS Outposts, announced in late 2018, extended this by allowing customers to run AWS infrastructure on-premises.
- Microsoft Azure — Integration with Azure enables hybrid and multi-cloud scenarios for Microsoft-centric enterprise customers.
- Google Cloud — Particularly relevant given Google’s role in originating Kubernetes.
- IBM Cloud and Alibaba Cloud — Broadening VMware’s reach into regulated industries and Asia-Pacific markets respectively.
Choosing a Cloud: Use-Case Over Loyalty
Fazzone’s advice to organisations evaluating cloud providers is to resist the instinct to pick a single winner. The analogy he draws is to server procurement a decade and a half ago — committing entirely to one server vendor removed all negotiating leverage. The same dynamic is playing out with cloud.
Rather than picking a strategic cloud partner and applying it to every workload, the recommendation is to evaluate clouds based on the specific needs of individual development teams and use cases:
- Breadth of developer tooling — AWS has historically offered the widest range of managed services and developer tools, making it a strong default for general-purpose application workloads.
- Specialist capabilities — Google Cloud has particular depth in machine learning and data analytics services, which may make it the better choice for AI-driven workloads.
- IaaS commoditisation — For undifferentiated compute, storage, and networking, capabilities across major clouds are increasingly comparable, meaning price and commercial terms can legitimately drive the decision.
The Cloud Cost Trap: Consumption Without Visibility
One of the more practically useful points Fazzone raises is cloud cost overrun — a problem that has caught many enterprises off guard. He described multiple cases where organisations, months into their public cloud journey, received bills so large they were forced to cut spending in other budget areas.
The core issue is that cloud consumption is frictionless by design. Developers can provision resources without the procurement cycles that governed on-premises infrastructure, and those resources often remain running — and billing — long after they are needed. This is compounded by the difficulty of right-sizing cloud workloads when engineers are not the ones paying the bill.
For organisations building a multi-cloud strategy, this makes FinOps practices — tagging, showback, automated shutdown policies, and reserved capacity planning — as important as the technical architecture decisions.
Private Cloud Is Not Dead
A recurring assumption during the early years of public cloud adoption was that private cloud would wither. Fazzone’s view, consistent with analyst data available at the time, was that private cloud continued to grow alongside public cloud rather than being displaced by it. The economics of certain workloads — particularly those with predictable, high-volume compute demand — can still favour private infrastructure over metered public cloud consumption.
The practical implication is that a mature multi-cloud strategy typically includes private cloud as a first-class participant, not a legacy holdover to be migrated away from.
Kubernetes as Enterprise Middleware: The Long-Term Bet
Fazzone’s most forward-looking claim is that Kubernetes will evolve into the next enterprise middleware platform — analogous to the role that application servers and messaging middleware played during the 2000s and 2010s in standardising how enterprise applications were built and integrated.
If that trajectory holds, the implication for architects is significant. Just as organisations that standardised on J2EE or .NET frameworks gained consistent tooling for application development across diverse infrastructure, organisations that standardise on Kubernetes today may gain a comparable consistency across cloud environments — normalising not just infrastructure operations but developer services as well.
The services built on top of Kubernetes — service meshes, observability stacks, CI/CD pipelines, policy engines — are, in this framing, the emerging enterprise middleware of the cloud-native era.
Key Takeaways for Enterprise Architects
- Do not commit exclusively to one cloud provider. Maintaining optionality preserves negotiating leverage and protects against provider-specific risk.
- Evaluate clouds by workload type. Match cloud-specific strengths (ML capabilities, breadth of managed services, regional availability) to the needs of individual teams and applications.
- Treat Kubernetes as your portability layer. Containerising and orchestrating workloads via Kubernetes reduces the cost of moving between cloud environments.
- Build cost governance in from the start. Frictionless cloud consumption without visibility leads to bill shock. Tagging standards, budget alerts, and automated resource lifecycle policies are not optional.
- Include private cloud in your multi-cloud model. For predictable, high-utilisation workloads, on-premises or co-located private cloud can be more cost-effective than public cloud metering.
Original reporting by ZDNet (2019). VMware product names, acquisition details, and partnership arrangements cited reflect publicly available information at the time of the original article.