This guide explains how to self-host Ollama on AWS with Cloud Forge. Cloud Forge does not host your app; resources are created in your cloud account.
Prerequisites
- An AWS account with permission to create compute, EIP, and security groups
- Access to the official AWS console
- Understanding of: Local LLM runtime and API server
Step 1: Choose launch settings
Choose AWS, us-east-1, and the application version in Cloud Forge. Select the AWS instance type in the launcher.
Step 2: Continue to CloudFormation
The deploy button opens the official provider console with the template and launch parameters prefilled.
Step 3: Review and create resources
Select VPC, subnet, key-pair, and other account resources, review costs, and confirm creation. Cloud Forge never receives your cloud credentials.
Step 4: Verify
After the stack succeeds, get the application URL from its outputs. Default service port: 11434.
Estimated cost
| Item | AWS (est.) | Aliyun (est.) |
|---|---|---|
| Compute | Quoted by AWS for the selected instance | Quoted by ROS for a region-available instance |
| Public IP | Current AWS rate | Current Alibaba Cloud rate |
| Service port | 11434 | |
| Marketplace AMI | Required; see the subscribed offer | N/A |
The catalog and launch service are free. AWS Marketplace software fees and provider resource charges are separate; review the official quote before creation.
AWS vs Aliyun differences
| Aspect | AWS | Aliyun |
|---|---|---|
| Default region | us-east-1 | cn-hongkong |
| IaC format | CloudFormation | ROS |
| First bootstrap | Pre-baked hardened AMI; total time varies by app | Public OS typically 8–15 min |
| Custom domain | Route53 parameters | Alidns parameters |
| HTTPS without domain | IP certificate | IP certificate |
Clean up
Delete the corresponding stack in the CloudFormation console to remove resources created by the template.