Google Cloud is going after AWS’s installed base with an AI agent that moves Kubernetes workloads from Amazon EKS to Google Kubernetes Engine. Announced October 5 as part of a portfolio called Google Cloud Modernize, the EKS-to-GKE agentic migration is in public preview, according to a report from Cloud Computing News. It targets one of the stickiest parts of a cloud contract: the container platform where applications actually run.

What the Agent Translates

The agent scans an EKS cluster or repository, identifies dependencies and AWS-specific configuration, and generates Terraform modules. It translates IAM Roles for Service Accounts into Google Cloud Workload Identity, Application Load Balancer ingress into the Kubernetes Gateway API, and Karpenter node configurations into GKE node provisioning tools. It also handles Kubernetes manifest translation and storage and networking mappings.

Migration runs in five stages: discovery, planning, infrastructure generation, application translation and validation. Validation uses Terraform and Kubernetes manifest checks before a human reviews the output.

Built With Guardrails for Enterprise Risk

Google is careful about what the agent does not do. It does not apply changes to a live cluster. Instead it creates pull requests that go through existing CI/CD and approval workflows. It does not move stateful data, and it generates runbooks that point to Database Migration Service or Storage Transfer Service. According to analysts cited by Network World, application testing, architecture decisions, data migration and production cutover remain outside the agent’s scope.

That limit defines the real offer. Google is automating the tedious translation work and leaving the risky decisions with customers, which lowers the cost of starting a migration but not the cost of finishing one.

A Broader Modernization Portfolio

Modernize brings existing migration and modernization services together with new AI-assisted tools. The components are Migration Center, Google Cloud VMware Engine, Google Cloud Mainframe Modernization, a new Modernization Hub and the EKS-to-GKE agent.

The Modernization Hub is a console interface for analyzing source code and mapping application dependencies for Java, .NET and mainframe applications. It includes an App Modernization CLI, called CodMod, which uses Gemini models to analyze repositories, identify dependencies and recommend changes, including moving .NET Framework apps to .NET on Linux containers.

Migration Center now builds on AWS discovery support that assesses EC2-related infrastructure, EBS, Application Load Balancer, ECR, DynamoDB, API Gateway, Glue, Kinesis and ElastiCache. Its Quick TCO Estimator adds Gemini-based features: customers can upload VMware inventory data or enter infrastructure manually, model workloads on Compute Engine, and get one-year and five-year total cost comparisons that include licensing, support, discounts and migration costs.

Mainframes and Licensing Costs

The mainframe tools include a Mainframe Assessment Tool that analyzes source code and dependencies, a Mainframe Connector that converts mainframe-format data such as EBCDIC for Google Cloud services, and Dual Run, which runs workloads on the existing mainframe and on Google Cloud in parallel and compares outputs before cutover.

On the infrastructure side, Google introduced the M4N series with 26.57 GiB of memory per vCPU and support for Hyperdisk Extreme. Google says that memory-to-core ratio can cut licensing costs by more than 20 percent for Oracle and other core-licensed databases. For VMware customers, Google offers self-managed VMware on Bare Metal Z3 instances running VMware Cloud Foundation 9.1, connected to Compute Engine, GKE and BigQuery through a Google Cloud VPC.

What It Means for AWS and Microsoft

Pareekh Jain, principal analyst at Pareekh Consulting, told Network World that organizations with mixed VMware, Java, mainframe, Oracle and Kubernetes workloads have traditionally needed separate assessment tools and teams. Google has now consolidated several of those under one name.

The competitive target is plain. AWS’s EKS is the default home for many Kubernetes workloads, and Google is offering to move them with an automated first pass. Microsoft, whose Azure Migrate and Hybrid Benefit licensing programs court the same VMware, .NET and SQL Server shops, faces a rival pitching lower Oracle licensing costs and agent-generated migration pull requests. Google’s approach is to attack the switching cost directly, and it leaves AWS and Microsoft to prove that staying put is cheaper than leaving.