The 100-server cloud migration is the most quoted reference point in cloud migration cost modelling. It is the smallest scope that exercises the full migration discipline (wave planning, partner-led labour, parallel running, programme governance) and the largest scope that most mid-market organisations face. This page is a fully worked 100-server scenario for 2026, with all line items, all three major hyperscalers compared, and the 3-year TCO comparison versus staying on-premise.
The scenario used throughout this page: 100 servers on-premise in a colocation facility, mixed Windows and Linux estate (60 Windows, 40 Linux), 20 TB total data across various databases and file shares, 12 production applications and 8 supporting applications, no mainframe component, no SAP component, no regulated workloads (no HIPAA, PCI-DSS, or FedRAMP), 12-month migration timeline, partner-led migration, mid-market SI partner blend.
The strategy mix: 60 percent rehost (lift-and-shift) to cloud-native VM instances, 25 percent replatform (databases to managed database services, web apps to managed PaaS), 10 percent retire (workloads identified as no longer needed during assessment), 5 percent refactor (one high-traffic application rebuilt as cloud-native microservices). Wave structure: 6 waves of approximately 15 servers each over the 12-month programme.
Worked 100-server cost build, AWS destination, 12 months
| Cost line | Low estimate | Typical estimate | High estimate |
|---|---|---|---|
| Assessment and discovery (Migration Evaluator + partner) | $40,000 | $75,000 | $140,000 |
| Wave planning and PMO (12 months) | $180,000 | $280,000 | $420,000 |
| Migration labour, 90 workloads, blended | $580,000 | $1,140,000 | $2,100,000 |
| AWS MGN tooling (mostly free, EBS staging) | $8,000 | $15,000 | $25,000 |
| AWS DMS for 5 databases | $2,000 | $4,000 | $8,000 |
| AWS Snowball Edge (2 devices, 14 days) | $8,000 | $15,000 | $25,000 |
| AWS Direct Connect (12 months, 200 Mbps) | $18,000 | $32,000 | $60,000 |
| Parallel running, 5 months blended (on-prem + AWS) | $320,000 | $680,000 | $1,150,000 |
| Cutover and downtime contingency (per wave) | $50,000 | $120,000 | $320,000 |
| AWS Enterprise Support (12 months) | $110,000 | $170,000 | $220,000 |
| Staff retraining (8 engineers) | $22,000 | $42,000 | $75,000 |
| Security rework (IAM, VPC, encryption, Security Hub) | $45,000 | $110,000 | $260,000 |
| Refactor work (1 application to microservices on EKS) | $80,000 | $180,000 | $420,000 |
| Source colocation decommission | $10,000 | $25,000 | $50,000 |
| Contingency at 15 percent | $220,000 | $430,000 | $780,000 |
| AWS MAP partner funding | ($280,000) | ($450,000) | ($800,000) |
| Net AWS estimate | $1,413,000 | $2,868,000 | $5,253,000 |
Worked 100-server cost build, Azure destination, 12 months
| Cost line | Low estimate | Typical estimate | High estimate |
|---|---|---|---|
| Assessment and discovery (Azure Migrate + partner) | $35,000 | $65,000 | $120,000 |
| Wave planning and PMO (12 months) | $175,000 | $270,000 | $410,000 |
| Migration labour, 90 workloads, blended | $550,000 | $1,080,000 | $1,950,000 |
| Azure Site Recovery for 100 servers | $6,000 | $12,000 | $22,000 |
| Azure DMS premium for 5 databases | $4,500 | $9,000 | $18,000 |
| Azure Data Box (2 devices, 14 days) | $5,000 | $10,000 | $18,000 |
| ExpressRoute (12 months, 200 Mbps Metered) | $15,000 | $28,000 | $50,000 |
| Parallel running, 5 months blended (Hybrid Benefit applied) | $280,000 | $590,000 | $1,000,000 |
| Cutover and downtime contingency (per wave) | $50,000 | $115,000 | $300,000 |
| Azure Premier Support (12 months) | $100,000 | $155,000 | $210,000 |
| Staff retraining (8 engineers) | $22,000 | $40,000 | $72,000 |
| Security rework (Entra ID, network, Defender) | $40,000 | $100,000 | $240,000 |
| Refactor work (1 application to Azure Container Apps) | $75,000 | $170,000 | $400,000 |
| Source colocation decommission | $10,000 | $25,000 | $50,000 |
| Contingency at 15 percent | $200,000 | $385,000 | $700,000 |
| FastTrack value (Microsoft-funded) | ($60,000) | ($90,000) | ($120,000) |
| AMMP partner funding | ($240,000) | ($380,000) | ($650,000) |
| Net Azure estimate | $1,267,500 | $2,584,000 | $4,790,000 |
Worked 100-server cost build, GCP destination, 12 months
| Cost line | Low estimate | Typical estimate | High estimate |
|---|---|---|---|
| Assessment and discovery (Migration Center + partner) | $35,000 | $60,000 | $110,000 |
| Wave planning and PMO (12 months) | $170,000 | $260,000 | $400,000 |
| Migration labour, 90 workloads, blended | $540,000 | $1,060,000 | $1,920,000 |
| Migrate to Virtual Machines (free) | $0 | $0 | $0 |
| GCP Database Migration Service for 5 databases | $3,000 | $6,000 | $12,000 |
| Transfer Appliance (1 device, 14 days) | $3,000 | $6,000 | $10,000 |
| Cloud Interconnect (12 months) | $8,000 | $22,000 | $45,000 |
| Parallel running, 5 months blended (SUD applied) | $260,000 | $540,000 | $920,000 |
| Cutover and downtime contingency (per wave) | $45,000 | $110,000 | $290,000 |
| Premium Support (12 months) | $95,000 | $145,000 | $200,000 |
| Staff retraining (8 engineers) | $20,000 | $38,000 | $70,000 |
| Security rework (IAM, VPC, CSCC) | $38,000 | $95,000 | $225,000 |
| Refactor work (1 application to Cloud Run) | $70,000 | $160,000 | $380,000 |
| Source colocation decommission | $10,000 | $25,000 | $50,000 |
| Contingency at 15 percent | $192,000 | $370,000 | $680,000 |
| RaMP / partner funding credit | ($200,000) | ($340,000) | ($580,000) |
| Net GCP estimate | $1,289,000 | $2,557,000 | $4,732,000 |
The migration cost is one-time; the run-rate cost is forever. The 3-year TCO comparison includes 36 months of cloud run-rate plus the one-time migration cost. The comparison versus on-premise includes 36 months of equivalent on-premise cost.
3-year TCO comparison, 100-server mid-market estate
| Scenario | Migration cost | 36-month run cost | 3-year TCO | Versus on-prem baseline |
|---|---|---|---|---|
| Stay on-premise (no migration) | $0 | $5,400,000 | $5,400,000 | Baseline |
| Migrate to AWS (typical) | $2,868,000 | $4,200,000 | $7,068,000 | +$1,668,000 |
| Migrate to Azure (typical, Hybrid Benefit) | $2,584,000 | $3,800,000 | $6,384,000 | +$984,000 |
| Migrate to GCP (typical, SUD) | $2,557,000 | $3,650,000 | $6,207,000 | +$807,000 |
| Migrate with aggressive refactor | $3,500,000 | $3,200,000 | $6,700,000 | +$1,300,000 |
| Migrate with retire 25% | $2,200,000 | $3,150,000 | $5,350,000 | -$50,000 |
The 3-year TCO honesty check
The migration cost is dominated by three line items: migration labour ($1.0M to $1.2M of the typical column), parallel running ($540K to $680K), and contingency ($370K to $430K). Together these three lines account for roughly 75 percent of the total cost. Tooling cost (MGN, ASR, DMS, Snowball, Direct Connect) accounts for roughly 3 percent of the total. The cost concentration suggests where discipline matters: efficient migration labour, compressed parallel running, and realistic contingency planning produce on-budget outcomes.
The most common cost overruns on 100-server programmes, in order of frequency and magnitude:
The 100-server migration is the most quoted reference in cloud migration cost modelling because it exercises the full discipline at a tractable scale. The typical cost band of $1.4M to $2.9M is reliable as a starting point; the variance comes from strategy mix, destination choice, and parallel-running discipline. The 3-year TCO comparison is rarely a cloud-wins-by-50-percent story; for most mid-market estates it is a small positive on the 3-year horizon and a larger positive on the 5-year horizon. The case is strongest when aggressive retirement removes 20 to 25 percent of workloads from the migration scope entirely.
A. A typical mid-market 100-server cloud migration over 12 months costs $1.4M to $2.8M all-in including labour, parallel running, partner consulting, tooling, and contingency. The 3-year total cost of ownership (migration plus 36 months of cloud run-cost) typically lands at $4.5M to $8M depending on the workload pattern, destination cloud, and strategy mix.
A. Yes, but the difference is smaller than the marketing might suggest. AWS migrations typically cost 5 to 10 percent more than Azure for Microsoft-heavy estates because Azure Hybrid Benefit reduces compute cost. AWS and GCP are typically within 5 percent of each other for general-purpose workloads. GCP tends to win on data-heavy workloads thanks to BigQuery economics; AWS tends to win on partner ecosystem breadth. For a 100-server mid-market migration the destination cost difference is typically $100K to $300K, meaningful but not dominant.
A. Most 100-server mid-market migrations run 9 to 15 months from kickoff to full source decommission, with 6 to 10 months of active cutover work concentrated in waves. Larger waves of 15 to 25 servers each over 4 to 6 week cycles is the typical pattern. Wave-based migrations of this size produce on-budget outcomes more reliably than single-cutover migrations.
A. Most enterprise 100-server migrations use a mixed strategy: roughly 60 percent rehost (lift-and-shift), 25 percent replatform, 10 percent retire, 5 percent refactor. The mix optimises for cost: rehost is cheapest to execute, replatform produces meaningful run-rate reduction, retire eliminates ongoing cost entirely, refactor is reserved for the highest-run-rate workloads where the additional migration cost pays back.
A. For a typical 100-server mid-market estate, staying on-premise costs roughly $4.5M to $7M over 3 years (including hardware refresh, software licensing, data centre rent, operations staff). Migrating to cloud costs $4.5M to $8M over the same period including the one-time migration cost. The cloud case typically pays back versus on-premise within 18 to 36 months, with the saving accelerating in years 2 and 3 as the migration cost is amortised.
A. Parallel running is the most common cost overrun, typically by 30 to 80 percent of the budgeted figure. Teams underestimate how long the parallel-running window lasts (typically 3 to 6 months, not the 2 months in the original plan) and underestimate the per-month cost. The second most common overrun is contingency on Oracle, SQL Server, or specialised hardware workloads that turn out to be more complex than initial discovery indicated.
Strategy cost tables ->
7Rs framework, per-workload
Lift-and-shift per server ->
$3K to $8K band
Refactor vs replatform ->
3-year TCO difference
On-prem to AWS ->
MAP, EC2, Snowball
On-prem to Azure ->
Hybrid Benefit, FastTrack
On-prem to GCP ->
Sustained-use, BigQuery
Data centre exit ->
Larger programme reality
10 hidden costs ->
Detailed playbook
Updated 2 May 2026