Light Detection and Ranging (LiDAR) has became an indispensable technology across diverse domains including agriculture, geology, urban planning, and transportation due to its capacity to generate precise, high resolution three dimensional spatial data. However, the substantial volume of data produced by high-density LiDAR scans imposes significant burdens on storage infrastructure, data transmission bandwidth, and computational resources. As such, effective compression techniques are essential to enable scalable LiDAR data management while maintaining the integrity and fidelity required for analytical applications. This paper introduces a new lossless compression framework tailored for LiDAR data stored in the standard LiDAR Aerial Survey (LAS) format. The proposed approach integrates a neural network–based autoencoder with the Lempel–Ziv–Markov chain (LZMA) compression algorithm to generate a highly compact latent representation of LiDAR data. Experimental results demonstrate a compression factor of 17.21:1 with an average efficiency of 94.16%, significantly outperforming both general purpose compressors (e.g. WinRAR) and specialized tools (e.g. LASzip). By substantially reducing data size without compromising quality, the proposed method offers a robust and scalable solution for the efficient handling of large-scale LiDAR files, making it well suited for both operational and resource-constrained environments.