<p>Support Vector Machine (SVM) is widely used for classification tasks in machine learning, but it struggles with high-dimensional data, scalability, and computational inefficiencies, particularly in cloud environments with large-scale virtualized workloads. Quantum computing provides a solution by enhancing computational efficiency and offering richer feature representation through quantum kernels. They uses high-dimensional Hilbert spaces, enabling more effective pattern capture and reduced training overhead compared to classical kernels. This paper introduces QKiSVM, a novel Quantum Kernel-infused Support Vector Machine algorithm that incorporates quantum principles to improve classical SVMs. The framework integrates Quantum Amplitude Encoding, Quantum Kernel Functions, Quantum Inner Product estimation, and optimizes the model using Quantum Gradient Descent and Quantum-Assisted Hyperparameter Tuning. Theoretical foundation of QKiSVM is established by proving two theorems that demonstrate its optimality in terms of convergence and classification performance. We implement QKiSVM for the classification of anomalous Virtual Machines (VMs) in public cloud environments. For validation, we conduct a comprehensive performance evaluation, comparing QKiSVM with PSO-SVM and QSVM using key metrics such as accuracy, precision, recall, F1-score, training time, resource utilization, computational cost, and hyperparameter tuning time. Experimental results show that QKiSVM consistently outperforms existing approaches, demonstrating superior accuracy and efficiency in both classification and resource optimization.</p>