Cloud ML Engine

The Cloud ML Engine offers two distinct subservices: distributed training and distributed prediction. Training refers to the process of building a machine learning model from a corpus of data, whereas prediction refers to the use of that model to actually make forecasts.

Depending on how long your ML training jobs run and whether you use batch or online mode for prediction, the costs vary significantly. A fairly heavy-duty set of ML workloads will cost about $4,000 per month: