Google Cloud previewed Lakehouse cross-cloud caching, storing frequently accessed remote data locally to accelerate BigQuery queries and reduce transfers.

Google Cloud announced the preview of cross-cloud caching for its borderless Lakehouse. The service stores frequently accessed remote data locally in Google Cloud to accelerate repeated BigQuery queries and reduce cross-cloud transfers. Google also announced preview BigQuery cross-cloud connections for querying non-Iceberg data in other clouds.

The existing Lakehouse architecture uses the Apache Iceberg REST catalog specification to federate to catalogs including Databricks Unity Catalog, AWS Glue and Snowflake Horizon. Google’s article situates the update across data estates spanning Amazon S3, Azure Data Lake Storage, Google Cloud Storage, operational databases and software-as-a-service platforms.

Caching operates at sub-file block granularity for columnar formats such as Apache Parquet. BigQuery caches the particular column chunks and dictionary pages projected by a query, rather than transferring entire multi-gigabyte files. On a cache miss, it retrieves the required remote data and retains a local copy for subsequent queries.

Google says that combining standard Iceberg columnar compression with caching often reduces transferred data to under 5% of the data processed across clouds. That is a vendor-described workload outcome, not a universal transfer guarantee or a measured 95% reduction in total bills. Google also describes the feature as requiring no cache-storage management or tuning controls from users.