Terraform

Terraform has been the default way to provision cloud infrastructure for more than a decade, and in 2026 it still anchors most of the platform stacks I see. The ground under it has shifted though. HashiCorp moved the tool to the Business Source License in 2023, the community answered with the OpenTofu fork under the Linux Foundation, and IBM closed its acquisition of HashiCorp in 2025. Teams that once treated Terraform as the single obvious choice now weigh two compatible tools that are slowly drifting apart in features and governance.

>>> The OpenTofu and Terraform split created a fragmented ecosystem, but OpenTofu has firmly established itself as a viable, open-source enterprise standard backed by the Linux Foundation. For a new platform team in 2026, standardizing on OpenTofu minimizes licensing and vendor lock-in risks while retaining semi-full syntax compatibility.

None of that churn has dented the fundamentals. The provider ecosystem remains the widest in the IaC space. The plan and apply loop is still the clearest contract between an engineer and a cloud API. State is still the thing that bites when process gets sloppy. Most failures I see in Terraform estates have nothing to do with licenses. They come from weak module boundaries, missing tests and drift that nobody notices until an apply goes sideways. The posts collected here deal with those problems directly.

The bigger shift is in who writes the code. AI coding agents now produce a growing share of new HCL. They scaffold modules, wire up providers and chase plan errors quickly, which moves the human effort elsewhere: reviewing plans, testing modules and holding the line on drift. Infrastructure code is a good fit for agents because the feedback loops are explicit. A plan tells you what will change before anything does, and policy checks run the same way whether a person or an agent opened the pull request. I track that intersection on theAI agents hub, and theClaude Code for infrastructure as code post shows what an agent-driven Terraform workflow looks like in practice.

>>> Before allowing AI agents to write Terraform against production state, you must enforce strict read-only least privilege access, changes still go through vetted CI/CD pipelines, automated policy-as-code validation (like OPA or Sentinel), mandatory human-in-the-loop plan reviews until you build trust, and state locking.

This page organizes the Terraform writing on the site into three tracks. Fundamentals collects the pieces I still send to engineers who are new to the tool: practices, output values and how Terraform relates to Kubernetes. Workflows and automation is about getting applies off laptops and into pipelines, whether that means GitHub Actions, Atlantis or a dedicated platform. Migration, testing and drift covers the operational side: moving to OpenTofu, testing strategies, drift detection and bringing existing AWS estates under management. Most entries are link posts, meaning short commentary here plus a pointer to the full article published with partners such as Spacelift, StackGuardian and the AWS blogs. Each link carries a one-line note so you can decide what to open. For everything in date order, use the tag archive at the bottom or the mainblog feed.