Why I Chose GCP over AWS and Azure for Ollama and Open WebUI

Adapted from my LinkedIn article: Why I chose GCP over AWS and Azure for ollama and open webui ?.

I wanted to run Ollama + Open WebUI in cloud with spot pricing. Local hardware was not giving the performance I wanted.

Evaluation goal #

• run GPU-backed spot instance
• keep cost predictable
• deploy quickly without platform friction

AWS experience #

I started with AWS because I already used it for Kubernetes work.

• tested small spot instances successfully
• GPU spot request failed due to service quota (All G and VT Spot Instance Requests)
• quota-increase request remained pending for hours

Azure experience #

I was newer to Azure at that time, so some friction may have been user-side.

• too many options and workflow branches for this specific task
• could not quickly reach a clear GPU spot workflow
• did not feel smooth for this weekend experiment

GCP experience #

GCP was the most practical for this use case.

• clearer UI for VM + pricing configuration
• one-click generation of CLI/API/Terraform snippets
• quota request flow was faster in my case
• successful launch of a GPU spot VM after approval

After SSH access, getting Ollama and Open WebUI running was straightforward with Docker.

Why GCP won for this test #

1. Faster path from idea to running GPU VM.
2. Better visibility into pricing while configuring resources.
3. Easier conversion from manual setup to IaC/CLI workflows.

Final note #

This was a practical comparison from one specific setup and time window, not a universal ranking.

My planned next step was to codify equivalent Terraform flows for all three clouds so experimentation can stay reproducible and vendor-neutral.