The Real Cost of "Cloud vs On-Prem": Why TCO Models Get the Headline Number Wrong
A five-year total cost of ownership model can be built two defensible ways and land on opposite conclusions. That's not a spreadsheet problem — it's a governance problem.
A better question than "which is cheaper"
Cost matters, but it shouldn't be the first gate. The organisations that make this decision well tend to run it through a staged framework before cost is allowed to dominate the conversation:
Why the same numbers can point two ways
A full-loaded TCO model accounts for everything: existing hardware refresh cycles, the labour cost of migration itself, contract exit penalties, retraining, parallel-running periods, and the "long tail" of legacy applications that don't move cleanly. Modelled this way, on-premises refresh can come out marginally cheaper over five years — because the cost of change is real and front-loaded, while cloud's efficiency gains compound slowly.
Strip the model back to an executive-ready comparison — steady-state run costs, headline licensing, straight-line depreciation — and cloud often wins instead, sometimes by a wide margin, because the model is no longer carrying the transition cost or the legacy drag.
Both models are legitimate. The mistake is presenting either one as the answer without naming which question it's answering.
Three things that quietly decide the outcome
What counts as sunk cost. Recently purchased infrastructure — a baseline refresh, a hardware investment mid-life — gets treated inconsistently across models. Whether it's carried as a cost in both scenarios or written off in neither changes the comparison materially, and this decision is often made implicitly rather than agreed up front.
The legacy contract tail. Large IT estates typically carry 30–40 active licensing and support contracts, a meaningful fraction of which are candidates for retirement, renegotiation, or consolidation regardless of the infrastructure decision. Rolling those savings into only one scenario skews the comparison before the cloud-versus-on-prem question is even asked.
Where the migration cost lands. The labour, risk, and disruption of the move itself is a real cost of the cloud scenario — but it's a one-off, not a steady-state cost. Amortising it correctly (or being transparent about not amortising it) is the difference between a model that informs a five-year decision and one that just describes year one.
Every infrastructure transformation program eventually arrives at the same slide: a total cost of ownership comparison, one column for staying on-premises, one for moving to the cloud, five years across the top, a number at the bottom that's supposed to settle the argument.
It rarely does. Not because the modelling is sloppy, but because "cloud vs on-prem" is not actually one question. It's a bundle of decisions — about workload placement, sunk investment, licensing exposure, operational maturity, and risk appetite — dressed up as a single dollar figure. When organisations skip the unbundling and go straight to the headline number, they get a TCO model that's technically correct and strategically useless.
This ordering matters because cost models are seductive — they produce a single number that feels objective, which makes it easy to skip the harder, more qualitative gates. A workload with poor strategic fit can still show a favourable TCO and still be the wrong thing to migrate.
The governance point underneath all of this
None of this is really about spreadsheets. A TCO model is a proxy for a governance decision: who gets to define the scope of the comparison, and what assumptions they're allowed to bake in before anyone sees the output. Getting the model "right" matters less than making that scoping decision visible, contestable, and owned by someone accountable for the outcome — not buried in a tab labelled "assumptions" that nobody outside the finance workstream reads.
For security and infrastructure architects, this is also where risk should enter the conversation — not as a footnote after the cost decision, but as an input to it. Migration risk, legacy application fragility, and operational maturity gaps are all cost drivers in disguise; they just don't show up until year two or three, by which point the original TCO slide has usually been forgotten.
Navigating a cloud or infrastructure transformation decision?
Arkovis helps government, critical infrastructure, and enterprise organisations build TCO and migration decision frameworks that hold up to board scrutiny.
© 2026 Arkovis Pty Ltd | ACN 699 673 526 | ABN 25 699 673 526