In an aggressive push to capture enterprise market share for artificial intelligence and big data analytics, Google Cloud has rolled out a suite of major enhancements to its borderless Lakehouse framework. The newly announced capabilities—highlighted by the public preview of intelligent cross-cloud caching and BigQuery cross-cloud connections—are designed to solve one of the most stubborn and expensive bottlenecks in modern enterprise IT: querying and moving massive, decentralized data estates spread across Amazon Web Services, Microsoft Azure, and diverse SaaS ecosystems.
For years, enterprises have struggled with the architectural friction of multi-cloud environments. As modern data stacks expanded to support resource-hungry AI agents, data scientists, and business intelligence teams, organizations found their critical information siloed across disparate platforms. Customer relationship management data lived in Salesforce, transactional logs resided in AWS S3 or Azure Data Lake Storage (ADLS), and operational telemetry was scattered across multiple cloud providers. Uniting these data sources traditionally forced engineering teams to rely on fragile, high-maintenance Extract, Transform, Load (ETL) pipelines, expensive data duplication, and punitive cross-cloud egress fees charged by public internet and cloud providers.
The Genesis of the Borderless Lakehouse
To dismantle these operational barriers, Google Cloud initially introduced the borderless Lakehouse concept earlier this year. The architecture was engineered to let organizations query and activate data directly where it lives without forcing them into costly data migration projects. By fully embracing the Apache Iceberg open table format and its REST catalog specifications, Google enabled seamless federation directly to competing and third-party catalogs, including Databricks Unity Catalog, AWS Glue, and Snowflake Horizon.
Simultaneously, Google introduced Partner Cross-Cloud Interconnect, establishing high-bandwidth, private network links between Google Cloud and other major cloud providers. This infrastructure significantly lowered per-gigabyte data transfer costs compared to routing traffic over the public internet. However, while federation and private interconnects solved connectivity and governance challenges, the volume of data traveling across network boundaries remained a significant financial hurdle for petabyte-scale enterprises running continuous AI workloads.
Addressing the Egress Cost Crisis
Recognizing that network transfer fees can quickly derail enterprise budgets, Google Cloud’s latest round of updates focuses on minimizing the sheer volume of data that must traverse the wire during query execution. The centerpiece of this update is the introduction of cross-cloud caching for Lakehouse, now available in preview.
The feature works transparently behind the scenes within BigQuery, intelligently caching frequently accessed remote data locally inside Google Cloud. By combining standard Apache Iceberg columnar compression with this new layer of cross-cloud caching, enterprises can dramatically reduce network strain. According to Google’s internal performance benchmarks, optimized queries often require transferring less than 5%—and in some scenarios, under 3%—of the total raw data processed across clouds. For an organization processing 1 terabyte of data, for instance, only about 26 gigabytes must cross the network boundary. This drastic reduction fundamentally alters the Total Cost of Ownership (TCO) equation, rendering cross-cloud analytics and large-scale AI model training economically viable.
Under the Hood: How Cross-Cloud Caching Operates
Designed to operate with zero administrative overhead, cross-cloud caching requires no manual tuning, custom indexing, or complex storage management from database administrators. It integrates natively with enterprise security and governance protocols.
When an analyst or an autonomous AI agent initiates a query against a remote dataset—such as a massive 10-terabyte sales table hosted in Amazon S3 and federated via Databricks Unity Catalog—BigQuery executes a series of optimized steps. During an initial "cold" run where the local cache is empty, the system applies partition pruning and column projection. Instead of pulling the entire table, it fetches only the exact Parquet byte ranges required to answer the query over the Partner Cross-Cloud Interconnect.
When a follow-on query is executed—a common scenario in exploratory data analysis where users or AI agents modify parameters, such as adding a shipping method or filtering by a new dimension—the engine intelligently leverages the previously cached columnar segments. This architecture prevents redundant network calls, accelerates response times from minutes to seconds, and compounds efficiency savings as query volumes scale.
BigQuery Cross-Cloud Connections Enter Preview
Complementing the caching engine is the simultaneous rollout of BigQuery cross-cloud connections, also available in preview. This feature allows organizations to link BigQuery directly to open-format data residing in Amazon S3 and Azure Storage without being strictly limited to Iceberg catalogs.
Industry analysts note a clear architectural distinction between catalog federation and cross-cloud connections. Catalog federation is ideal for organizations already invested in unified metadata layers like Databricks Unity Catalog or AWS Glue, allowing them to maintain external governance while executing queries in Google Cloud. In contrast, cross-cloud connections represent a modern architectural leap by utilizing standard BigQuery compute workers located within Google Cloud regions rather than spinning up compute resources in external clouds.
This compute consolidation unlocks several vital advantages for enterprise architects. First, it provides global region availability, ensuring consistent performance regardless of where the source data originates. Second, it delivers full BigQuery feature parity. By keeping the compute workload inside Google Cloud, organizations can seamlessly leverage advanced capabilities such as BigQuery AI and Gemini on remote files without running into regional or processing limitations imposed by third-party environments. Notably, the intelligent cross-cloud caching capabilities apply equally to data queried via cross-cloud connections and traditional Lakehouse catalog federation.
Industry Reactions and Strategic Implications
Enterprise technology experts and early adopters have responded positively to Google’s multi-cloud strategy, viewing it as a pragmatic acknowledgment of the multi-cloud reality that defines modern corporate IT infrastructure. Very few enterprises operate within a single cloud ecosystem; mergers, acquisitions, and multi-vendor strategies mean that data fragmentation is a permanent fixture of the corporate landscape.
"For years, vendors have tried to convince enterprises to centralize all their data into a single repository, but business reality dictates otherwise," notes a prominent enterprise data architect who has evaluated the new tooling. "Solutions that respect existing multi-cloud investments while removing the financial penalty of data movement are precisely what the market needs. By cutting remote transfer requirements down to a fraction of the total dataset, Google is making real-time cross-cloud AI agents economically feasible for the Fortune 500."
Market observers also point out the competitive pressure this places on other major cloud providers and data platform vendors. As artificial intelligence workloads demand faster, more unified access to enterprise data, solutions that bridge the artificial divides between AWS, Azure, and Google Cloud will increasingly dictate enterprise purchasing decisions. By removing friction from multi-cloud data access, Google is positioning its data cloud not as a walled garden, but as an orchestration layer capable of unifying the entire enterprise data estate.
Looking Ahead
As the preview phases for cross-cloud caching and BigQuery cross-cloud connections progress toward general availability, enterprise data teams are expected to begin restructuring their analytical workflows to take advantage of the lower TCO and enhanced performance. Organizations looking to implement these features can reference Google Cloud’s updated intelligent caching and cross-cloud connection setup documentation to configure secure, high-performance connections across their distributed data environments.
Ultimately, Google’s latest advancements signal a mature shift in cloud computing: moving away from forced data consolidation and toward intelligent, federated architectures where data remains where it is most useful, securely accessible, and cost-effective to analyze at any scale.
