The rapid growth of web services is increasing the energy consumption and carbon footprint of Information and communications technology (ICT) infrastructure. While HTTP caching is widely deployed to improve performance, its quantitative energy and CO2 benefits remain insufficiently characterized. This paper presents a measurement-calibrated framework to estimate the energy use and associated carbon emissions of HTTP content delivery with and without caching. This study introduces a Projection-Calibrated Web Energy-Carbon Calculator (PC-WECC) that maps web workloads to energy and emissions by integrating data-center power usage effectiveness (PUE), network energy intensity, and country-level grid carbon intensity projections through 2030. To capture application-level effects, we develop an Empirically-Fitted Cache Efficiency Regression Model (ECERM) based on controlled server experiments that measure joules-per-gigabyte for cached and non-cached delivery. Results show that HTTP caching is 5−20% more energy efficient than non-cache delivery for typical web objects (< 5MB), with efficiency strongly dependent on file size and protocol overhead. When scaled to realistic traffic volumes, caching can yield substantial emissions reductions; for example, an average news website (∼ 300k daily visits) can avoid approximately 150tCO2 per year. This framework provides a transparent and reproducible pathway to translate HTTP architectural choices into energy and carbon impacts, supporting sustainable design and deployment of web infrastructure.