We are introducing Terraform support for managing and provisioning Verda infrastructure.
With the Verda Terraform provider, our users can define GPU instances, storage, and startup scripts as code, enabling reproducible deployments and integration into existing infrastructure-as-code (IaC) workflows.
Terraform, developed by HashiCorp, is a popular IaC tool that allows teams to provision and manage infrastructure declaratively using version-controlled configuration files.
Terraform provider
The Verda Terraform provider is available on the Terraform Registry and supports provisioning and managing core Verda resources, including GPU instances, storage volumes, SSH keys, and startup scripts.
The current version does not support Instant Clusters or Serverless Containers.
To get started, see:
- The Verda Terraform provider documentation
- A complete example deployment of DeepSeek-R1 inference with SGLang, using NVFP4 quantization on NVIDIA Blackwell Ultra GPUs
OpenTofu compatibility
The Verda Terraform provider is fully compatible with OpenTofu, the open-source fork of Terraform.
OpenTofu is licensed under MPL 2.0 (as opposed to Terraform’s BSL), making it a preferred option for teams with strict open-source or licensing requirements.
The provider is also available through the OpenTofu Registry with complementary documentation.
Provisioning options
Terraform and OpenTofu complement existing ways of interacting with Verda infrastructure. Depending on your workflow, you can also provision resources using:
- Cloud Console (UI)
- API
- Python SDK
- dstack