Terraform : Infrastructure as Code 2026

Terraform est le standard IaC depuis 2014. Mais 2023 : HashiCorp change la license (BSL non-OSI), fork OpenTofu créé par Linux Foundation. 2026 : OpenTofu adoption croissante, Terraform Cloud devient HCP Terraform, Pulumi grignote part premium.

TL;DR

- Terraform : standard IaC, mais license BSL controversée 2023.

- OpenTofu : fork OSS Linux Foundation, adoption en 2024-2026.

- HCP Terraform (cloud) : enterprise managed.

- Alternatives : Pulumi (programming langs), CDK (cloud-specific).

Terraform & OpenTofu

Différences 2026

IBM acquiert HashiCorp (2024)

• $6.4B deal annoncé 2024
• Concerne Terraform, Vault, Consul, Nomad, Boundary
• Incertitudes communauté → adoption OpenTofu accrue

HCL example typique

`hcl

terraform {

required_providers {

aws = {

source = "hashicorp/aws"

version = "~> 5.0"

}

}

backend "s3" {

bucket = "tf-state-clientelis"

key = "prod/terraform.tfstate"

region = "af-south-1"

}

}

provider "aws" {

region = "af-south-1" # Cape Town

}

resource "aws_instance" "api" {

ami = "ami-0c55b159cbfafe1f0"

instance_type = "t3.medium"

tags = {

Name = "kolonell-api"

Environment = "prod"

}

}

resource "aws_db_instance" "postgres" {

identifier = "kolonell-db"

engine = "postgres"

engine_version = "16.1"

instance_class = "db.t3.medium"

allocated_storage = 100

storage_encrypted = true

}

`

HCP Terraform (anciennement Cloud)

• Remote state managed
• Workspaces collaborative
• Run pipelines (plan/apply automated)
• Policy-as-code (Sentinel)
• Free tier : 500 resources
• Standard : $0.00014/resource/h (~$10/mois 100 resources)
• Plus : $20/user/mois

Alternatives 2026

Pulumi

• Code IaC en Python, TypeScript, Go, C#, Java
• Pas DSL HCL custom
• Pricing : Free → $700+/mois Enterprise
• Adoption : équipes dev qui veulent typage IDE

AWS CDK / CDKtf

• AWS-specific (CloudFormation under hood)
• TypeScript, Python, Java
• Free (CloudFormation gratuit)

Crossplane

• Kubernetes-native IaC
• Infrastructure as Kubernetes resources
• GitOps-friendly

Ansible (immutable infra)

• Plus config management que pure IaC
• YAML playbooks
• Pas state file

Bicep (Azure)

• Azure-specific DSL
• ARM templates compilés

OpenTofu migration

Trivial pour 99 % cases :

`bash

# Replace binary

brew uninstall terraform

brew install opentofu

# Same commands

tofu init

tofu plan

tofu apply

# State compatible

# Existing .tf files work as-is

`

Pattern moderne 2026

`

Code .tf (Git versioned)

→ CI/CD (GitHub Actions / GitLab CI)

→ tofu plan (preview)

→ Review PR (Atlantis ou HCP Terraform)

→ tofu apply (after merge)

`

Use cases Africa

Cloud providers Africa

• AWS : Cape Town (af-south-1)
• Azure : Johannesburg (south africa north)
• GCP : Johannesburg
• OVH, Scaleway : pas natif Africa mais Europe close
• Digital Ocean : Bangalore (proche), expanding

Tenant typique startup

• 50-200 resources (VPC, subnets, RDS, EC2, S3, IAM)
• 2-5 environnements (dev, staging, prod)
• Coût Terraform/OpenTofu : $0 (self-managed state)
• Coût HCP Terraform : $10-50/mois

FAQ

Q : Migrer vers OpenTofu ?

R : Oui si compliance OSI strict ou méfiance IBM. Non si stack mature avec Sentinel policies (HCP-only) ou HashiCorp support contracts.

Q : Terraform vs Pulumi ?

R : Terraform/OpenTofu plus mature, écosystème vaste. Pulumi meilleur DX pour dev devops avec typage IDE. Coût Pulumi enterprise tier élevé.

Q : State file stockage ?

R : S3/Spaces + DynamoDB lock standard. HCP Terraform pour managed. JAMAIS state file en Git (secrets).

Conclusion

Terraform 2026 reste standard IaC malgré controverse BSL. OpenTofu monte rapidement comme alternative OSS. Pour startups Africa, OpenTofu + state S3 (DO Spaces ok) coût zéro. Enterprise : HCP Terraform pour collaboration + policies. Pulumi excellent niche pour équipes dev-first.