The Infrastructure Reckoning, Part 2: The Virtualization Reckoning

The invoice landed and nobody wanted to open it.

Not because the number was surprising. Because it confirmed what everyone in the room had suspected for months: the platform that enterprise IT had built its entire operational model around had just repriced itself out of the relationship. No negotiation. No grandfathering. New terms, new math, and a renewal window that didn't leave much room for alternatives.

That moment — repeated across thousands of enterprise environments between 2023 and 2024 — was not a licensing dispute. It was a stress test. And a lot of stacks failed it.

59% of VMware customers saw prices rise by more than 25% after the Broadcom acquisition, with the median increase landing between 25% and 49%. Extreme cases hit 350% to 700%. Gartner's Peer Community survey found 74% of IT leaders actively exploring VMware alternatives. And yet most of them haven't moved. 86% of organizations are actively reducing their VMware footprints, but just over half plan phased, partial transitions rather than full migrations. The frozen middle is real — evaluating options, running pilots that go nowhere, stretching hardware into its sixth and seventh year because every alternative looks expensive and uncertain. Yahoo Finance

Here's what the invoice exposed. Enterprise IT had spent fifteen years consolidating onto a single hypervisor, building operational muscle memory around it, training every engineer on it, and treating it as infrastructure rather than a product with a vendor behind it. The abstraction was so complete that most organizations couldn't answer a basic question: what would it actually take to run this environment on something else? The answer, in most cases, was that nobody had tried.

That's not a technology problem. That's a strategy problem that technology made invisible.

The hypervisor isn't going away. But it is no longer the center of gravity, and the teams that treat it like it still is are building on a foundation that is actively shifting under them.

Look at where workloads actually live now. 82% of container users are running Kubernetes in production as of the 2025 CNCF Annual Survey. That number was 66% in 2023. Kubernetes didn't win because it was easy — it won because the operational model for containerized workloads doesn't need a hypervisor underneath it. Bare metal came back for AI inference and training because GPU utilization numbers drop fast when you add hypervisor overhead, and nobody running a $30,000 GPU wants to give 15% of its throughput to a virtualization layer that isn't adding value for that workload. The workloads that still need a hypervisor are real. But that's a subset now, not the default. Cloud Native Computing Foundation

The modern virtualization stack looks like three things running in parallel, not one thing running everything.

A hypervisor for the workloads that genuinely need it — legacy application estates, Windows workloads, anything with hard isolation requirements. Not as the default deployment target, but as a deliberate choice for specific workload profiles. The evaluation criteria for that hypervisor — cost, licensing model, operational tooling, migration path — matter more now than they did when there was effectively one option.

Kubernetes for anything containerized. 54% of AI and ML workloads are now running on Kubernetes, with over 90% of teams expecting that number to grow within 12 months. The organizations that resisted this in 2020 and 2021 are now playing catch-up on tooling, skills, and operational practices their peers built years ago. The Open Source Post

Bare metal for AI and high-performance workloads. The infrastructure underneath a GPU cluster is not a virtualization problem. It's a fabric problem, a networking problem, a storage throughput problem. Teams that try to solve it with the same hypervisor layer they use for their ERP system are going to be confused about why their utilization numbers look wrong.

The skills gap here is specific and worth naming directly.

Most senior infrastructure engineers built their careers on hypervisor administration. Those skills are still valuable for the environments that still run on hypervisors. But they don't transfer automatically to Kubernetes operations, bare metal provisioning at scale, or the GPU fabric management that AI infrastructure requires. The engineers who stay relevant through the next five years are the ones who treat their hypervisor expertise as one tool in the kit, not the kit itself.

Practically, that means getting hands-on with Kubernetes — not just deploying it, but operating it under load, managing persistent storage for stateful workloads, and understanding where it breaks. It means being able to stand up and tear down bare metal infrastructure without a GUI. It means having a conversation about why a GPU workload's performance profile looks different on virtualized versus bare metal infrastructure and what to do about it. Certifications are worth having as a signal, but the organizations hiring for roles that matter in 2027 and 2028 are looking for people who have operated these environments, not people who passed a test about them.

One more thing. Gartner predicts 35% of VMware workloads will migrate to alternative platforms by 2028. The switching cost is still high — tooling, training, operational process, third-party integrations — but it is no longer prohibitively high. Organizations that start evaluating alternatives now, even without an immediate migration plan, will be in a fundamentally different position at their next renewal than the ones that don't. SoftwareSeni

The invoice already came. The question is whether the next one catches you in the same position.

The platform is still useful. The dependency is the problem.

This blog is written by AI and concepted by a human.