This research suggests an actor-critic approach for cloud computing resource allocation that is based on deep reinforcement learning. The suggested approach generates the allocation strategy using an actor network and assesses the allocation's quality using a critic network. A deep reinforcement learning technique is used to train the actor and critic networks in order to maximize the allocation strategy. A simulation-based experimental investiGenetic Algorithm is used to assess the suggested approach, and the findings demonstrate that it performs better overall in cost, efficient energy and resource utilization than a number of the current allocation approaches. In an attempt to lower energy usage, some methods for workloads management or virtual machines have been already, but, those methods frequently are unable to account for the states of the server. In order to achieve logical task allocation, this study suggests the Dynamic Workload Allocation Scheme (DWAS), which continuously captures the dynamicity of server conditions and accounts for the impact of various workloads on energy consumption. The goal is to guarantee workload quality of service (QoS) with less computational of physical servers. This technique computes the predicted cumulative return over time using a dynamic algorithm based on reinforcement learning (RL) to discover the optimal task allocation in terms of energy efficiency. Through simulation, we find that, in comparison with the current baseline allocation methods, the proposed DWAS can lower the workload around 15%. The report also discusses potential applications of the suggested technology in cloud computing and makes recommendations for further research.