Why cloud migrations come in over budget on run cost
The migration project usually lands close to estimate. The monthly bill afterwards is where the surprise arrives.
By Vilindor Cloud Practice
Migration project costs are relatively easy to estimate because the work is bounded and visible. Run cost is harder, and it is where most cloud programmes disappoint their sponsors — not because cloud is expensive, but because lift-and-shift sizing is nearly always wrong in the same direction.
The main causes, in rough order of impact
- Sizing cloud VMs to match on-premises hardware that was itself over-provisioned five years ago
- Running non-production environments twenty-four hours a day when they are used eight
- Choosing premium storage tiers by default for workloads with no performance requirement
- Paying on-demand rates for steady-state workloads that should be on reservations or savings plans
- Egress charges that nobody modelled because on-premises networking was a sunk cost
- Orphaned resources — unattached disks, idle IPs, forgotten test environments
The on-premises server you are replacing was sized for a five-year peak. The cloud instance replacing it can be resized next Tuesday. Stop buying the five-year peak.
Fix it in design, not in a later optimisation project
Right-sizing after the fact is possible but politically harder — it requires admitting the original sizing was wrong, and it usually means downtime to change instance families. Collecting genuine utilisation data before the migration and sizing to observed load plus headroom is the same work done at a fraction of the cost.
Put a name against the bill
Tag everything at deployment time and make one person accountable for reviewing spend monthly. Cloud cost decays exactly like alert configuration does: gradually, invisibly, and only when nobody owns the number.