The server in the technical room has an accounting superpower: once you have paid for it, it looks free. It appears on no monthly invoice, so nobody sees it any more — not the electricity it draws day and night, not someone's hours updating it, not the day it will stop. The cloud has the opposite problem: its cost arrives monthly, in plain sight, and that is why it always looks expensive by comparison.
Both impressions lie. The honest comparison between “servers at our company” and “servers rented from the cloud” is made on the total cost of ownership, over a full equipment life cycle — that is, over five years. This article gives you all the lines of the spreadsheet, including the ones each camp conveniently omits. No amounts: your figures come from the quotes you request; here you get the structure that makes them comparable.
Why 5 years, not 1 year
A physical server is bought once and depreciated over its useful life — after which the cycle starts again: the equipment ages, the warranty expires, parts become hard to find, and the risk of failure rises exactly when the data on it is at its greatest. Any one-year comparison catches the cloud with twelve invoices and the local server with zero — and it is as false as a comparison made in the month of purchase, which shows the opposite. Five years cover a complete cycle: purchase, operation, ageing and replacement.
The full cost of the on-premises server
The lines you put in the “local” column — including the ones nobody puts in:
- The equipment. The server itself, the disks, the network equipment — sized for the activity peak of year five, not the average of year one. That over-sizing is a real cost of the model: you pay today for capacity you may only need in three years' time.
- The licences. The server operating system, the access licences, the backup and protection software — some with annual costs, not one-off.
- Electricity and cooling. A server runs non-stop, 8,760 hours a year, and in summer it needs cooling too. Ask your electrician or supplier for the estimated consumption in kWh and multiply by your tariff — it is a line that five years turn from “negligible” into visible.
- The UPS and the room. The uninterruptible power supply (with its batteries, replaced periodically) and a locked, ventilated room that can no longer be anything else.
- Maintenance. Security updates, monitoring, checking the backups — hours of work, month after month, in-house or contracted. The server “nobody touches” is not cheap, it is unpatched.
- Off-site backup. The safety copy kept next to the server burns together with it; an external one means extra equipment or a subscription.
- Downtime on failure. When the local server goes down, the company waits until the part or the technician arrives; factor in at least the risk, if not an amount.
- Replacement. At the end of the cycle, you repeat the purchase — the five-year sheet contains at least the provision for it.
The full cost of the cloud
The “cloud” column is not just the subscription either:
- The monthly subscription — for the resources consumed or per user, depending on the service. The structural advantage: you pay for what you use now, not the peak of three years from now; you scale up and down without purchases.
- Growth over time. Data accumulates, and the invoice grows gradually. Budget the trend, not the first month's value.
- Data transfer. Some services charge for taking data out of the cloud; ask explicitly what it would cost to leave with all your data — the answer matters commercially, not just technically.
- The internet becomes critical. Without a connection there is no “server”. A second internet line, from another provider, becomes practically mandatory — it is a cost line of the cloud, even if the invoice does not say “cloud” on it.
- The initial migration. Moving the applications and the data is a one-off project, with its own cost — ask for it itemised in any offer.
- Configuration and supervision. The cloud moves the work from the screwdriver to the console, but does not eliminate it: permissions, security, cost monitoring. The surprise cloud invoice is almost always a resource someone left running — monthly supervision is a cost line too.
What disappears completely from the cloud column, on the other hand: electricity, cooling, the UPS, the technical room, spare parts and the replacement provision — the equipment risk is the provider's problem.
The 5-year spreadsheet
Put the two columns side by side, over 60 months, with the lines above. Two rules make the difference between a calculation and a guess:
- The same requirements in both columns. Request quotes for exactly the same needs — the same applications, the same users, the same level of backup and availability. The quotes become the basis of the calculation; any figure before them is indicative.
- Add the risk row. What happens in each scenario on a major failure, on a one-day internet outage, when the number of users doubles? Not everything has an exact price, but everything has consequences — and they belong in the comparison.
Who wins in which scenario
The on-premises server remains the rational choice where an old but essential application does not run in the cloud; where the internet in the area is weak or unstable; where production machinery demands an instant response from a server right next to it; or where contracts or regulations require keeping data in a specific place.
The cloud usually wins for teams spread across several sites, for companies growing unpredictably, wherever there is neither a technical room nor a dedicated person — and anywhere continuity matters: a virtual server restarts somewhere else, yours in the closet does not.
In practice, the answer for many companies is hybrid: the production application stays local, while email, files and backup move to the cloud. It is not indecision — it is often precisely the economic optimum. The typical questions of this transition are gathered at the frequently asked questions about cloud and migration.
Before you decide: do you know what you have?
The comparison assumes you know exactly what runs in the company today — which applications, on which equipment, how old and with what dependencies between them. Surprisingly many companies do not, and a spreadsheet built on assumptions produces false results with maximum confidence. If that is your situation, start with a systematic inventory — we described in the article about the IT audit exactly what gets checked and what you receive at the end.
The next step
Build the spreadsheet with the lines in this article and request two quotes on the same requirements — one for the local scenario, one for the cloud. If you want someone to go through the calculation with you, with the advantages and the risks laid on the table without bias towards either option, Neoxis carries out such assessments as part of its IT consulting services.