Three years on: how 'Terraform, OpenTF, and the Shrinking Gap' panned out

Three years ago I wrote about how I saw things panning out with Terraform and other parts of the Hashi stack in response to Hashicorp introducing the BUSL licence. Let’s see how that panned out 3 years on.

On OpenTF/OpenTofu specifically

Credibility arrived faster than predicted (“within the first year.”): OpenTF was renamed OpenTofu with a first stable release (v1.6) by January 2024 - about 5 months after the original announcement. OpenTofu has overtaken rather than just tracking Terraform: native state encryption in v1.7, provider-defined functions, provider for_each, loopable import/removed blocks, variables in backend blocks (v1.8), etc. - and other features including many that Hashicorp themselves strangely never quite got around to implementing despite requests. It became clear that this was the trajectory once the sponsors were funding more full time developers for years ahead than Hashicorp and Martin Atkins joined them. Governance landed with the CNCF (accepted April 2025), a credibility marker I couldn’t have specified at the time. HC Terraform is still very much around, although I would hazard that a lot of HC Terraform use is at this point legacy- old versions that have not yet been updated.

On “Terraform won’t hold its crown” (writing HCL by hand becoming irrelevant)

I would say directionally right but different in detail. AI-generated/assisted IaC is now completely normal practice, not a novelty - the “shrinking gap” between provider contract and language spec has kept shrinking. HCL is still the artifact being generated, reviewed, and debugged however- nobody’s yet successfully abstracted a level above it, e.g. Wing shut down. On the other hand, it is perfectly possible to vibe code infrastructure, just with the risk of it having signifficant problems if you don’t know what you’re doing.

The Stack Call (Consul/Packer/Boundary/Nomad/Vagrant/Waypoint)

I will call this largely correct: none of the “niche, little impact” products as identified 3 years ago triggered a fork or community backlash. They just continued under BUSL, now IBM-owned. One partial exception: Waypoint’s open-source Community Edition was archived in January 2024 as Hashicorp changed direction for it from a deployment service to a workflow service catalogue SAAS. People joked about Otto an earlier, now-erased HashiCorp product at the time - abandoned within about a year, and omitted from HashiCorp’s own later histories. Turns out that was the right call, the thing that existed in 2023 is now gone.

What I did not anticipate in 2023 was Vault getting the same treatment as Terraform- forked in late 2023 as OpenBao, reaching a production-ready 2.0 by September 2024, OSSF sandbox status June 2025, with real adopters (Nvidia). Ironically it was built by IBM’s own engineers - the same company that then acquired HashiCorp outright.

Sale to IBM

Whilst it was obvious at the time that the licence change was intended to make Hashicorp a more attractive market for prospective purchasers, it wasn’t apparent that there actually were any at that time. IBM acquired HashiCorp for $6.4B, closing February 2025. Rather than reversing the BSL move, IBM has built further on it.

In Summary

I’d say a lot of the previous inividual predictions were broadly correct but with some important differences in the real world, especially around the emphasis on the shrinking gap. Whilst HC Terraform is still very much around, it’s clear that OpenTofu is very much alive, credible and advancing.