Currently, television companies use upscaling services or tools to improve the image quality of non-high-definition content, allowing old videos from the analog television era to be broadcast in high resolution. In this work, we present a low-cost upscaling system that is scalable on demand and easy to implement on any cloud computing platform. We utilize deep learning network architectures, open-source codes and tools, resulting in a cost of less than 1 dollar per hour of processed video.