<p>Machine learning (ML), as a data-hungry technology, is transforming a broad spectrum of industries, including the Internet of Things, by providing sophisticated and robust data-driven solutions to a wide range of complex problems. However, the training of ML algorithms is usually computationally intensive, requiring significant energy resources and contributing to environmental concerns such as increased carbon emissions and depletion of natural resources. Despite the extensive focus of research in artificial intelligence (AI) on enhancing algorithmic performance, the energy consumption of ML models remains underexplored. This paper investigates this issue by estimating the power consumption and energy footprint of two widely used machine learning algorithms: support vector machine (SVM) and random forest (RF). The study focuses on the training phase and presents a methodology for quantifying their energy consumption. We conduct experiments on three heterogeneous machines using the MNIST dataset, comparing SVM and RF in terms of training time, central processing unit (CPU) utilization, power consumption, energy consumption, and performance metrics, including accuracy, precision, recall, and F1-score. We provide insights into the energy-performance trade-offs and present the relationship between CPU utilization, power consumption, and energy usage. Our findings reveal that SVM consumes significantly more energy (up to 40 kJ) than RF (9 kJ), despite SVM demonstrating marginally higher accuracy (97.65 vs. 97.11%). The conducted analysis and its findings contribute to the advancement of sustainable AI practices, guiding researchers, engineers, and policymakers toward more eco-conscious AI solutions.</p>