Google Cloud Expands Borderless Lakehouse with Direct Cross-Cloud Apache Iceberg Access

Google Cloud has rolled out a direct cross-cloud data access capability within its borderless Lakehouse platform, providing a modern alternative to legacy BigQuery Omni architectures. The feature enables BigQuery and managed open-source compute engines, such as Apache Spark, to query datasets residing in external object storage like Amazon S3 without staging or migrating underlying files. Authentication is handled via Secret Manager credentials or OIDC token federation against remote Apache Iceberg REST catalogs—including Databricks Unity Catalog, Snowflake Horizon, and AWS Glue—with traffic routed over either the public internet or private Cross-Cloud Interconnect links. This architectural evolution fundamentally alters multi-cloud data operations for platform architects and FinOps engineers. Historically, organizations operating across AWS and Google Cloud faced an unappealing trade-off: duplicate vast petabytes into redundant cloud storage—incurring massive egress fees and pipeline maintenance—or run proprietary remote compute runtimes that lacked deep optimization. By establishing metadata federation alongside direct transport protocols, data teams can query remote Iceberg tables in place with native query engines, slashing pipeline maintenance overhead and making distributed lakehouses viable without recurring ETL costs. The shift reflects a broader convergence toward open table formats, particularly Apache Iceberg, as the universal abstraction layer across multi-cloud environments. Rather than competing strictly to lock storage within proprietary formats, major hyperscalers and data platforms are increasingly standardizing on the Iceberg REST catalog specification. This pattern mirrors earlier infrastructure developments where standard APIs neutralized proprietary control planes. In modern enterprise IT, data gravity is giving way to federated interoperability, enabling specialized analytical and AI workloads to execute seamlessly across multi-provider footprints. Practitioners should evaluate this capability carefully against latency and egress budgets. Querying remote S3 data directly from Google Cloud over public endpoints will still trigger standard cloud egress charges unless routed over Dedicated or Partner Cross-Cloud Interconnects, which provide discounted transfer rates and deterministic network performance. Engineering teams should audit their table layout and partitioning schemes: highly pruned analytical queries will see optimal performance, while unindexed full-table scans across cloud boundaries remain cost-inefficient. Moving forward, teams should standardize their schemas on open Iceberg catalogs to ensure future-proof portability across multi-cloud analytical stacks.