AWS to Google Cloud Migration in India, 2026 Guide
Saurav Jagdale
Technical Lead, Unico Connect
In this article
- Quick Answer
- Key Takeaways
- Why do Indian companies move from AWS to Google Cloud?
- When should you not migrate from AWS to Google Cloud?
- How do AWS services map to Google Cloud?
- What changes beyond the service names?
- What does an AWS to Google Cloud migration cost?
- What changes for Indian companies leaving AWS?
- How long does an AWS to Google Cloud migration take?
- What does an AWS to Google Cloud migration plan look like?
- What are the most common AWS to Google Cloud migration mistakes?
- Frequently Asked Questions
- How can Unico Connect help you move off AWS?
Teams rarely leave AWS because Google Cloud is better at everything. Usually they want BigQuery and Gemini next to their data. Sometimes Google funding tips the business case for a large move, or the company already runs on Google Workspace and wants one vendor. The service mapping gets most of the attention and is the easy part. Moving the data without a large egress bill is harder. Hardest of all is getting engineers who know AWS inside out to work with a different identity and network model. This guide walks through an AWS to Google Cloud (GCP) migration as it runs for Indian teams in 2026, with the costs, the billing changes and an illustrative plan.
Quick Answer
An AWS to Google Cloud (GCP) migration maps every AWS service to a Google Cloud equivalent, builds the landing zone first, then moves workloads in waves while data replicates, so each cut over is short. Migrate to Virtual Machines moves EC2, Storage Transfer Service moves S3, Database Migration Service moves RDS and Aurora, and the BigQuery Data Transfer Service moves Redshift. DynamoDB, Step Functions and Cognito change shape on the way, so plan them as rewrites.
For Indian teams, the regions do not line up. Google Cloud has no Hyderabad region, so a Mumbai and Hyderabad recovery pair on AWS becomes Mumbai, asia-south1, and Delhi, asia-south2. Ask AWS Support for free data transfer out before any part of the move starts. AWS credits it for customers leaving AWS, but only after it approves the request. Any fixed saving quoted before someone has read your architecture is a guess.
Key Takeaways
- Most AWS compute, storage and relational database services have a Google Cloud equivalent and a Google tool that moves them. DynamoDB, Step Functions, Cognito and deep IAM policies cannot be lifted across, so budget them as rewrites from the start.
- Identity is the biggest shift. Design the Google Cloud organization, folders and projects before anything moves, because every later permission decision hangs off it.
- Budget for a few weeks of running both clouds. That overlap is what makes a rollback possible.
- With no Google Cloud region in Hyderabad, workloads there usually move to Mumbai, and Delhi becomes the in country recovery site.
Why do Indian companies move from AWS to Google Cloud?
Data and AI usually lead. BigQuery is a serverless warehouse with no cluster to size, and the Gemini Enterprise Agent Platform, formerly Vertex AI, works on the same data without copying it anywhere. Containers come next, because GKE Autopilot and Cloud Run take a lot of cluster work off a small team. Companies that already run on Google Workspace often want one vendor for everything. And for larger estates, Google funds part of the move through its Rapid Migration and Modernization Program, which can change the numbers on its own. The wider platform decision is in our guide on when to choose GCP over AWS or Azure.
We would not move for cost alone. For steady workloads, AWS Savings Plans and Reserved Instances close much of any price gap, so if a smaller bill is the only goal, optimise AWS first. If you want a data platform, an AI roadmap or a Google centred stack, the move can earn its cost. Still choosing a cloud? Start with our AWS vs Azure vs GCP comparison.
When should you not migrate from AWS to Google Cloud?
Wait if you have just signed a multi year AWS commitment, because you pay for it whether you use it or not. Think hard if the application is built deep into AWS specific services such as DynamoDB single table designs, Step Functions orchestration or Cognito, because then the move is a rebuild and should be budgeted as one. A team with no Google Cloud experience and no budget to learn it or hire it in will struggle. And never schedule the cut over in your peak season. Where the doubt is about skills or rewrites, we would move one low risk workload first and decide after that.
How do AWS services map to Google Cloud?
The mapping is where the migration plan starts, and Google publishes AWS migration guides for the main services. Most compute, storage and relational database services have a direct equivalent and a Google tool that moves them. The rows marked as rewrites are where the budget goes.
AWS services and their Google Cloud equivalents in 2026
| AWS service | Google Cloud equivalent | What changes in the move |
|---|---|---|
| EC2 | Compute Engine | Migrate to Virtual Machines moves EC2 instances directly. Choose the machine series before pricing, because prices and discounts differ by series |
| EBS and EFS | Persistent Disk or Hyperdisk, and Filestore | Moves with the VM or by file copy. Recheck performance tiers rather than matching sizes |
| S3 | Cloud Storage | Storage Transfer Service reads straight from S3, including over a Google managed private network from ap-south-1 and ap-south-2 that avoids AWS S3 internet egress. Pick storage classes again, the tiers do not map one to one |
| RDS and Aurora PostgreSQL | Cloud SQL or AlloyDB | Database Migration Service replicates continuously so cut over is short |
| RDS and Aurora MySQL | Cloud SQL for MySQL | Same continuous replication path |
| RDS for SQL Server | Cloud SQL for SQL Server | Database Migration Service lists Amazon RDS as a supported SQL Server source |
| DynamoDB | Firestore, Spanner or Bigtable | Google publishes migration guides from DynamoDB to Spanner and to Bigtable. The data model and access code still change, so plan this as a rewrite |
| Redshift | BigQuery | BigQuery Data Transfer Service copies schema and data, staging through an S3 bucket that adds AWS side costs. The BigQuery batch SQL translator converts Redshift SQL. The cost model moves from cluster to scanned bytes or slots |
| Lambda | Cloud Run functions or Cloud Run | Handler code ports. Triggers, IAM and packaging change |
| ECS and Fargate | Cloud Run | Containers move as they are. Cloud Run scales to zero |
| EKS | Google Kubernetes Engine | Manifests mostly port. Replace AWS specific controllers, load balancer annotations and IAM for pods |
| SQS, SNS and Kinesis | Pub/Sub, Cloud Tasks and Dataflow | Delivery semantics differ. Test ordering and retries |
| Step Functions and EventBridge | Workflows and Eventarc | Definitions are rewritten, not converted |
| ElastiCache | Memorystore | Warm the cache after cut over |
| CloudFront and Route 53 | Cloud CDN and Cloud DNS | Lower DNS time to live days before the switch |
| CloudWatch | Cloud Monitoring and Cloud Logging | Dashboards and alerts are rebuilt |
| IAM and Organizations | Cloud IAM with organization, folders and projects | The biggest conceptual change. Design the hierarchy before moving anything |
| Cognito | Identity Platform or Firebase Authentication | Users and sign in flows move to a new identity service, so plan it as a rebuild with its own cut over |
| CloudFormation | Terraform, which Infrastructure Manager can run as a managed service | Google offers no converter, so templates are rewritten |
| SageMaker and Bedrock | Gemini Enterprise Agent Platform, formerly Vertex AI | Models, endpoints and pipelines are redeployed. Anthropic Claude models are also offered in Model Garden, so teams using Claude on Bedrock can keep it, and prompts need tuning only if you switch to Gemini |
What changes beyond the service names?
Networking is built differently. A Google Cloud VPC is global with regional subnets, while an AWS VPC is regional. Multi region designs get simpler, but the CIDR plan and firewall rules have to be redrawn. Google Cloud load balancing can also serve a global application from a single anycast IP.
Identity is the bigger shift. AWS organises around accounts and attaches policies to identities and resources. Google Cloud organises around an organization, folders and projects, and grants roles to principals at any level of that hierarchy. Workloads use service accounts where AWS uses instance profiles, and GKE pods use Workload Identity Federation for GKE, the counterpart of IAM roles for service accounts and EKS Pod Identity. Design this hierarchy before anything moves. Every later permission decision hangs off it.
Terraform stays, but every resource block changes provider. CloudFormation has no Google converter, so those templates are rewritten, usually in Terraform, which Google Cloud can run as a managed service through Infrastructure Manager. Skip Deployment Manager, which reached end of support on 31 March 2026.
Operations is what teams underestimate. Dashboards, alerts, runbooks and on call habits all get rebuilt in Cloud Monitoring and Cloud Logging, and nothing tells you they are wrong until the first incident on the new platform.
What does an AWS to Google Cloud migration cost?
There are two numbers to get right, the one time cost of moving and the change in what you pay each month.
One time costs
The move itself costs an assessment, the landing zone build, engineering time for each wave, any rewrites of AWS specific services, and a stretch of running both clouds while traffic shifts. That last line is the one most budgets leave out. Expect to pay for two production environments for a few weeks. It is what makes a rollback possible.
Data transfer out of AWS
Egress is usually the number that worries finance most, and for a full move most of it can be avoided. AWS gives free data transfer out to customers moving all of their data off AWS, or all of their data off one service. It is not automatic. You contact AWS Support and wait for approval before any workload or data leaves AWS. AWS reviews the plan at the account level and applies a temporary credit sized to the data you hold. Eligible customers then have 90 days to finish, and anyone who needs longer should tell Support straight away. By the end of the window the data and workloads left on the services you exited must be deleted, though AWS does not ask you to close the account. Under 100 GB, the standard monthly free allowance already covers it. Transfers through CloudFront, Direct Connect, Snow Family and Global Accelerator are not covered, so get in writing what your credit includes. Because the window is short, we would file the request once the landing zone is built and the waves have dates, so approval lands just before wave one.
For S3 there is a second route. Storage Transfer Service can pull from S3 in ap-south-1 Mumbai and ap-south-2 Hyderabad over a Google managed private network, where you pay no AWS S3 internet egress and instead pay Google a data processing fee, published at $0.08 per GiB for transfers within APAC in September 2026. AWS request charges such as LIST and GET still apply. Standard agentless transfers from S3 carry no Storage Transfer Service fee, so the private network route earns its cost mainly when the AWS credit does not apply, for example when only part of the data moves.
Running cost
Architecture decides this far more than the choice of cloud. Both clouds cut the price of steady and interruptible workloads heavily, so compare a committed AWS rate with a committed Google Cloud rate for the same workload and machine series. Price it in INR for the region you will use. Indian accounts also carry a separate India surcharge that Google applies to the billing account whatever region is used, so moving workloads to a US region does not remove it. BigQuery is the line that can swing either way. You pay for bytes scanned or reserved slots instead of a cluster, which suits spiky analytics and punishes badly written queries.
The levers behind the GCP vs AWS cost question sit side by side below, as each vendor publishes them in September 2026.
Google Cloud vs AWS cost levers as published in September 2026
| Cost lever | Google Cloud | AWS |
|---|---|---|
| Automatic discount | Sustained use discounts apply automatically on self serve Cloud Billing accounts once a VM of an eligible machine series runs for more than a quarter of the month, up to 30 percent on N1, M1 and M2 and up to 20 percent on N2, N2D and C2. E2, N4, C3 and C4 get none, and invoiced billing accounts do not receive them | None. Discounts need a Savings Plan or Reserved Instance |
| Commitment discount | One or three year committed use discounts reach up to 55 percent for most machine types and up to 70 percent for memory optimized ones. Flexible spend commitments 28 percent for one year and 46 percent for three on most series | Savings Plans up to 72 percent off on demand prices, plus Reserved Instances |
| Spare capacity | Spot VMs up to 91 percent off | Spot Instances up to 90 percent off |
| Managed Kubernetes fee | GKE $0.10 per cluster hour in every mode, free tier covers one Autopilot or zonal cluster. $0.60 on the Extended channel after standard support ends | EKS $0.10 per cluster hour in standard support, $0.60 in extended support |
| Data warehouse | BigQuery bills bytes scanned on demand, or reserved slot capacity | Redshift bills provisioned nodes, or Serverless compute in RPU hours |
| Moving S3 data out | Standard agentless transfers from S3 carry no Storage Transfer Service fee. The managed private network route costs $0.08 per GiB within APAC and removes AWS S3 internet egress | Free data transfer out on request through AWS Support when you move all data off AWS or off one service, 90 days to complete the move |
Two things in a proposal should make you pause. The first is a fixed saving percentage, because nobody can know it without your architecture. The second is a claim that the Kubernetes control plane is free on Google Cloud. It is not. GKE charges $0.10 per cluster hour in every mode, the same as EKS in standard support, and the free tier covers one Autopilot or zonal cluster per billing account. Both clouds charge $0.60 per cluster hour for versions kept past standard support, GKE through its Extended channel. So if your EKS clusters already pay the extended support fee, the saving comes from upgrading them during the move, not from the change of cloud.
Migration funding from Google
For a large estate, funding can do more for the business case than any discount. Google Cloud funds qualifying migrations through its Rapid Migration and Modernization Program, RaMP, once each workload has a projected annual run rate of at least $60,000 of eligible Google Cloud spend. Google decides who qualifies and sets the terms, so check that floor and raise RaMP before scoping starts.
What changes for Indian companies leaving AWS?
If your AWS account sits with AWS India, you already pay in INR with GST, and that does not change. AWS documents that some accounts, including member accounts in an organization, have a seller of record other than AWS India, and for those teams rupee billing is itself new. Add your GSTIN to the new Google Cloud billing account on day one so input tax credit carries on without a gap. AWS India supports automatic payments only up to 15,000 INR per bill, the same RBI e-mandate ceiling that applies on Google Cloud, so if you pay AWS by hand today, expect to do the same on Google Cloud until you qualify for invoiced billing.
The regions do not line up city for city. AWS runs ap-south-1 in Mumbai and ap-south-2 in Hyderabad. Google Cloud runs two regions in India, Mumbai as asia-south1 and Delhi as asia-south2. A Mumbai workload maps cleanly. A Hyderabad workload, or a recovery design built on the Mumbai and Hyderabad pair, has to be redesigned around Mumbai and Delhi. That adds latency for users in the south and puts the recovery site much further from the primary one. The Reserve Bank of India circular of 6 April 2018 requires payment system data to be stored only in India. If you run a payment system, or serve as a service provider to one, pin those workloads to the two Indian regions with organization policies before anything moves.
How long does an AWS to Google Cloud migration take?
Google puts a move from AWS at days, weeks or months depending on scope, and the size of your AWS bill says little about where yours will land. What sets the pace is how many AWS specific services need rewriting, how much data has to move and how fast it changes, and how many teams have to change the way they work. A handful of containerised services on one PostgreSQL database is a single short project. An estate with DynamoDB, Redshift and many accounts is a programme of several waves over months.
The order of work is what shortens it. Assess and map dependencies first in Google Cloud Migration Center, which imports AWS inventory directly and estimates Google Cloud cost. Then build the landing zone. Move stateless services early, because they are easy to roll back and they teach the team the new platform. Leave the databases and the final DNS switch until replication has been proven.
What does an AWS to Google Cloud migration plan look like?
The plan below is illustrative. It is not a client project. It takes a shape common among Indian SaaS products and shows where each tool in this guide fits.
Picture a B2B SaaS product in AWS Mumbai. Twenty containerised services on EKS, one PostgreSQL database on RDS of around 500 GB, about 3 TB of user uploads in S3, a small Redshift cluster for reporting, and SQS between services. The team uses Google Workspace and wants to put Gemini on its reporting data.
Wave zero, foundations. Build the Google Cloud organization with folders for shared, staging and production, a Shared VPC with subnets in asia-south1, Workload Identity Federation for GKE, and organization policies that keep data in India. Set up billing with the GSTIN and budgets on day one. With the foundations built and each wave scheduled, file the AWS data transfer out request, and start wave one only after AWS approves it.
Wave one, stateless services. For twenty services like these we would default to Cloud Run and keep a GKE Autopilot cluster only for the services that depend on Kubernetes features. SQS becomes Pub/Sub. The new services talk to the existing RDS database over an HA VPN link between the two clouds, which Google documents step by step, so they can take a small share of live traffic early.
Wave two, data. Start Database Migration Service replication from RDS PostgreSQL to Cloud SQL or AlloyDB, copy S3 to Cloud Storage with Storage Transfer Service and keep it in sync, and move Redshift to BigQuery with the BigQuery Data Transfer Service before rebuilding the reports.
Wave three, cut over. Lower DNS time to live in advance, freeze writes briefly, let replication catch up, promote the Google Cloud database, and switch traffic. Keep AWS running and ready for rollback until the agreed threshold passes. Once the Google Cloud database takes writes, a rollback means carrying those writes back to AWS, so agree how that would work before the switch.
Wave four, clean up and optimise. Decommission AWS inside the 90 day data transfer window. Then rightsize, commit where usage is steady using the levers in our Google Cloud cost optimization guide, and put Gemini to work on the reporting data now sitting in BigQuery. That was the reason for the move.
What are the most common AWS to Google Cloud migration mistakes?
Several come from treating Google Cloud as AWS with new names.
- Copying the AWS network and IAM design across. The models differ, and a copied design carries the old assumptions with it.
- Leaving DynamoDB and Step Functions until last. They are rewrites and belong in the plan from the first week.
- Keeping a Mumbai and Hyderabad recovery design. Google Cloud has no Hyderabad region, so recovery moves to Delhi.
- Going live before billing is ready. A new account without a GSTIN, or a large bill on card autopay, causes trouble in the first month.
- No rollback threshold. Agree in advance what would make you switch back, so nobody has to decide at two in the morning.
Frequently Asked Questions
How do I migrate from AWS to Google Cloud?
To migrate from AWS to Google Cloud, assess the estate in Google Cloud Migration Center, which imports asset inventory directly from AWS and produces cost estimates and total cost of ownership reports. Map each AWS service to its Google Cloud equivalent, build a landing zone with the organization, IAM and network design, then move workloads in waves. Use Migrate to Virtual Machines for EC2, Storage Transfer Service for S3, Database Migration Service for RDS and Aurora, and the BigQuery Data Transfer Service for Redshift, with continuous replication so each cut over is short.
Is GCP cheaper than AWS in India?
Neither GCP nor AWS is cheaper by default in India. Sustained use discounts apply automatically on self serve Cloud Billing accounts once a VM of an eligible machine series runs for more than a quarter of the month, up to 30 percent on N1, M1 and M2 and up to 20 percent on N2, N2D and C2. E2, N4, C3 and C4 get none, and invoiced billing accounts do not receive them. AWS has no automatic discount and needs a Savings Plan or Reserved Instance instead. Google Cloud committed use discounts reach up to 55 percent for most machine types and AWS Savings Plans up to 72 percent, but each is measured against a different on demand price, so the ceilings do not compare directly. BigQuery can be much cheaper than Redshift for spiky analytics. Compare INR quotes for your own architecture on both clouds.
Does AWS charge to move data to Google Cloud?
AWS offers free data transfer out to customers moving all of their data off AWS, or off one AWS service. Contact AWS Support and wait for approval before any part of the move starts. AWS reviews your plan and applies credits, and eligible customers have 90 days to complete the move. Specialised services such as CloudFront and Direct Connect are not covered. For S3 data, Storage Transfer Service can also pull from ap-south-1 and ap-south-2 over a Google managed private network with no AWS S3 internet egress charge, paying Google $0.08 per GiB within APAC instead.
What is the Google Cloud equivalent of AWS Lambda?
Cloud Run functions, previously called Cloud Functions, or Cloud Run for containerised services. Handler code usually ports with small changes, while triggers, permissions and packaging are redone.
What replaces DynamoDB on Google Cloud?
Firestore for document style workloads, Spanner where you need SQL and strong consistency at scale, or Bigtable for very large wide column workloads. Google publishes migration guides from DynamoDB to both Spanner and Bigtable, but there is no drop in replacement, so the data model and access code are rewritten.
Is the GKE control plane free?
No. GKE charges $0.10 per cluster hour in every mode, and the free tier covers the fee for one Autopilot or zonal Standard cluster per billing account. Amazon EKS charges $0.10 per cluster hour in standard support and $0.60 in extended support, and GKE clusters on the Extended channel also reach $0.60 per cluster hour once their version passes standard support.
Which Google Cloud region replaces AWS Mumbai and Hyderabad?
AWS Mumbai, ap-south-1, maps to Google Cloud Mumbai, asia-south1. Google Cloud has no Hyderabad region, so workloads in AWS ap-south-2 usually move to Mumbai, the closer region, with Delhi, asia-south2, as the in country recovery site.
Is there funding to migrate from AWS to Google Cloud?
Yes, for larger estates. Google Cloud funds qualifying migrations through its Rapid Migration and Modernization Program, RaMP, once each workload has a projected annual run rate of at least $60,000 of eligible Google Cloud spend. A general workload is then eligible for service funds of up to 20 percent of that run rate, capped at $2 million, plus quarterly Google Cloud credits on the growth of its tagged spend, and specialised types such as VMware, Oracle and SAP carry higher percentage rates. Smaller moves should be planned without counting on funding.
Will my billing still be in INR after moving from AWS?
Yes. Accounts with an Indian billing address are billed in INR by Google Cloud India Private Limited, with 18 percent GST on the invoice, and a registered business that adds its GSTIN can claim input tax credit. That matches how AWS India bills the accounts it sells. If you buy through a reseller such as Unico Connect, the reseller invoices you in INR with GST instead.
How can Unico Connect help you move off AWS?
Unico Connect is a certified Google Cloud partner and reseller in India, based in Mumbai, so your Google Cloud usage can be billed by us in INR with GST. We run AWS to Google Cloud migrations the way this guide lays them out, from the assessment of your AWS account and a landing zone for Indian regions to moves in waves with a rollback at every step, then cost optimization, FinOps and DevOps work after go live. Among the systems we built on Google Cloud, Highlands Brain, an AI powered learning platform for a charter school network in California, serves more than 15,000 students.
