NASA’s Global Imagery Browse Services (GIBS) delivers an enormous amount of Earth observation imagery through publicly accessible, standard web services. Billions of image tiles have been served to users around the world, across over 1100 Earth Science data products. GIBS serves as the backbone to popular NASA websites such as Worldview (web client for GIBS imagery) and Eyes on the Earth, and has become popular due to its straightforward integration into GIS applications such as QGIS and ArcGIS. This year, GIBS completed its multi-year transition from an on-premises implementation to a cloud-native implementation, allowing for greater scalability to accommodate new Earth science missions with massive amounts of data, such as PACE, SWOT, and NISAR. This transition involved creating a new open-source cloud-optimized implementation of OnEarth, the open-source image server of GIBS. Currently, imagery served is pre-rendered and stored in S3 buckets as PNG and JPG MRFs.Due to ever-growing storage demands and associated costs, the OnEarth team recently added support for a number of GDAL-supported image compression algorithms such as LERC, brunsli, and ZenJPEG, each serving an important purpose across different parts of the image collection. For example, LERC provides floating-point precision that is crucial for scientific usability and is not available through formats such as JPEG, which suffers from compression loss in pre-rendering as images are converted and compressed, and PNG, whose 8-bit depth incurs precision loss. Furthermore, LERC support is a subset of a larger effort to investigate potential solutions for imagery to be generated dynamically from cloud-optimized data. For compression and visualization of JPEG data, ZenJPEG improves the storage of NoData values, while brunsli can compress JPEG images up to 22% with no additional loss. A deep dive into the pros, cons, and performance evaluation of these compression algorithms will be performed.