Google Cloud Introduces Cross-Cloud Caching to Eliminate Data Migration Bottlenecks

Google Cloud announced the public preview of cross-cloud caching within its borderless Lakehouse architecture, alongside new BigQuery cross-cloud connections. The new capability allows BigQuery compute engines running in Google Cloud regions to query and process datasets stored natively across third-party environments—including Amazon S3 and Azure Data Lake Storage—while transparently caching frequently accessed sub-file Parquet blocks locally. By combining Apache Iceberg REST catalog federation (integrating with AWS Glue, Databricks Unity Catalog, and Snowflake Horizon) with local cache retention and columnar compression, the feature reduces actual cross-cloud data transfer volume to under 5% of processed query bytes. This announcement fundamentally alters the financial and operational calculus of enterprise cloud data migrations. Traditionally, moving analytics and AI workloads to a new cloud provider required either full, upfront dataset migration or continuous, brittle extract-transform-load (ETL) pipelines that duplicate petabytes of storage and incur massive egress charges. For data architects and cloud migration leads, cross-cloud caching offers a non-disruptive migration path: teams can immediately migrate analytical compute, agentic workflows, and query engines to Google Cloud without waiting for multi-month bulk data transfer projects to complete. The development aligns with the broader industry shift away from monolithic data consolidation toward open-table-format federation. As open table formats like Apache Iceberg become the universal storage standard across hyperscalers, data gravity is becoming decoupled from the query compute plane. Rather than forcing all storage assets into proprietary cloud silos, modern migration strategies increasingly leverage federated metadata catalogs, Dedicated Cross-Cloud Interconnects, and intelligent edge caches to treat multicloud estates as a unified data plane. In practice, infrastructure and data engineering teams should re-evaluate active data migration schedules. For organizations evaluating BigQuery or Gemini-driven agentic pipelines on top of data parked in AWS or Azure, cross-cloud caching allows immediate cutover of reporting and reasoning layers without migrating underlying object storage. Practitioners should benchmark their initial 'cold' cache read latencies against subsequent 'hot' cached queries and verify that their Apache Iceberg datasets utilize modern compression like Zstandard to maximize cache density and minimize network egress costs.