Lift-and-shift cost per server is the most quoted figure in cloud migration cost references. The $3,000 to $8,000 per workload band has held for the past five years across the major cloud migration cost surveys (AWS, McKinsey, Forrester, Gartner) and is the standard starting point for migration cost estimates. This page decomposes the band: what each dollar buys, what drives variance, how programme size shifts the number, and how to read the headline figure honestly when scoping a migration.
A representative cost build for a single Windows or Linux server lift-and-shift, mid-market scope, partner-led, no special compliance overhead, standard application validation.
Per-server lift-and-shift cost decomposition, mid-market
| Line item | Low cost | Typical cost | High cost | Drives variance |
|---|---|---|---|---|
| Discovery and assessment | $200 | $400 | $800 | Dependency complexity |
| Agent installation and configuration | $150 | $250 | $500 | OS variant, network constraints |
| Replication setup and monitoring | $100 | $200 | $400 | Source IO, replication tool |
| Network and security configuration | $300 | $500 | $1,200 | VPC complexity, security group rework |
| Identity and access mapping | $150 | $300 | $700 | AD integration, service accounts |
| Application dependency validation | $400 | $800 | $1,800 | App complexity, integration count |
| Cutover preparation and rehearsal | $300 | $600 | $1,200 | Downtime tolerance, validation depth |
| Cutover execution and hypercare | $400 | $800 | $1,500 | Cutover complexity, on-call coverage |
| Source decommission preparation | $100 | $200 | $400 | Asset disposal requirements |
| Programme overhead (per-server allocation) | $300 | $600 | $1,200 | Programme size, governance |
| Contingency at 15 percent | $370 | $735 | $1,470 | Spread proportionally |
| Per-server total | $2,770 | $5,385 | $11,170 | Excluding parallel running infra |
The per-server number above excludes parallel running infrastructure cost (which is paid at the workload level rather than per-server migration cost) and excludes the destination cloud compute cost (which is the actual destination cost, not the migration cost). The true total per-server cost including parallel running typically adds another $1,500 to $3,000 per server for a 3-month parallel run, bringing the all-in per-server number to $4,500 to $8,500 for typical mid-market scope.
The $8,000 per server number is reached when several variance drivers compound. The most common patterns:
The variance rule of thumb
Programme overhead has both fixed and variable components. Fixed overhead (PMO setup, governance structure, tooling licensing) spreads better over larger programmes. Variable overhead (wave coordination, cutover management, post-cutover validation) scales roughly linearly with workload count. The net effect: per-server cost typically decreases as programme size increases up to a point, then plateaus or slightly increases as wave-management overhead becomes a significant line item.
Per-server cost by programme size, typical mid-market lift-and-shift
| Programme size | Per-server cost band | Notes |
|---|---|---|
| 10 servers | $6,000 to $10,000 | Fixed overhead dominates; small programme penalty |
| 25 servers | $5,000 to $8,500 | Standard small-business migration |
| 50 servers | $4,500 to $7,500 | Sweet spot for mid-market |
| 100 servers | $4,000 to $7,000 | Strong economies of scale |
| 250 servers | $3,800 to $6,500 | Wave structure begins to matter |
| 500 servers | $4,000 to $7,000 | Wave-management overhead returns |
| 1,000+ servers | $4,500 to $8,000 | Wave overhead plus long-tail complexity |
The headline per-server numbers throughout this page assume partner-led migration with a mid-market SI partner blended onshore-offshore rate. Self-led migrations have a different cost shape: the cash cost per server is lower because there is no partner margin, but the internal engineering time has its own cost that is often hidden in steady-state team budgets.
Self-led migrations typically run at 50 to 70 percent of the cash cost of partner-led for equivalent scope, but consume 1.5 to 2.5 times the internal engineering hours because of the learning curve and lack of cross-customer migration experience. For a 100-server migration, self-led typically costs $250K to $400K in cash and $400K to $800K in fully loaded internal engineering time. Partner-led typically costs $400K to $700K in cash and $150K to $300K in internal engineering time. The total cost is broadly comparable; the choice often comes down to internal team capacity rather than cash cost optimisation.
The $3K to $8K per server figure is a useful starting point but easily misused. The most common misuse patterns:
A reliable rule of thumb for scoping. Take the workload count, multiply by the per-server number adjusted for complexity (typical band $5,000 to $7,000 for mixed mid-market estates). Add 30 to 50 percent for parallel running and destination compute during the migration window. Add 15 percent contingency. The resulting number is a reasonable estimate range for the migration cost line of the cost case.
Per-server rule-of-thumb scope estimator
| Workload count | Per-server cost (typical) | Subtotal (migration) | Plus 40% parallel run | Plus 15% contingency | Total estimate |
|---|---|---|---|---|---|
| 10 | $7,500 | $75,000 | $105,000 | $120,750 | $120,750 |
| 25 | $6,500 | $162,500 | $227,500 | $261,625 | $261,625 |
| 50 | $6,000 | $300,000 | $420,000 | $483,000 | $483,000 |
| 100 | $5,500 | $550,000 | $770,000 | $885,500 | $885,500 |
| 250 | $5,200 | $1,300,000 | $1,820,000 | $2,093,000 | $2,093,000 |
| 500 | $5,500 | $2,750,000 | $3,850,000 | $4,427,500 | $4,427,500 |
| 1,000 | $6,000 | $6,000,000 | $8,400,000 | $9,660,000 | $9,660,000 |
The $3K to $8K per server figure is reliable as a starting point but always carries variance. The teams that produce on-budget lift-and-shift outcomes are the ones that decompose the per-server number by line item, identify variance drivers at assessment rather than at cutover, and budget honestly for parallel running and destination compute alongside the migration cost itself. The number itself is rarely the problem; the discipline of using it accurately usually is.
A. The range reflects variation across discovery complexity, agent compatibility, application dependencies, and validation depth. Simple Linux VMs with no application dependencies and standard cutover land at the low end. Windows servers with custom AD integration, multiple application dependencies, and heavy validation requirements land at the high end. Most mid-market lift-and-shift programmes blend out at $5,000 to $6,000 per server.
A. Three factors push lift-and-shift cost above the standard band. First, legacy operating systems (Windows Server 2008, Linux distributions out of support) requiring OS upgrade alongside the migration. Second, hardware-dependent workloads (specific HBA cards, USB licence dongles, serial connections) requiring re-architecture rather than simple rehost. Third, regulated workloads (HIPAA, PCI-DSS, FedRAMP) requiring additional documentation and compliance validation. Any of these can double the per-server cost.
A. Lift-and-shift is the cheapest migration cost per workload at $3K to $8K, compared with $8K to $25K for replatform and $25K to $80K+ for refactor. The trade-off is run-rate: lift-and-shift workloads typically run 5 to 15 percent more expensive than equivalent on-premise on a 3-year horizon because they are not optimised for cloud pricing. The break-even versus replatform on run-rate is typically 18 to 36 months; above that, the higher replatform migration cost pays back in lower run-rate.
A. Most workloads can technically lift-and-shift, but not all should. Workloads with active development that could benefit from cloud-native services (managed databases, serverless, container orchestration) often save more by replatform than they cost in additional migration effort. Workloads scheduled for retirement within 12 months should not be migrated at all. The most successful migration programmes use lift-and-shift as the default for stable workloads while strategically replatforming or retiring the rest.
A. For a mid-market lift-and-shift programme at $5,500 per server: discovery and assessment $400, agent setup and replication $300, parallel running infra $1,800, cutover labour and testing $1,500, validation $400, decommission $300, contingency at 15 percent $800. The cost concentration is in parallel running and cutover labour; tooling cost is small thanks to free AWS MGN or Azure ASR free windows.
A. Not exactly. Programme overhead (PMO, wave governance, cutover coordination) has fixed-cost elements that spread better over larger programmes. A 10-server migration typically costs $6K to $10K per server because the fixed overhead dominates. A 100-server migration typically blends to $4K to $7K per server. A 1,000-server migration typically blends to $3K to $5K per server but adds wave-management overhead that brings the all-in number back up.
Strategy cost tables ->
7Rs framework, per-workload
Refactor vs replatform ->
3-year TCO difference
100-server worked scenario ->
Full mid-market reference
On-prem to AWS ->
MAP, EC2, Snowball
On-prem to Azure ->
Hybrid Benefit, FastTrack
AWS MGN cost ->
Free-for-90-days window
Azure Migrate cost ->
31-day free ASR trial
10 hidden costs ->
Detailed playbook
Updated 2 May 2026