Terraform

Terraform is better for teams that want a stable, battle-tested tool with the largest provider ecosystem and community — its HCL language, while domain-specific, is simple enough for non-developers (DevOps/platform engineers) to learn, and there's a massive library of community modules and examples. Pulumi is better for developer-led teams that want to use general-purpose programming languages (TypeScript, Python, Go, C#) for infrastructure — enabling unit testing, loops, functions, and abstractions that HCL handles clumsily. Pulumi's real-programming-language approach also integrates naturally with application code in the same repository. For most teams without strong software engineering culture, Terraform's simplicity wins. For developer-led platform teams who find HCL limiting, Pulumi's expressiveness is valuable.

In August 2023, HashiCorp changed Terraform's license from the Mozilla Public License (MPL 2.0 — permissive open source) to the Business Source License (BUSL 1.1). The BUSL restricts 'competing with HashiCorp' — specifically, building products or services that compete with Terraform Cloud/Enterprise. This means you can use Terraform internally (free), but you cannot build a managed Terraform service to sell commercially without a HashiCorp commercial license. The change prompted significant community backlash. The Linux Foundation quickly established the OpenTofu project (formerly OpenTF) as a true open-source fork of Terraform under the MPL license. OpenTofu is now at feature parity with Terraform 1.x and is supported by major cloud providers and infrastructure vendors as the community alternative.

Terraform state is a JSON file (terraform.tfstate) that maps your Terraform configuration to real-world infrastructure resources. When you create an AWS EC2 instance with Terraform, the state file records that specific instance's ID — so Terraform knows which real resource corresponds to which configuration block. Without state, Terraform can't determine what exists vs. what needs to be created/modified/deleted. State also stores metadata for dependency tracking and performance optimization. The challenge with state: it can contain sensitive values (passwords, connection strings), it must not be lost (losing state = orphaned infrastructure), and concurrent modifications cause state conflicts. Terraform Cloud, S3 backends, and Terraform Enterprise solve these problems with remote state storage, locking, and versioning for team environments.