AWS Cloud Migration Cost in 2026: Real Budgets, MAP Credits and the Bill After

The short answer

Those are 2026 project budgets for US companies: assessment, landing zone, migration work, testing and cutover, blended across vendor types. They do not include what you pay AWS during the migration or after it. The low end of each range is a team in Latin America, Central Europe or Asia moving clean, well-documented systems. The high end is a US consultancy, or any team facing old operating systems, Oracle databases with years of stored procedures and an audit trail to preserve.

This guide is about AWS specifically: its pricing tools, its incentive programs and the parts of its bill that surprise people in month three. If you are still choosing between AWS, Azure and Google Cloud, or want the general picture of migration strategies, start with our guide to cloud migration cost. Disclosure: Gilzor is a software company with engineering teams in Poland and Cyprus. For US clients that makes us offshore, and our rates sit in the Central and Eastern Europe row of the table further down.

The three bills of an AWS migration

When a CFO asks what the move costs, they mean one number. There are three:

1. The project billPeople: discovery, the landing zone (accounts, networking, IAM, logging), moving servers and databases, changing application code where needed, testing and cutover. This is 60–80% of the one-off cost in almost every estimate we see, and the part you can control best through strategy and vendor choice.
2. The overlap billWhile workloads replicate and get tested, you pay AWS for the new environment and keep paying the data center, colocation contract, hardware support and licenses for the old one. Two to six months of overlap is normal. Long migrations make this bill bigger than the tooling and transfer costs combined.
3. The run billYour monthly AWS invoice after cutover: compute, storage, databases, data transfer, support plan and the managed services you add. It's the only one of the three that never ends, and it decides whether the business case holds up in year two.

Overruns are common enough to plan for. McKinsey's survey of roughly 450 CIOs and IT leaders found companies spent on average 14% more on migration than planned each year, 38% had seen migrations delayed by more than a quarter, and the firm estimated well over $100 billion of wasted migration spend globally over three years. After cutover the waste moves to the run bill: Flexera's 2025 State of the Cloud report put self-estimated wasted cloud spend at 27%, and 84% of respondents named managing cloud spend as their top cloud challenge.

What AWS gives you for free (and what it doesn't)

AWS wants your workloads, so most of its migration tooling costs little or nothing. The catch: free tools don't do the engineering.

What we see in estimates and first calls: teams read "DMS is free" and budget the database move at near zero. Moving the data is the easy 20%. Converting a SQL Server or Oracle database with a few hundred stored procedures to Aurora PostgreSQL, then fixing and retesting every application query that depended on engine-specific behavior, is where weeks go. If the engine stays the same (MySQL to RDS for MySQL, PostgreSQL to Aurora PostgreSQL), database work drops to a fraction.

MAP credits: how AWS pays for part of your migration

The Migration Acceleration Program is the main reason the AWS migration price you pay can be lower than the one you were quoted. It runs in three phases, and funding can attach to each:

1. AssessAn inventory and business case, often built with Migration Evaluator. Funding at this stage typically covers part or all of the assessment work.
2. MobilizeLanding zone, security baseline, skills and a pilot migration. Funding can cover part of the partner effort and training.
3. Migrate and ModernizeThe waves themselves. This is where AWS service credits come in, offsetting spend on the migrated workloads while you still pay for the old environment.

AWS doesn't publish a fixed formula, and terms are negotiated per migration. Partner write-ups describe Migrate-phase credits scaled to the annualized AWS spend of the workloads you move, commonly in the range of 15–25% for larger commitments, and eligibility that starts at a meaningful committed spend (small estates may not qualify at all). MAP engagements run through AWS or an AWS Partner with migration specialization, and the partner of record matters for the paperwork. Whoever does the engineering, ask the question early: "Is this migration MAP-eligible, and who files it?"

Built by Gilzor

Results we’ve shipped

70+products launched

98%delivered on time

85%clients come back

Real estateModernizing a Multi-Service Real Estate Marketplace→FinTechAndroid application for FinTech industry→Pharmacy · DACHRedesigning and modernizing an Online Pharmacy mobile app→

Talk to the people who build it. Tell us about your project and get a free estimate of scope, timeline and cost.

See how we’d approach yours →

The overlap curve: where dual-running money goes

Here is a typical monthly cost profile for a mid-size migration: a $30,000-a-month data center footprint moving to AWS over eight months. Note the hump in the middle and the step down after commitments.

Two things move the overlap bill most. The first is your exit date: if the colocation contract renews in March for another year, a migration that finishes in April costs you a year of rent. Work backward from contract, hardware support and license renewal dates before you set the wave plan. The second is how long replicated servers sit in AWS before cutover. Each test environment left running "for safety" for an extra month is a month of EC2 and EBS at On-Demand rates.

Savings Plans, Reserved Instances and the run bill

Most of the run bill is decided by pricing models, not engineering. AWS publishes these maximum discounts versus On-Demand:

The trap is timing. Commit too early and you lock in the oversized instances you copied from on-premises, where servers were bought for peak load plus three years of growth. Commit too late and you pay On-Demand rates for months you didn't need to. The sequence that works:

1. Run On-Demand for one to three monthsLong enough to see a full business cycle (month-end batch jobs, seasonal traffic).
2. RightsizeUse Compute Optimizer and CloudWatch data. Lift-and-shift estates commonly have instances running at single-digit CPU averages; dropping one or two sizes halves their cost before any commitment.
3. Commit to the baseline, not the peakCover the steady part with a one-year Compute Savings Plan, keep the variable part On-Demand or Spot, and add commitments as the picture settles.
4. Turn off what doesn't need to runDev and test environments scheduled to office hours run about a third of the week.

Licensing deserves its own line. Windows Server and SQL Server on AWS are often cheapest license-included, because Microsoft's 2019 licensing changes restrict bringing your own Windows Server licenses to AWS (licenses bought before October 2019 can still run on dedicated hosts), while SQL Server can move under License Mobility only with active Software Assurance. SQL Server Enterprise license-included rates can make a database instance several times more expensive than the same instance running PostgreSQL. That single line is often the strongest financial argument for an engine change, and for paying the engineering cost that comes with it.

Data transfer and egress: free in, metered out

Moving data into AWS costs nothing. Getting it out, or moving it around inside AWS, is metered, and architectures copied from a data center generate more of that traffic than anyone expects.

For the move itself: up to a few terabytes, replication over your internet link or a site-to-site VPN is fine. For tens of terabytes, a Snowball device or a Direct Connect connection is often faster and more predictable. The hybrid phase is the expensive one. If an application moves to AWS in wave two but its database stays on-premises until wave four, every query crosses the wire in both latency and dollars. Group applications with their data when planning waves.

Partner and labor rates by region

The hours for a migration depend on your estate and strategy. The rate depends on who does it. Senior cloud and DevOps engineering rates for US clients in 2026:

Time zones cut both ways. A Central European team can run a cutover that starts in its morning and finishes before US users log in. What doesn't work is a migration where every decision waits for a stakeholder who is only free after 2 p.m. Pacific. For the full offshore math, see cost of offshore software development.

AWS migration cost calculator

Enter your server and database count, how much change you want along the way, and the AWS bill you expect after migration (the Pricing Calculator or Migration Evaluator gives you a first figure). The calculator returns project hours and cost, the dual-running bill, an indicative MAP credit and the run bill after commitments.

The defaults (40 servers, four same-engine databases, a rehost and replatform mix, Central European team) land around 1,240 hours and $87,000 in project work, plus about $37,000 of AWS spend during the overlap. Switch the databases to an Oracle conversion and the project grows by roughly 700 hours. Switch to a US consultancy and the people cost multiplies by about 2.7. Watch the last line: a 1-year Savings Plan after rightsizing takes about 15% off the whole run bill, every month, which over three years is usually worth more than the MAP credits.

Where will your money leak?

Every AWS migration has one cost lever that matters more than the rest. Six questions to find yours.

The hidden costs of an AWS migration

These are the lines that turn up after the statement of work is signed, roughly in the order they surprise people:

Three budgets, worked through

Typical requests from US companies, priced with the ranges above. Illustrative, not Gilzor quotes.

1. A SaaS startup leaving a VPS provider

Twelve Linux servers and one PostgreSQL database move to AWS, the app goes into containers on ECS and the database into Aurora PostgreSQL. About 350–450 hours including the landing zone, CI/CD changes and load testing: $25,000–35,000 with a Central European or Latin American team, $65,000–85,000 with a US consultancy. Expected AWS bill around $4,000 a month, too small for meaningful MAP credits. The biggest saving comes from rightsizing and a one-year Compute Savings Plan after month two.

2. A mid-size company leaving colocation

Ninety servers, a third of them Windows, and eight SQL Server databases, two of which move to Aurora PostgreSQL to escape Enterprise licensing. About 2,800–3,500 hours over seven months: roughly $200,000–270,000 offshore or nearshore, $500,000–650,000 onshore. The AWS bill ramps to about $35,000 a month, so dual-running adds around $100,000–150,000 depending on how fast the waves go. At that spend MAP is worth applying for and may offset a meaningful part of the overlap. The colocation contract renews in month eight, which sets the deadline for every wave.

3. A healthcare platform with HIPAA obligations

Forty servers and an Oracle database with heavy PL/SQL move to AWS under a Business Associate Addendum, with encryption, audit logging and access reviews built into the landing zone. The Oracle-to-Aurora conversion alone takes 600–900 hours, including retesting every report. Total around 2,500–3,000 hours: $175,000–255,000 offshore or nearshore, $475,000–570,000 onshore. The payback comes mainly from the Oracle license, not from the servers.

How to reduce AWS migration cost without breaking things

1. Retire before you migrateMost inventories include applications nobody uses. Every server you switch off instead of moving saves migration hours and a line on the run bill forever. A short business analysis phase pays for itself here.
2. Rehost first, modernize what earns itLift and shift is the cheapest way to exit a data center on time. Refactor the three or four systems where licensing, scaling or release speed justify the cost, after you're out.
3. Change database engines where licenses justify itAn Oracle or SQL Server Enterprise conversion is expensive once and cheap forever. A MySQL instance doesn't need one.
4. Move applications with their dataSplitting them across waves creates latency problems and data transfer charges during the hybrid phase.
5. Don't buy commitments on day oneRightsize first. A three-year plan sized to on-premises servers locks in waste.
6. Tag everything and set budgets before cutoverCost allocation tags, AWS Budgets alerts and Cost Explorer reports per team turn a surprise invoice into a Tuesday alert.
7. Test like it mattersPerformance, failover and data integrity tests before each cutover cost far less than a rollback. Our QA team runs regression and load tests against migrated environments, not just smoke tests.

Migrations are often the moment companies decide to rebuild parts of a product rather than move them as they are. If that's where you're heading, our guides to SaaS development cost and digital transformation cost cover the budgets beyond infrastructure.

FAQ

Where Gilzor fits in an AWS migration

Our part of a migration is the application side: the code changes behind a replatform or refactor, database conversions and the application queries that depend on them, CI/CD for the new environment, and testing before every cutover. We work as a dedicated team or extend yours through development support, and when an inherited system misbehaves after a move, our tech troubleshooting engineers find the root cause. Our teams sit in Poland and Cyprus, with 70+ projects launched and 85% of clients coming back for more work. In an estimate from us, the lift and shift, the engineering work and the testing are separate lines, so you can see which systems drive the cost and compare them with any other quote.