Ask whether the cloud is cheaper than on-premises and the honest answer is: it depends on what you count. Most comparisons fall apart at the first step, because they put a monthly cloud bill next to the purchase price of a server — two numbers that don’t measure the same thing.
To compare fairly, you need the full built-up cost of each option over the same period, including the costs that never appear on an invoice. Here’s how we work it out for clients, and a formula you can use yourself.
Why the sticker price misleads
A server you buy outright looks cheap the day you sign the order, because the price tag ignores the three or four years you’ll spend powering, cooling, patching, securing and eventually replacing it. A cloud bill has the opposite problem: it’s gloriously transparent about the monthly figure, but easy to let sprawl, and it quietly adds extras such as data egress and premium support.
Compare like with like and the picture changes completely. That means counting everything, on both sides, over the same time horizon.
Building up the on-premises cost
On-premises spend splits into three buckets. The trick is to spread the one-off costs across the life of the kit, then add the annual running costs.
- Capital (one-off): servers, storage, networking, firewalls, software licences and any data-centre or comms-room fit-out. Spread this over the refresh cycle — typically three to five years.
- Running (annual): power and cooling, rack or floor space, hardware maintenance and support contracts, connectivity, plus backup and disaster recovery.
- People and risk (annual): the engineer time spent patching, monitoring and firefighting, the cost of downtime when something fails, and the eventual cost of decommissioning and disposal.
The people bucket is the one most businesses forget, and it’s rarely small.
Building up the cloud cost
Cloud has its own hidden corners, so build it up the same way:
- Consumption (monthly): compute (cheaper if you commit to reserved or savings plans rather than on-demand), storage, and data egress — the cost of getting data out, which catches people out.
- Licensing: whether you bring your own licences or pay for them inside the service.
- Platform and support: managed services, and the support plan you actually need rather than the free tier.
- One-off and ongoing: the migration project itself, spread over your time horizon, plus the ongoing optimisation effort to stop spend drifting upward.
The comparison formula
Normalise both options to an annual cost over the same N-year horizon, then divide by something meaningful — workloads, or users — so you’re comparing a rate, not a lump sum:
On-prem annual TCO = (Capex ÷ N) + annual running + people + risk allowance
Cloud annual TCO = (12 × monthly consumption) + (migration ÷ N) + management
Compare: Annual TCO ÷ workloads → £ per workload per yearTry it: the comparison calculator
The maths is simple once it’s all in one place. Plug your own figures into the calculator below — it builds up the full annual cost of each option and compares them like for like. It’s pre-filled with illustrative numbers to show the method; change them to match your estate.
Cloud vs on-premises cost calculator
Enter your own figures to compare the full annual cost of each. Pre-filled with illustrative numbers — change them to match your estate. Nothing is sent anywhere; it all runs in your browser.
On-premises — annual TCO
£47,000/yr
£4,700 / workload / yr
£141,000 over 3 yrs
Cloud — annual TCO
£45,600/yr
£4,560 / workload / yr
£136,800 over 3 yrs
On these figures, Cloud works out £1,400/yr cheaper — about 3%.
A guide, not a quote. The right answer also depends on utilisation, growth, resilience and opportunity cost.
On the illustrative defaults it’s close — and that’s exactly the point. The decision rarely turns on the headline numbers; it turns on the factors below.
The factors that actually decide it
- Utilisation. Cloud shines for variable, spiky or unpredictable load, because you only pay for what you use. Steady, predictable, high utilisation often still favours on-premises.
- Growth. If you’re scaling fast, paying for peak capacity up front in tin is wasteful. If you’re flat, you may be paying a premium for elasticity you never use.
- Resilience and risk. Building genuine redundancy on-premises is expensive; in the cloud a lot of it is included. Put a real number on the cost of an outage.
- Opportunity cost. Capital tied up in hardware is capital not invested in the business.
So which is cheaper?
Usually, it isn’t all-or-nothing. For a lot of UK businesses the lowest true cost is a hybrid approach: steady, predictable workloads stay on-premises or in a private cloud, while variable or growth workloads run in public cloud such as Microsoft Azure or AWS. A hybrid cloud model lets you put each workload where it’s genuinely cheapest to run.
We’re vendor-independent, so we’ve no reason to push you one way or the other — we’d rather help you do the maths properly and land on the answer that’s right for your business. If you’d like a hand building up the numbers for your own estate, get in touch. We’re happy to give honest, no-pressure advice.
