In this study, an exploration is made on applying agent-based models and deep reinforcement learning to cloud resource optimization. Experimental environments are backed by an agent-based cloud simulator developed using the modeling language Netlogo. Several existing service placement algorithms are utilized to formulate an alternating approach to explore further optimization opportunities. A deep reinforcement learning model is built to study energy efficient service placement strategies. More specifically, deep Q-learning is employed for handling service placement. In the evaluation, the performance of the alternating approach and the deep reinforcement learning model is comparatively evaluated against existing algorithms.