Moving petabytes of data out of your data warehouse to train a model is slow, expensive, and insecure. BigQuery ML (BQML) solves this by allowing you to train ML models directly inside BigQuery using standard SQL.
CREATE MODEL `my_dataset.churn_model`
OPTIONS(model_type='logistic_reg') AS
SELECT * FROM `my_dataset.customer_data`;
You can train linear regression, k-means clustering, and even deep neural networks without ever leaving the database.