Compute Model
The fundamental unit of compute each cloud/hosting platform runs your code as.
AWS
Offers essentially every compute model available: raw virtual machines (EC2), containers (ECS/EKS/Fargate), and functions (Lambda) — you choose the level of abstraction rather than the platform choosing for you.
GCP
Similarly broad: Compute Engine for VMs, Google Kubernetes Engine for containers, Cloud Run for serverless containers, and Cloud Functions for individual functions.
Azure
Also broad: Virtual Machines, Azure Kubernetes Service for containers, Azure App Service for managed web apps, and Azure Functions for serverless.
Vercel
Deploys applications as a combination of static assets, serverless functions, and edge functions, automatically deciding the right compute target based on how a Next.js (or other framework) app is structured.
Cloudflare
Centers on Workers, V8-isolate-based functions that run at the edge with near-instant cold starts, a fundamentally different (and lighter-weight) unit than a traditional container or VM.
Netlify
Deploys static assets plus serverless and edge functions, similar in shape to Vercel’s model but usable with a broader range of frontend frameworks and static site generators.
DigitalOcean
Offers Droplets (VMs), a managed Kubernetes service, App Platform (a PaaS for containers/apps), and Functions (serverless), covering a similar spread to the hyperscalers at smaller scale.
Render
Runs long-lived web services and background workers as containers, alongside static site hosting and managed Postgres/Redis — a PaaS-style model without a raw VM option.
Fly.io
Runs applications as lightweight, hardware-virtualized VMs called Fly Machines that start in milliseconds, positioned as full application containers rather than short-lived functions.
Heroku
Runs applications in “dynos,” lightweight containers that scale horizontally, the original model that popularized the git-push PaaS pattern.