Data center TCO over 1, 3 and 5 years: reading cost through time
Asking "what does this hosting cost?" without stating the horizon is an incomplete question: the same solution is expensive at one year, competitive at three, unbeatable at five, or the reverse. Total cost of ownership (TCO) reads through time.
Why the horizon changes the answer #
Each model spreads its costs differently over time. Colocation concentrates spending at the start, hardware, setup, migration, then costs little each month. Cloud asks nothing upfront and bills every month. Comparing without a horizon is like comparing a bought house and a rented one on the first quarter alone.
What weighs at 1 year #
One-off costs dominate: hardware purchase or refurbishing, setup fees, cabling, migration, the project's human time. At this horizon, solutions with no initial investment, cloud or rented servers, almost always look like winners. It is the horizon that flatters renting.
What weighs at 3 years #
Entry costs are amortized; recurring costs take over, and energy often becomes the first line, ahead of rent. It is the horizon of truth for a stable load: where ownership in colocation crosses and overtakes usage-based renting, the famous tipping point.
What weighs at 5 years #
Two events enter the scene: hardware refresh, as gear reaches end of life, and contract renegotiation, with its exit terms. A 5-year TCO must provision a partial refresh and read the exit clauses. It is also the horizon where indexation clauses, endured or negotiated, make the difference.
Building your table, line by line #
- Space and energy: rent, consumed kW × PUE × kWh price, expected indexation.
- Connectivity: transit, cross-connects, inter-site links.
- Hardware: purchase or rental, warranties, refresh provision at 4-5 years.
- Services: setup, remote hands, managed services if any.
- People: internal operating time, often forgotten, never free.
- Exit: termination, decommissioning, future migration costs.
Stress-test before deciding #
A good TCO gets stressed: what happens to the gap if the kWh price rises 20%, if the load doubles, if the project stops at two years? The solution that stays reasonable in all three scenarios is the right one, even if it wins none outright. Robustness beats a fragile optimum.
FAQ #
Three years is the reference: it amortizes entry costs without too many unknowns, and matches the life of hardware and commitments. Then check the decision also holds at one year (cash flow) and five years (refresh).
Yes, otherwise the comparison is skewed toward ownership. Internal operating hours have a real cost; so do the managed services that replace them. Count both honestly.
It depends on utilization: the more stable and continuous the load, the earlier it comes, often before three years. A highly variable load may never tip, and that is an equally useful answer.
Written on 1 September 2026.
From reading to comparing: relevant data centers
Estimate my cost → Compare data centers