IT infrastructure review before a budget cycle: facts for a defensible 2027 budget line

What an IT infrastructure review before the budget collects

An IT infrastructure review before the budget cycle collects the facts that let someone outside IT check a budget line: which releases lose vendor support in the budget year, which licences and contracts renew, how much of the installed capacity is in use, where backup and recovery have gaps, and what it costs to leave everything as it is. For the 2027 IT budget the fixed dates include 12 January 2027 for Windows Server 2016, February 2027 for Veeam Backup & Replication 12 and 11 October 2027 for vSphere 8. Each line of the request then names its trigger, the source of its figure and the decision it supports.

One sentence of a budget request shows the difference. "The virtualisation hosts are old" opens a discussion about how old is too old. "General support for the installed hypervisor release ends on 11 October 2027, and the vendor writes that after this day there will be no technical support and no new patches" names a day, a document and a consequence. This guide goes through the facts in the order in which they can be collected. What an audit of the whole estate finds beyond them is the subject of our companion article on the independent IT infrastructure audit.

Support end dates in the 2027 budget year

documents read on 12 November 2026, except the VCF 9.0 date, which we take from our published article on vSphere 8 (not re-read on that day). Veeam's page gives months and no days.

Windows Server 2016 comes first in the year. The vendor's release information page prints 12 January 2027 as the end of extended support; mainstream support ended in January 2022. A server still on this release in January is therefore a line in the 2027 request or a decision already taken in 2026.

Veeam's lifecycle page gives months and defines the month as "the last month when the relevant service is offered". For Backup & Replication 12, released in February 2023, it lists End of Fix in November 2025 and End of Support and Security Fix in February 2027. A backup server on version 12 belongs at the front of the plan, together with a check of which platform versions the next release supports; our article on backup for VMware after Broadcom covers that side.

RHEL 9 needs a careful reading. Red Hat's life cycle data shows Full support until 31 May 2027 and Maintenance support from 1 June 2027 to 31 May 2032. The date marks a change of phase, and RHEL 9 does not reach an end of support in 2027.

For vSphere 8, Broadcom's KB 459135 states: "The End of general support for vSphere 8.0 is October 11, 2027." KB 456708 lists the same day for VCF 5.2.x, NSX 4.x, HCX 4.x and VMware Live Recovery 9.0.x, among others, and adds that after it "there will be no Technical Support and no new patches for these product versions." For customers in the European Economic Area, Broadcom's VCF blog of 4 November 2026 keeps vSphere Standard 8 and Enterprise Plus 8 on sale until 31 October 2027 and includes Extended Support in active subscriptions to them until the subscription ends or 31 October 2028, whichever is earlier. The upgrade target has a date of its own, as the VCF 9.0 row shows. The details and the upgrade order are in our guide to vSphere 8 end of general support, and the dates of the 9.x releases in vSphere 9 end of support dates.

For servers and storage the dates that count are in your support contracts with the makers. The list needs, for each system, the end of that contract and whether the hypervisor release planned for it still supports its processors and devices. The generations concerned are listed in our guide to ESX 9 hardware requirements, and the replacement itself is described in VMware host refresh.

Dates in 2028 and 2029 that a model over three to five years reaches

Broadcom's VCF blog of 13 October 2026 (VCF 9.1) and of 4 November 2026 (EEA), both as cited in our published vSphere articles and not re-read on 12 November 2026; Veeam's product lifecycle page; the vendor's "Windows Server release information" (28 October 2026); Red Hat's product life cycles data; the last three read on 12 November 2026.

A financial model over three to five years runs past 2027, so the step chosen now is checked against its own end date. A move from Windows Server 2016 to 2019, for example, returns as a line inside the model period. The release information page gives 14 October 2031 for Windows Server 2022 and 14 November 2034 for Windows Server 2025, and Red Hat's data shows Full support for RHEL 10 until 31 May 2030.

Renewals that fall due in the budget year

Every subscription and maintenance contract with an end date in the budget year is a line, whether anything else changes or not. The review lists each one with its end date, its metric, the quantity under contract and the quantity in use. For VMware the metric is the core: Broadcom's KB 313548 sets a minimum of 16 licensed cores per CPU, so the count follows from the hosts and their processors. Which tier fits which estate is in our VMware renewal guide, where the end dates are shown in how to check VMware licence expiry, and what to compare in the quote in our renewal quote checklist.

The order of work matters here. A host that can be removed before the renewal lowers the number of cores to license, so capacity is measured before the renewal is priced.

Licences and contracts are one of the six domains of our audit, examined as editions and subscriptions, renewal dates and alignment with actual usage; send us the renewal dates you know and the versions you run, and we reply within one business day.

Capacity in use, measured before anything is bought

Installed capacity is in the inventory. Used capacity is in the performance history of the hypervisor and the storage systems. The review needs the peak over a period long enough to contain month-end and other regular loads, and the reserve each cluster must hold to carry the loss of a host. The result sorts hosts into three decisions: buy, defer or remove.

Our guide to vSphere capacity planning builds a three-year forecast from 90-day peaks, and host consolidation under per-core licensing shows how far a cluster can shrink. Budgets for GPU servers start from models and users and follow their own order; they are covered in AI hardware budget for 2027.

Backup and recovery gaps that turn into budget lines

Two figures decide this part: when a restore of each critical system was last tested, and how long it took compared with the time the business has agreed to wait. Where no test is on record, running one before the budget is submitted produces the figure. NIST's Cybersecurity Framework 2.0 puts the expectation into subcategory PR.DS-11: "Backups of data are created, protected, maintained, and tested".

A gap can lead to three kinds of line: storage for the retention the data owners expect, capacity at a second site, and time and capacity for tests. How a recovery site is sized is in our guide to DR site sizing. A way to find gaps in the plan with the team at one table is the tabletop exercise for IT outages.

IT budget planning by question, data source and decision

our working table for the review; it contains no external figures.

A review sequence for the weeks before the budget

1. Set the budget year and the model period of three to five years, and note the day on which the request must be in.
2. Pull the inventory: systems with versions, cores and purchase dates; licences and contracts with their end dates.
3. Put the vendors' dates beside the version list and mark everything that ends inside the budget year or the model period.
4. Read the capacity figures from the performance history and note the peak and the failover reserve per cluster.
5. Collect the records of restore tests; where a critical system has none, run a test.
6. Price the estate as it is over the model period. This is the baseline called "change nothing".
7. Build the scenarios against that baseline and write each budget line with its trigger, source and decision.

The steps stand in this order because each one uses the output of the step before it. Renewals are priced after capacity, and scenarios after the baseline. For a vSphere estate, the exports behind steps 2 to 4 are listed in our article on what a VMware health check covers.

TCO analysis of IT infrastructure for modernisation, cloud, hybrid and no change

Gartner's glossary defines total cost of ownership as "a comprehensive assessment of information technology (IT) or other costs across enterprise boundaries over time." For IT, it continues, TCO "includes hardware and software acquisition, management and support, communications, end-user expenses and the opportunity cost of downtime, training and other productivity losses." The definition names cost categories and no period. The horizon of 3 to 5 years is the one our audit uses for its financial model.

The European Commission's page on life-cycle costing for green public procurement describes the same idea: "LCC means considering all the costs that will be incurred during the lifetime of the product, work or service". For how such approaches can be used, the page refers to Article 68 of Directive 2014/24/EU. The Commission's calculation tools cover computers and monitors and four other kinds of product, none of them servers, so an infrastructure model is assembled from the elements by hand.

Gartner IT glossary, "Total Cost of Ownership (TCO)"; European Commission, Green Public Procurement, "Life-cycle costing"; both read on 12 November 2026. The examples in the third column are ours.

Each scenario is priced with the same elements over the same period. Modernisation keeps the systems on your premises and renews hardware, releases and licences. Cloud moves workloads to a provider and exchanges purchases for running fees, and hybrid splits the estate between the two; the hosting models are compared in IaaS vs private cloud vs hybrid cloud. Changing nothing is the fourth column, and it has entries of its own: the renewals that fall due anyway, support for releases past general support where the vendor offers it, hardware outside the maker's support, and downtime, which Gartner's definition counts as a cost. For vSphere 8 the entry depends on region and edition, as the Extended Support rule above shows. As of 5 November 2026 we found no public terms for what this Extended Support covers, so ask for them in writing before the column is priced.

Our audit builds this comparison as an ROI/TCO financial model over 3 to 5 years and presents it to your management; write to us with your budget date and the decisions that are open.

What an audit adds to an internal review

The review described here can be done by the IT team with its own data. An audit is the larger piece of work. It records the estate in a full inventory of infrastructure, licences and access across up to six domains, and it orders what it finds in a prioritised risk map, so that the budget carries the critical items first. It also comes from outside: our audit is priced and delivered independently of any purchase, and the roadmap with stages and a budget estimate belongs to you, to carry out with us, on your own or with another contractor. We work from the documentation and exports you provide and from interviews with your team, without access to your production systems, and scope and price are fixed in writing before the start.

FAQ