Understanding the AFA Tax

The AFA tax is the premium organizations pay for storage systems that use off-the-shelf components but are priced as high-margin appliances. All-flash arrays (AFAs) were marketed as a path to simplicity. Instead, they introduced one of the most persistent cost burdens in IT infrastructure.

Vendors like NetApp and Dell continue to mark up standard SSDs by 5–10x and wrap them in rigid support contracts, opaque licensing, and locked-down hardware. Pure Storage does create its own custom flash modules, but at a high cost for almost no ascertainable value.

The AFA tax goes far beyond drive cost. It affects every part of infrastructure planning—from licensing and patching to backup and recovery. It leaves IT professionals looking for an AFA alternative.

The Flash Markup Problem

Most AFAs utilize standard U.2 or E1.S NVMe SSDs sourced from vendors such as Solidigm, Kioxia, or Samsung. A 4TB enterprise NVMe drive, which costs $400–$600 retail, is priced over $4,000 once installed in an AFA.

That’s the first layer of the AFA tax. And it’s only the beginning.

Storage controllers are x86 servers running the vendor’s software. There’s no proprietary silicon. No hardware acceleration. Yet the pricing model treats these servers as premium infrastructure. Multiply that markup across a full array and support contract, and IT teams are locked into a cost structure that has remained unchanged for over a decade.

Some vendors now offer software-defined storage (SDS) solutions, but the model is based on the same logic: mark up commodity components and resell them at appliance margins.

AFAs Still Leave You With Work

The AFA tax doesn’t end with hardware. It extends into operations. Even after buying a modern AFA, teams still need to manage VMware, backups, patch cycles, and networking. The array doesn’t reduce complexity—it carves out a new island in the middle of it.

Patching is a good example. AFAs must be patched independently of the hypervisor. These updates require coordination and, in many cases, downtime. They are unaware of hypervisor workloads and don’t coordinate with HA policies.

Compare that to a system like VergeOS, where the hypervisor, storage services, and network services are built into the same OS. VergeOS knows what’s running where and can shift workloads automatically before applying patches—keeping everything online and consistent.

That workload-aware patching removes friction. The AFA model preserves it. That’s operational tax, layered on top of hardware tax.

The False Promise of HCI

HCI was supposed to remove the AFA tax. Instead, it replicated it—in software.

Nutanix runs its storage services inside controller VMs. These are managed by the hypervisor and subject to its limitations. VMware vSAN takes a similar approach, embedding storage services in a VM and maintaining a separate control plane. Both

In both vSAN and Nutanix, storage relies on the hypervisor and doesn’t function independently. It competes with virtual machines for resources such as compute and memory, thereby increasing management complexity and causing new performance issues. Vendors require YOU to buy higher-end hardware to address these challenges, even when it shouldn’t be necessary.

That’s the second form of AFA tax—paying for software-defined infrastructure that recreates the same silos it was meant to collapse.

Beware of the “New” Hypervisors

The Broadcom acquisition of VMware has created a vacuum. Storage and HCI vendors are racing to fill this gap by adding KVM-based hypervisors to their stacks. Many of these offerings are in early development.

Here’s the problem: hypervisors aren’t plug-ins. They take years to build, test, and integrate. These vendors are repackaging open-source code and running it as a separate module to their storage stack. They lack native APIs, coordinated HA models, and proven upgrade paths.

Many charge based on sockets, cores, GPU, or storage capacity. Even if the software is open source, the AFA tax remains, although it is now hidden in a different line item.

If a vendor has a new hypervisor, ask why they have it. Then ask how well it works with the rest of the system—and who is responsible for recovery when it doesn’t.

The Backup Tax

While AFAs and HCI are different approaches to the same problem, they share one challenge: backups are external.

Most vendors treat data protection as a point of integration. Customers must license backup software, buy separate appliances, and coordinate retention policies across systems that don’t communicate.

In many cases, the costs of backup infrastructure are comparable to those of primary storage. That’s a major part of the AFA tax. It’s not only about the array. It’s the backup software, second-tier storage, replication, and orchestration overhead.

All of that adds up to a second infrastructure stack that exists solely to protect the first.

What a Better Model Looks Like

A better model eliminates appliance pricing and operational sprawl. It utilizes standard NVMe drives in standard servers to create a distributed storage fabric.

The storage system should present block and file services natively. It should not require an external backup product. It should protect itself. It should support live patching, integrated replication, and instant recovery.

That’s the only way to eliminate the AFA tax.

One example of this model is VergeIO. The company recently launched its AFA Replacement Kit, which combines VergeOS and Solidigm enterprise NVMe drives in a ready-to-deploy configuration.

VergeOS pools SSDs across nodes, manages compute and networking in the same OS, and includes built-in snapshotting, cloning, replication, and data availability tools. Storage is an equal part of the operating system—not a guest process or a separate layer. That means it performs better, scales more cleanly, and doesn’t depend on outside software for core services.

VergeOS also supports GPU-based AI workloads without a separate Kubernetes stack. Virtual machines, storage, and AI pipelines all run in the same environment.

Licensing is per server. There are no capacity tiers or per-feature upcharges. VergeOS installs in minutes and runs on the hardware IT teams already own.

Tactical Takeaways for the Next Refresh

Before committing to another AFA, ask:

• What markup am I paying on SSDs?
• Can I reuse my existing servers?
• Does it help me solve my VMware problem?
• Is the hypervisor included, or bolted on?
• How does patching work?
• Is backup built in or extra?
• Can I run AI workloads on the same platform?
• Does the licensing model follow the hardware, or fight it?

If the answer to any of these questions is “no,” then the AFA tax is in effect.

StorageSwiss Take

The AFA tax is more than inflated hardware pricing. It’s an entire architecture that separates storage, compute, and protection across disconnected layers—and then charges for each.

There are alternatives. The VergeIO + Solidigm AFA Replacement Kit is one example. But what matters more is the model: no markup, no layering, and no hidden costs.

The best way to save money on your next AFA isn’t to negotiate harder. It’s to replace the AFA entirely—with a platform that doesn’t charge the same tax in a different form.