Cloud Migration Cost in 2026: Budgets by Size, the 6 Rs and a Calculator

The short answer

These are 2026 labor ranges for US companies, blended across vendor types and providers. The low end of each band is a mostly rehosted estate migrated by a team in Latin America, Central Europe or Asia. The high end is a US consultancy, or any team dealing with legacy databases, undocumented dependencies and regulated data. In hours, the three tiers are roughly 300–1,200, 1,500–7,000 and 8,000 and up.

The numbers come from migration estimates we write, proposals we compete against and the rate data in our guide to nearshore software development rates. Disclosure: Gilzor is a software company with engineering teams in Poland and Cyprus. For US clients that makes us offshore, and our pricing sits in the Central and Eastern Europe row below. If you have already chosen AWS, our AWS cloud migration cost guide goes into AWS pricing tools, migration credits and the Database Migration Service. This article stays provider-neutral.

Why cloud migrations run over budget

Overruns are the norm, not the exception. In McKinsey's survey of about 450 CIOs and IT decision makers on cloud migration, 75% of companies said their migration ran over budget and 38% said it ran behind schedule. McKinsey estimated that this kind of overspend would add up to about $100 billion of wasted migration spend over three years. The money keeps leaking after go-live too: Flexera's 2025 State of the Cloud report found respondents estimate 27% of their IaaS and PaaS spend is wasted, cloud budgets were already exceeded by 17%, and 84% named managing cloud spend as their top cloud challenge.

What we see in estimates and first calls is consistent. The overruns rarely come from the provider's price list. They come from four places:

1. An inventory that isn't oneA spreadsheet of servers is not a dependency map. The batch job that writes to a file share, the hard-coded IP in a partner integration, the license dongle on one VM: each one found in week six instead of week one costs days of rework and pushes the cutover.
2. Dual-running that lasts longer than plannedEvery month the old data center and the new cloud account both run, you pay twice. Plans assume four weeks of overlap. Reality is often three to six months when testing slips or one stubborn application can't move.
3. Strategy decided too lateTeams start lifting and shifting, discover a database that should go to a managed service, and replatform mid-flight. Deciding the strategy per application up front is cheaper than changing it during the move.
4. Nobody owns the run costThe migration team hits its deadline. The bill arrives a month later, 20–40% higher than the old hosting, because everything was sized for peak and left on demand.

The 6 Rs and what each one costs

The 6 Rs started as Gartner's five migration strategies and were extended by AWS (which now lists seven, adding "relocate" for moving VMware estates as they are). They're provider-neutral and they matter for budgeting because every application gets one, and the mix sets the price.

The cheapest line in that table is "retire." In assessments it's common to find that 10–20% of what's running is unused, duplicated or kept "just in case." Every app you retire is one you don't migrate, test or pay for in two places.

Cost per application, by strategy

Here's the same picture as bar lengths: typical labor for one mid-complexity business application (a web app, its database and a few integrations) under each strategy.

Databases deserve their own line. Moving a SQL Server or Oracle database with near-zero downtime means schema assessment, replication setup, validation runs and a rehearsed cutover. Budget 40–120 hours per production database for a straightforward one, and several times that for a large Oracle estate with stored procedures you plan to convert to PostgreSQL.

Migration cost vs run cost

A cloud migration business case lives or dies on the monthly bill afterward, not on the project quote. Gartner forecast worldwide public cloud end-user spending at $723.4 billion for 2025, and a large share of that is companies paying for capacity they sized during migration and never revisited.

The pattern looks the same on almost every estate we review. Before migration there's a flat on-premises cost. During migration both environments run. Right after cutover the cloud bill is often higher than the old hosting, because servers were copied at their on-premises size and left on on-demand pricing. Then optimization brings it down, if somebody does it.

The orange block in that chart is a real budget line. For an estate costing $20,000 a month, three months of dual-running is $60,000 or more, which can match the entire labor cost of a small migration. Short, well-rehearsed cutovers are the single biggest lever on the total.

Built by Gilzor

Results we’ve shipped

70+products launched

98%delivered on time

85%clients come back

Aviation33%faster baggage weigh-inWeb and Mobile application for Airlines→Healthcare+30%lab test ordersMedical Lab Web and Native Mobile Applications→Retail · Startup700+stores aggregatedMobile application for Grocery shops→

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 →

Rates by region for the same migration

Hours depend on your estate. The rate depends on who does the work. Here's a mid-size migration of about 2,000 hours (15 applications, four databases, a mixed rehost and replatform plan) at 2026 vendor rates for cloud and DevOps engineers:

Time zones matter more in migrations than in most projects, but in an unexpected direction. Cutovers happen at night and on weekends, US time. A team in Warsaw is in its morning when a US East Coast business closes on Friday evening, so a Saturday-morning cutover in Europe lands in a quiet window for US users. What still needs shared hours is the assessment, the dependency workshops and acceptance testing with your application owners.

Cloud migration cost calculator

Enter your estate, pick the dominant strategy and the vendor region. The calculator gives migration hours, labor cost, dual-running and data transfer costs, a one-time total with contingency, and a rough monthly cloud bill after the first round of optimization. If one or two applications will be refactored, run them separately with the "refactor-heavy" option and add the results.

The default (15 applications, four databases, a mixed plan, built in Central Europe) lands around 1,700 hours and $110,000 in labor. Three months of dual-running at $20,000 a month adds $60,000, more than half the labor cost again. Switch the vendor to a US consultancy and labor multiplies by about 2.6. Switch to "mostly rehost" and the project gets cheaper, but the monthly bill estimate goes up: that's the trade-off the 6 Rs are about.

The hidden costs of cloud migration

These are the lines that show up between the old bill and the new one, roughly in the order buyers are surprised by them:

Egress deserves a note because it works in both directions. Moving data into AWS, Azure or Google Cloud is free. Since early 2024, all three also waive egress fees for customers who leave entirely: Google Cloud announced it in January 2024, AWS and Microsoft followed in March, with conditions such as moving the data within 60 days and, for Azure, closing the subscriptions. Everyday egress (serving files to users, replicating to a backup elsewhere, calling a partner's API) is still billed and grows with your traffic.

Which migration path fits your estate?

Six questions about your applications, deadline and team. The result points to the strategy mix that usually gives the best ratio between the one-time cost and the monthly bill.

Three budgets, worked through

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

1. A small SaaS leaving a hosting provider

A B2B SaaS runs on six VMs at a managed hosting company: two app servers, a PostgreSQL database, a worker, a file server and a staging box. The plan is to rehost the app servers, move the database to a managed service and the files to object storage. About 350–450 hours including a landing zone, infrastructure as code and a rehearsed cutover: $23,000–29,000 with a Central European or Latin American team, $60,000–75,000 with a US consultancy. One month of overlap at $6,000. The cloud bill after rightsizing lands close to the old hosting cost, with backups and failover the old setup never had.

2. A mid-size company closing a server room

Twenty applications, mostly Windows, three SQL Server databases and a file share, with a colocation contract ending in nine months. Assessment retires four apps and repurchases one as SaaS. The rest are rehosted, with the databases replatformed. Around 1,800–2,300 hours: $115,000–150,000 offshore or nearshore, $300,000–390,000 onshore. Licensing is the line to watch: Windows Server and SQL Server costs can rival compute unless existing licenses carry over.

3. A healthcare platform moving clouds

A patient-facing platform moves from one cloud provider to another to consolidate with its parent company, with 80 TB of data and HIPAA obligations. The exit waiver covers the bulk data transfer if the move completes within the provider's window. The cost sits in re-creating the security controls, audit logging and network design on the new provider, plus a parallel run for validation. About 3,500–5,000 hours: $230,000–330,000 offshore or nearshore, $600,000–850,000 with a US firm.

How to reduce cloud migration cost without breaking things

1. Pay for a real assessment firstDependency mapping, a strategy per app and a run-cost model typically take 10–15% of the budget and protect the other 85%. Our business analysis team runs this as a fixed first phase.
2. Retire before you migrateEvery unused app, stale environment and forgotten VM you switch off is one less thing to move, test and pay for twice.
3. Migrate in waves and decommission as you goShort dual-running per wave beats one long overlap at the end. Cancel hardware and colocation contracts on the wave schedule.
4. Rightsize from usage data, not from today's server specsOn-premises servers are sized for peak plus headroom. Use a few weeks of CPU and memory metrics to pick instance sizes.
5. Buy commitments only after things settleReserved instances and savings plans cut compute cost substantially, but lock in the wrong size if bought on day one. Wait for four to eight weeks of real usage.
6. Negotiate credits and licensingAll three providers run migration incentive programs and license benefits for qualifying customers. Ask before you sign, not after.
7. Don't cut testing and rehearsalsA rehearsed cutover is what keeps dual-running short. Our QA team tests migrated apps against production-like data and failure scenarios, not just "it starts."

FAQ

How we approach cloud migrations at Gilzor

We migrate the products we build and maintain, and we take on migrations for teams whose apps already exist. Our DevOps work centers on AWS, Docker, Kubernetes, Ansible and CI/CD pipelines in GitHub Actions or CircleCI; the full list is on our tech stack page. We price the assessment, each migration wave and the post-migration optimization as separate lines, and we give you a run-cost estimate next to the project quote so the business case covers both. If you'd rather compare several vendors first, our list of cloud deployment companies in the USA is a reasonable starting point. If the migration is part of a wider modernization program, our guide to digital transformation cost covers the bigger budget.