Last Year of Terraform (encore.dev)

A recent analysis highlights a significant shift in software development, revealing that AI agents are now capable of generating production backend code, with enterprise code being 30-50% AI-generated. However, many teams encounter delays due to the traditional infrastructure provisioning processes, primarily utilizing tools like Terraform. The discrepancy arises because while AI can rapidly produce application code, the corresponding infrastructure setup often requires extensive manual review, planning, and synchronization, leading to bottlenecks that hinder delivery. This situation underscores a pressing need for a new approach to infrastructure management. Many companies are beginning to explore alternatives that eliminate the separation between application code and infrastructure code. By integrating infrastructure provisioning directly within the application code—without relying on separate Terraform files and associated review processes—developers can achieve continuous and frictionless deployments. Such innovations not only streamline workflows but also bring infrastructure under the same guardrails and verification as application logic, enhancing overall agility. As AI continues to transform development practices, this shift toward unified codebases may redefine how teams approach backend architecture, making the case for outdated systems like Terraform increasingly untenable.