GCP vs Azure, from someone who shipped both
Two months ago I was operating a production stack on Azure. Giant Context, the website platform I build now, has run on Google Cloud since its first commit. The industry treats switching clouds as a career change, acquired once and altered only through years of paperwork. It is closer to learning the other name for things.
At my previous company I ran two years of production on Azure. Twenty-one function apps, a service bus, a Kusto cluster for analytics, a key vault, B2C identity, blob storage, and App Insights watching all of it. Real scale, real incidents, real bills. When I left and started Giant Context on GCP, the entire conceptual translation came to this:
| Azure | Google Cloud |
|---|---|
| AD B2C | Identity Platform |
| API Management | API Gateway |
| App Insights | Cloud Logging + Monitoring |
| Azure OpenAI | Gemini on Vertex |
| Blob Storage | Cloud Storage |
| Data Explorer (Kusto) | BigQuery |
| Functions | Cloud Run |
| Key Vault | Secret Manager |
| Service Bus | Pub/Sub |
A compute runtime is a compute runtime. A queue is a queue. A secret store is a secret store. Every operational instinct I built on the left column (how to think about cold starts, dead letters, log retention, key rotation) transferred whole to the right column.
Both of these stacks are incredible
Neither cloud is a mistake, and any account claiming otherwise would be false. Azure and GCP are both astonishing feats of engineering. In two years on one and two months on the other, I have not once hit a wall where the platform could not do the thing I imagined needing. Anything I can conceive of building, both of them host, queue, store, secure, and observe. Capability is a tie.
What is not a tie is the experience of being one person inside them.
Who each cloud is built for
Azure feels like enterprise software. GCP feels like it was built for a solopreneur. That is a statement about who each platform believes you are, not a euphemism for better and worse. Azure assumes an organization, with procurement, subscriptions arranged in management groups, portals within portals, and a governance story before a deploy story. GCP assumes a person with a project ID and a terminal. The CLI tools are the cleanest expression of the difference. gcloud is built for a human to type, and I reach for it without dread. The interface repeats it. I simply like being in GCP's console, and after two years I never learned to like being in Azure's.
And GCP is cheaper. For this company's shape (one operator, many small services, AI at the center) the bill runs lower for the same work, and it scales the same direction Azure does. Cheaper and as scalable ends the deliberation for a startup, even before preference gets a vote.
Azure feels like enterprise software. GCP feels like it was built for a solopreneur.
The AI bet
One reason in the choice was a hunch about the future. I thought Google would lead the way in AI, and betting the platform's center of gravity meant betting on whose models I would live closest to. Keeping everything under one roof (models, data, identity, billing) was already the doctrine, and Gemini sitting in the same cloud as the database made Google's the natural choice. Two months in, they might be proving the hunch right. We will see.
What Giant Context actually runs on today is a short list:
What Giant Context runs on today
Cloud SQL Postgres, pgvector includedCloud Storage every customer file and media assetIdentity Platform auth (the one piece of Firebase that stayed)BigQuery the analytics and cost events, immutableGemini every generation the product performsThe transfer is the point
Cloud skills are treated, by hiring pipelines and by engineers themselves, as if they were the platform. Azure engineer, AWS shop, GCP-certified. My experience says the platform-specific layer is a few weeks thick, and everything under it (the instincts about queues, identity, blast radius, cost, observability) is the actual skill, and it moves with you.
I know this because I wanted the skills transfer. Two years of Azure production converted into a GCP platform without a course, a cert, or a consultant. The phrasebook was one table long. If you can run one of the big three in production, you already know the other two. You just haven't met them yet.
Have questions?
If you're interested in my work or Giant Context, contact me!