Twenty years ago, virtualising your servers was a project with a business case, a payback period and a board paper. Today it is an assumption. Almost every organisation runs virtualised workloads, so virtualisation on its own says little about the quality of an infrastructure estate.
A more useful measure is what each level of maturity looks like in practice.
Why is infrastructure strategy a live question again?
Two things have made infrastructure strategy an active topic rather than a settled one.
The first is commercial. Changes to hypervisor licensing and packaging have materially altered the cost base for many organisations. European cloud providers told the European Commission in 2025 that VMware price rises after Broadcom’s acquisition ran from 800 to 1,500 per cent for some of their members, as reported by The Register. Increases vary widely by customer, but the effect has been to reopen decisions left alone since the original virtualisation programme, and to make alternatives, including other hypervisors, containers and public cloud, worth costing properly.
The second is operational. Recovery from a cyber attack, rather than from hardware failure, is now the design case. Workloads move between on-premises and cloud more often. And the business expects new environments in hours rather than weeks.
A five-level infrastructure and virtualisation maturity model
Level one: physical and fragmented
Some virtualisation, but significant physical estate remains, often because an application vendor will not support anything else. Provisioning is manual and slow. Nobody has a complete inventory. Recovery is by tape and untested. This is rarer than it was, but it persists in pockets, usually around a legacy line-of-business application.
Level two: virtualised and consolidated
The estate is largely virtualised, consolidation ratios are decent and hardware utilisation is reasonable. But provisioning is still a ticket and a person, configuration drifts because changes are made in a console, and disaster recovery is documented rather than demonstrated. This is a functional place to be. It is also where the cost of change is highest, because nothing is described in a form you can reproduce.
Level three: standardised and monitored
Templates and standard builds are used consistently. There is a real inventory, and monitoring covers capacity as well as availability. Patching is systematic. Disaster recovery (DR) is tested at least annually. Provisioning is faster but still fundamentally manual.
Level four: automated and code-defined
Infrastructure is described as code and provisioned from a pipeline. Configuration drift is detected and corrected automatically. Environments can be rebuilt from source, which is what makes both DR and cyber recovery credible. Full observability is available, and workload placement is a deliberate decision against defined criteria rather than a default.
Level five: portable and cost-optimised
Workloads are placed on the basis of measured cost and performance, and can move when the economics change. Capacity is managed against forecast rather than reactively. Resilience is designed per workload rather than applied uniformly. Few organisations of any size are here, and for many it is not worth the investment.
How do you work out where your estate really sits?
Maturity models invite optimism, so a few concrete tests help.
- Can you rebuild your most important server from source, without a person remembering anything? If not, you are below level four whatever tooling you own.
- When did you last complete a full DR test, and what broke? “We tested it and nothing broke” usually means the test was too easy or not representative.
- How long does it take from request to a running environment? Measure it on real tickets rather than estimating.
- Do you know your cost per workload? Not total infrastructure spend, but what a specific application costs to run. The answer shows how mature your level of observability is.
- How many configuration exceptions exist, and does anyone know why each one was made?
What moves you up a level?
Levels two and three are mostly a documentation and discipline problem, and level four is mostly an automation problem. Very little of it is solved by buying something.
From two to three: build a real inventory, standardise builds into templates, get patching onto a schedule with evidence, and run one realistic DR test.
From three to four: describe the estate as code, starting with the environments that change most often. Infrastructure as code is usually the most valuable change at this stage, because it improves provisioning speed, reduces drift and makes recovery reproducible at the same time.
From four to five: measure cost properly, per workload, and start making placement decisions from data.
“Almost every infrastructure conversation we have starts with a technology question and ends up being about reproducibility,” says Tom Richards, Lead, Systems and Storage at Northdoor. “If you can rebuild it from a definition, you can recover it, you can move it, and you can cost it. If you can’t, every one of those becomes a project.”
How does virtualisation strategy fit with cloud?
Modernising infrastructure and moving to cloud are not the same thing, and treating them as synonyms leads to poor decisions in both directions.
Some workloads belong in public cloud and are worse off on-premises. Some are the reverse, particularly steady-state, high-throughput, licence-heavy workloads where the cost over the hardware’s life can be lower on owned infrastructure. Many organisations should expect a mixed estate for the foreseeable future, and the maturity that matters is the ability to place each workload deliberately rather than by default. Our article on choosing between on-premises, cloud and hybrid storage sets out the same logic for data, and our cloud and on-premises infrastructure page covers how we approach it.
What this means for your organisation
For most mid-sized organisations, level four is the right target, while level five risks becoming over-engineering. Fast provisioning, low drift, credible recovery and enough cost visibility to place workloads effectively are achievable within 12 to 18 months from level two, without the need for a major capital programme.
If you would value an independent view of where your virtualisation estate sits today, our infrastructure team can help assess its maturity, identify the most valuable opportunities for improvement and define a practical path forward. See our managed infrastructure services or talk to us to discuss your current environment.