← Back to all articles Server-Side Tracking & CAPI Clients only November 13, 2026 ⏱ 15 min read

Hosting sGTM on AWS Fargate vs. Google Cloud Run: Cost and Performance

The choice of cloud provider for the sGTM container affects latency and the IT budget. We take a detailed look at AWS Fargate versus Google Cloud Run.

Random Performance Engineering Paid traffic & server-side tracking specialists

November 13, 2026

Hosting sGTM on AWS Fargate vs. Google Cloud Run: Cost and Performance

TECHNICAL DIAGNOSIS · OPERATIONS > BRL 100K/MONTH 15 min read

Operating scenario: Technology teams face unpredictable cloud computing costs, or server container outages during traffic peaks from promotional campaigns.

Technical root cause: Incorrectly sized containers and a poor grasp of how egress traffic is billed and how cold start behaves on AWS and GCP.

Engineering guideline: An architectural decision based on the company's profile: Cloud Run for fast deployments and automatic serverless scaling; AWS Fargate for operations with an existing Amazon ecosystem.

The Infrastructure Decision for Server-Side GTM

Deploying a Google Tag Manager Server-Side (sGTM) container requires provisioning dedicated cloud servers to process the stream of analytics requests. While Google offers an automated default setup on Google Cloud Platform (GCP), most mid-sized and large companies have contracts and corporate infrastructure hosted on Amazon Web Services (AWS).

Choosing between Google Cloud Run and AWS Fargate involves technical trade-offs across four critical dimensions: response time (network latency), tolerance to traffic spikes, operational maintenance complexity and monthly financial cost.

Architecture 1: Google Cloud Run (Google's Native Standard)

Cloud Run is Google's serverless compute platform that runs containers directly on the company's global infrastructure. Its main engineering characteristics include:

One-Click Automated Deploy: Native integration with the Google Tag Manager panel, letting you spin up test and production instances without writing infrastructure-as-code (IaC) manifests.
Per-Container Concurrency: A single Cloud Run instance can process up to 80 simultaneous requests by default, reducing the total number of active containers needed.
Scale to Zero: The ability to shut down instances overnight when there is no traffic, although in production we recommend keeping a minimum of 2 instances always active to avoid cold start latency.

Architecture 2: AWS Fargate (Amazon Corporate Environment)

AWS Fargate lets you run Docker containers inside Amazon ECS (Elastic Container Service) without provisioning or managing EC2 virtual instances:

Integration with the Internal AWS Network: Ideal for organizations whose application servers, databases and CI/CD pipelines already live on AWS, keeping all traffic inside the same VPC (Virtual Private Cloud).
CPU Steadiness: Fargate instances stay continuously active with dedicated memory and vCPU allocation, completely eliminating cold start variation.
Load Balancer (ALB) Required: You need to set up an Application Load Balancer in front of the instances to distribute connections and manage SSL certificates.

Google Cloud Run

Simplified deploy native to GTM
Billing by fraction of a second and concurrency
Ultra-fast elastic scalability
Estimated cost: $40 to $110/month for medium traffic

AWS Fargate (ECS)

Full integration with VPC and corporate security
Stable performance with no cold start at all
Requires DevOps knowledge and ALB configuration
Estimated cost: $70 to $160/month with a Load Balancer

Frequently Asked Questions about sGTM Infrastructure

Which of the two offers lower latency in Brazil?

Both have data regions in São Paulo (southamerica-east1 on GCP and sa-east-1 on AWS). When configured in the Brazilian region, response latency for local users stays under 30 milliseconds.

Can I run sGTM on ordinary VPS servers (such as DigitalOcean or Hetzner)?

Technically, you can run the sGTM Docker image on any Linux server. However, you lose automatic elastic scaling during traffic peaks from large campaigns, which calls for constant manual monitoring by a sysadmin.

Need specialist support to provision your sGTM container?

Random Marketing's engineering team designs and implements resilient cloud infrastructure for large advertisers.

I want to hire Random Marketing for my company ↗

RESERVED DOCUMENT · CLIENTS ONLY

Restricted access for advisory clients

The architecture models, verification checklists, integration scripts and operating pipelines in this dossier are restricted to companies advised by Random Marketing.

✓ Audited technical implementation checklist

✓ Architecture and configuration models

✓ Data governance and engineer follow-up

How do I become a client?

Advisory for operations investing BRL 100k or more per month in media.

TECHNICAL RIGOR & PRIMARY SOURCES

Official documentation & engineering references

Architecture guidelines, API specifications and official technical documentation consulted to support this dossier:

Google Cloud Cloud Run Pricing and Concurrency Architecture Billing model by millisecond of allocated CPU and concurrency per container.

Amazon Web Services AWS Fargate Serverless Compute for Containers Specification of containerized instances without managing the underlying EC2 servers.

Indexing keywords: #sgtm aws fargate #sgtm google cloud run #sgtm cloud infrastructure #sgtm server cost #server-side tagging performance