The seismic migration is one of the seismic processing stages employed by the oil and gas industry to generate subsurface images. This process is responsible for relocating the recorded seismic events to its correct spatial position and collapse the diffractions to their scattering points. Reverse-Time Migration (RTM) is one of the most common methods because it generates subsurface images with high quality in scenarios with complex structures. However, the method implies a high computational cost because it uses the solution of the wave equation to find the source wavefield, increasing the runtime. In this work, we propose a Reverse-Time Migration implementation using a GPU cluster, by taking advantage of the independence of the seismic data, more specifically the shots gathers acquired in the field. The proposed strategy consists in split the data to be processed in different GPUs using Message Passing Interface (MPI). Then, each GPU independently applies the RTM method to its corresponding portion of the total data. The final migrated image is generated by adding the results obtained from each GPU. The advantage of this strategy is its scalability, because the performance can be improved by adding more hardware (GPUs). We tested the method by using the synthetic Marmousi II model, obtaining a speed up factor of 5:61 when 3 nodes are used in comparison to the implementation in a single GPU.