Cloud migration estimates are typically 40–60% understated. The compute and storage line items everyone budgets for are usually accurate. The line items below are not. Each has a cost range, severity, and a mitigation playbook.
| # | Cost category | Range | Severity |
|---|---|---|---|
| H01 | Data Egress During Migration | $50–$92 per TB | Critical |
| H02 | Cutover Downtime | Revenue × downtime hours | Critical |
| H03 | Parallel Running Costs | 1.5–2.5× monthly infra spend | High |
| H04 | Staff Retraining | $1,000–$5,000 per technical employee | High |
| H05 | Application Re-architecture Labour | $25,000–$200,000 per app | High |
| H06 | Right-Sizing Over-Provisioning | 20–40% of first 3–6 months cloud spend | Medium |
| H07 | Licence Changes | $50,000–$500,000+ for Oracle/SQL | High |
| H08 | Cloud Support Plans | 8–10% of monthly cloud spend | Medium |
| H09 | Security and Compliance Rework | $20,000–$150,000 | Medium |
| H10 | Vendor Lock-in Exit Costs | $100,000–$5,000,000+ to switch later | Medium |
Moving data out of your current environment (or between clouds) triggers egress fees. AWS charges $0.09/GB for the first 10TB. A 50TB migration costs $4,500 in egress alone, before any other costs.
Mitigation
Use AWS DataSync or Azure Data Box for large transfers. Physical shipping of data (AWS Snowball) eliminates egress for datasets over 10TB.
Every hour offline during cutover costs money. For an e-commerce site doing $100K/day, a 4-hour cutover window costs $16,700 in lost revenue. Factor in SLA penalty exposure too.
Mitigation
Plan blue-green deployments or phased cutovers. Run parallel systems for at least 48 hours before DNS switch. Test rollback procedures obsessively.
During migration you run both on-premise and cloud environments simultaneously. A typical 6-month migration with $10K/month on-premise spend adds $60,000–$150,000 in parallel running costs.
Mitigation
Minimise parallel running window with a phased cutover. Negotiate to delay on-premise contract renewals. Right-size cloud resources before the migration to reduce cloud spend during this phase.
AWS, Azure, and GCP certifications cost $300–$400 each. Training time is 2–4 weeks per engineer. For a 10-person engineering team, budget $10,000–$50,000 for training alone.
Mitigation
Identify cloud champions early. Use provider training credits (AWS gives $500–$1,000 per certification attempt). Stagger training before, during, and after migration.
Apps that seemed simple on-premise often need significant rework for cloud. Configuration management, secret handling, logging pipelines, networking, these all need rebuilding.
Mitigation
Audit apps before migration. Classify each by refactor complexity. Plan refactor work separately from migration work to avoid scope creep blowing the budget.
Teams routinely over-provision cloud instances in the first 3–6 months. A common pattern: lift-and-shift a 32-core on-premise server to a comparable cloud instance, then discover actual utilisation is 15%.
Mitigation
Use AWS Compute Optimizer, Azure Advisor, or GCP Recommender from day one. Schedule a right-sizing review at 30, 60, and 90 days post-migration.
Oracle Database on AWS EC2 requires bringing your own licence (BYOL), Oracle's cloud licence terms can double or triple your Oracle costs. SQL Server on Windows EC2 is expensive; SQL Server on Linux EC2 is cheaper.
Mitigation
Audit all software licences before migration. Evaluate managed alternatives (AWS RDS, Azure SQL Managed Instance) which often include licensing. Consult a licence specialist before moving Oracle workloads.
AWS Enterprise Support is 10% of monthly spend (min $15,000/month). Azure and GCP have similar structures. Many teams underestimate this until their first month's invoice arrives.
Mitigation
Budget for Business or Enterprise support from day one if you're running production workloads. Factor it into your TCO comparison, on-premise support costs exist too but are often hidden in staff time.
Cloud security posture is different from on-premise. IAM, VPC configuration, encryption at rest and in transit, compliance controls (SOC2, PCI, HIPAA) all need cloud-specific implementations.
Mitigation
Engage a cloud security specialist early. Use AWS Security Hub, Azure Security Center, or GCP Security Command Center from day one. Budget 10–20% of project cost for security work.
The decision you make today about cloud-native services (AWS Lambda vs ECS, Azure Cosmos DB vs AWS DynamoDB) creates future switching costs. Egress alone for a 1PB dataset from AWS would be $92,000.
Mitigation
Use open standards where possible (Kubernetes over ECS, open-source databases over proprietary). Design for portability in strategic apps. Accept lock-in pragmatically for commodity services.
The 25% buffer rule
Add 25% to any quoted vendor or partner number for predictable hidden costs (egress, support, retraining). This is a floor, not a ceiling.
Separate predictable from unpredictable
Egress, parallel running, support, retraining are all predictable. Re-architecture and rollback are not. Budget the predictable line items, hold contingency for the rest.
Build contingency into the business case
A 15–25% rollback / re-architecture contingency, signed off in advance, prevents programme stalls when something inevitably surprises you.
The roadmap calculator includes the predictable hidden costs above. Plan the source-to-destination transition with a number you can take to the board.
Run the migration roadmap →Updated 2 May 2026