Ok, so first - AWS Copilot is CloudFormation under the hood, but the way it works results in you not thinking about CFN anymore. AWS found the right balance with Copilot - it's insanely simple to setup production-ready multi-account environment with many services inside, with CI/CD out of the box etc etc. It's pretty new, but even now it was enough to launch Transcripto, which uses may be a dozen of different AWS services, all bound together by Copilot.

Because Pulumi uses real programming languages, you can actually write abstractions for your infrastructure code, which is incredibly empowering. You still 'describe' your desired state, but by having a programming language at your fingers, you can factor out patterns, and package it up for easier consumption.

We use Terraform to manage AWS cloud environment for the project. It is pretty complex, largely static, security-focused, and constantly evolving.

Terraform provides descriptive (declarative) way of defining the target configuration, where it can work out the dependencies between configuration elements and apply differences without re-provisioning the entire cloud stack.

Terraform is vendor-neutral in a way that it is using a common configuration language (HCL) with plugins (providers) for multiple cloud and service providers.

Terraform keeps track of the previous state of the deployment and applies incremental changes, resulting in faster deployment times.

Terraform allows us to share reusable modules between projects. We have built an impressive library of modules internally, which makes it very easy to assemble a new project from pre-fabricated building blocks.

Software is imperfect, and Terraform is no exception. Occasionally we hit annoying bugs that we have to work around. The interaction with any underlying APIs is encapsulated inside 3rd party Terraform providers, and any bug fixes or new features require a provider release. Some providers have very poor coverage of the underlying APIs.

Terraform is not great for managing highly dynamic parts of cloud environments. That part is better delegated to other tools or scripts.

Terraform state may go out of sync with the target environment or with the source configuration, which often results in painful reconciliation.

I personally am not a huge fan of vendor lock in for multiple reasons:

• I've seen cost saving moves to the cloud end up costing a fortune and trapping companies due to over utilization of cloud specific features.
• I've seen S3 failures nearly take down half the internet.
• I've seen companies get stuck in the cloud because they aren't built cloud agnostic.

I choose to use terraform for my cloud provisioning for these reasons:

• It's cloud agnostic so I can use it no matter where I am.
• It isn't difficult to use and uses a relatively easy to read language.
• It tests infrastructure before running it, and enables me to see and keep changes up to date.
• It runs from the same CLI I do most of my CM work from.

GCE is much more user friendly than EC2, though Amazon has come a very long way since the early days (pre-2010's). This can be seen in how easy it is to edit the storage attached to an instance in GCE: it's under the instance details and is edited inline. In AWS you have to click the instance > click the storage block device (new screen) > click the edit option (new modal) > resize the volume > confirm (new model) then wait a very long time. Google's is nearly instant.

• In both cases, the instance much be shut down.

There also the preference between "user burden-of-security" and automatic security: AWS goes for the former, GCE the latter.