Cloud infrastructure is a growing contributor to global carbon emissions. However, existing sustainability assessment tools predominantly focus on post-deployment monitoring or require access to application code, limiting their utility for proactive, pre-deployment decision-making. To address this gap, we present LEAF, a lightweight, layered framework – developed with Lloyds Offshore Global Services Private Limited – that enables early-stage estimation of carbon footprint from terraform configurations. LEAF employs a three-layer modeling approach to (i) interpret Infrastructure-as-Code (IaC) definitions, (ii) simulate expected workload behavior, and (iii) integrate region-specific carbon intensity data. By parsing IaC files and modeling system behavior, LEAF provides energy and emission estimates without requiring access to proprietary source code. Evaluations of LEAF across three real-world projects demonstrates its ability to enable pre-deployment comparisons, with carbon footprint simulations that align with post-deployment measurements within 15.7% for realistic (high) workloads, enabling architects and developers to make sustainability-informed decisions early in the infrastructure provisioning process.