Amazon Elastic Inference will reduce deep learning costs by ~75%

AWS today announced Amazon Elastic Inference, a new service that lets customers attach GPU-powered inference acceleration to any Amazon EC2 instance.