The future of CDKTF

A. CDKTF Sunset & Community Fork

• IBM/HashiCorp archived CDKTF org without notice
• Community rallied via Open Constructs Foundation (OCF)
• TerraConstructs initiated the fork, and helped gather contributors
• Internal code bases suddenly surfaced with real TF-CDK usage
• Conclusion: out of crisis comes community ownership

What Happened

While preparing for the 2025 year in review and rounding off a productive year for TerraConstructs, a significant event sent a shockwave through the CDK community.

IBM, the company that acquired HashiCorp, announced its intention to sunset the CDK for Terraform (CDKTF) project. The announcement was abrupt. The README was updated to indicate a future shutdown date - which turned out to be the same day - and shortly thereafter the core repository, along with every pre-generated CDKTF provider bindings repository, was archived.

For many teams, this was not an abstract decision. CDKTF underpins real production systems, internal platforms, and years of accumulated infrastructure code. The sudden nature of the announcement surfaced something that had been quietly true for a long time: despite years of minimal maintenance and unresolved community issues, CDKTF had become a critical dependency for a surprising number of organizations.

There is, however, a positive outcome.

The King is dead, Long live The King

Because of earlier efforts to explore a community fork - motivated initially by the growing backlog of issues and stagnation in upstream development - the groundwork already existed. With a clear signal that the project would no longer be stewarded by IBM/HashiCorp, the community responded quickly.

The Open Constructs Foundation (OCF) offered a neutral home for a community-driven continuation of the project. This is the same foundation that hosts the long-standing CDK community Slack at cdk.dev, organizes CDK Day, and maintains newsletters and community infrastructure. Their involvement provided immediate legitimacy and continuity.

Once the future of CDKTF was no longer ambiguous, contributors surfaced almost overnight. Multiple engineers came forward to describe substantial internal codebases built on top of CDKTF — systems that now required a maintained and trustworthy path forward.

As part of this effort, We at TerraConstructs, created a fork of the Terraform CDK codebase and adopted a new project name (currently provisional). Discussions are ongoing with OCF and IBM/HashiCorp to explore whether stewardship of the original project can be formally returned to the community it originated from.

Most importantly, the decisive action changed the tone of the conversation. The same day we had a working and renamed version. However we held off on publishing to avoid confusion when the final name is confirmed.

At the same time, community members stepped up and also started submitting fixes and new features for TerraConstructs itself, and making their first contributions. Momentum replaced uncertainty.

The initial news was undeniably destabilizing. But with clarity comes coordination, and with coordination comes progress. The outcome so far suggests that the future of typed, higher-level Terraform abstractions is not ending - it is simply changing hands.

In that sense, the future is brighter than it first appeared.

B. HCP Classic Sunset & Real Workload Cost Shifts

The HCP free tier / legacy plan is being sunset

This week significant news spread through the Terraform community: HashiCorp announced the legacy HCP Terraform Free plan will reach end-of-life on March 31, 2026. The plan will be migrated to the enhanced Free tier introduced in 2023, which supports up to 500 managed resources with unlimited users and additional capabilities like SSO and policy-as-code support.

• Forced migration to enterprise pricing (e.g. $15k+ annually for some teams)
• Teams respond by moving to S3 backends or self-hosted solutions
• A compelling argument for solutions like GridAPI + Atlantis workflows