The Borderless Lakehouse: Extending Google AI to AWS and Azure
The corporate tech sector in 2026 has evolved beyond experimenting with generative models and into the large-scale operationalization of agentic systems. This period is defined by the merging of vast computational power with self-governing data frameworks that oversee the entire data lifecycle, transforming everything from initial ingestion to active business coordination.
Rather than merely producing insights, the fundamental challenge for today’s organizations is developing the “connective tissue” necessary for AI agents to function within disparate, multi-cloud settings.*
To address this, executives are focused on developing an adaptable, organization-wide “living AI backbone” that operates in real-time. By leveraging Google’s cross-cloud, AI-native advancements, businesses can successfully navigate this shift and dissolve the conventional silos between cloud providers.
Dismantling the Walled Garden: The Borderless Lakehouse Vision
The concept of the “walled garden” has long been a point of friction in multi-cloud strategies. While many providers promise “open” systems, these promises often extend only to the boundaries of their proprietary walls. When an enterprise attempts to run an engine on-prem or use a different cloud’s ML platform, governance often vanishes, security breaks, and the cost of data movement spirals.
Modernizing Infrastructure for Zero-Copy Inference
Legacy architectures were designed for human scales, where a few analysts might run ten queries a day, rather than for agent scales, where autonomous systems might run ten thousand queries a minute. In this high-velocity environment, the latency of moving data to a model compromises the user experience and the agent’s effectiveness.
The borderless lakehouse addresses this by enabling “zero-copy inference,” allowing AI to reason over data exactly where it resides, such as on AWS S3 or Azure Blob Storage. The Google Cloud Agentic Data Cloud is engineered specifically to close this gap by integrating connectivity, context, and computation into a vertically integrated stack. This approach allows organizations to move from reactive intelligence to proactive action by providing agents with the universal context they require to execute complex tasks with high accuracy and low hallucination rates.
Strategic Challenges of Multi-Cloud Data Silos
Fragmented visibility remains the top security and operational concern for enterprises. Over 80% of enterprises now run workloads across two or more cloud providers, yet only 8% report having achieved a single, unified view of their data.*
| Multi-Cloud Challenge | Technical Root Cause | Business Impact |
|---|---|---|
| Fragmented Visibility | Siloed monitoring tools; no unified inventory. | Uncontrolled costs; shadow IT; security blind spots. |
| Inconsistent Policy | Manual replication of IAM across cloud consoles. | High risk of human error; privilege escalation. |
| Data Gravity | High egress fees; unsustainable transfer costs. | Stalled innovation; rigid tech roadmaps. |
| Trust Gap | Lack of universal context for AI models. | Hallucinations; unreliable autonomous agents. |
To overcome these barriers, technical decision-makers are shifting toward “modular, cloud-native platforms” that securely connect and govern all data types. The “borderless lakehouse” is the architectural manifestation of this shift, providing the essential “connective tissue” that enables autonomous agents to operate across the enterprise without hindrance.*
Innovations in Cross-Cloud Interconnects and NetworkingAt the heart of the “borderless lakehouse” are major innovations in connectivity that provide the throughput and latency required for agentic reasoning. Google’s Cross-Cloud Network acts as a unified foundation to connect users, data, and AI services across environments at a “planetary scale”.* The Virgo Network and AI-Native Cloud InterconnectOne of the most significant announcements from Google Cloud Next 26 is the Virgo Network, a breakthrough scale-out fabric for AI workloads. It is designed to handle the massive machine-to-machine traffic generated by autonomous “reasoning loops,” where agents call other agents and multiple LLMs in rapid succession.* Technical capabilities of the Virgo Network include:
To complement this fabric, Google introduced AI-native Cloud Interconnect, which supports 400 Gbps circuits and provides fixed-price options for petabit-scale data transfers.* This innovation is particularly relevant for businesses looking to standardize their multi-cloud infrastructure while maintaining a predictable budget. Partner Cross-Cloud Interconnect for AWS and AzureGoogle Cloud has expanded its native connectivity to other hyperscalers, making Partner Cross-Cloud Interconnect for AWS generally available.* This allows for simplified, high-bandwidth (up to 100 Gbps) and encrypted-by-default connectivity between Google Cloud and AWS, eliminating the need for complex third-party intermediate hardware. A similar experience for Azure is currently in preview, further extending the reach of the borderless lakehouse. Integration with the Network Connectivity Center (NCC) allows these cross-cloud links to be managed as a first-class connectivity type, enabling secure, high-performance data paths that the Agentic Data Cloud can leverage for zero-copy queries. |
Eliminating the Egress Tax with Cross-Cloud Caching
For many businesses, the cost of moving data between cloud providers is the primary barrier to multi-cloud analytics. Traditional methods require ongoing data extraction and loading (ETL) processes that create a growing technical burden and incur massive egress fees.*
The Intelligent Cross-Cloud Cache
The 2026 lakehouse architecture introduces cross-cloud caching (Preview Q2 2026), a capability that integrates directly into the data plane. This intelligent cache stores cross-cloud data on the first read, allowing follow-on queries to run at cloud-native speeds without re-fetching data from the source cloud.
Benefits of cross-cloud caching include:*
- Slash Egress Fees: By reducing the amount of data that must be pulled from AWS or Azure for repetitive queries, organizations can significantly lower their monthly cloud networking bills.
- Accelerate Performance: Data residing in the cache is served with local throughput, delivering optimized price-performance characteristics similar to cloud-native solutions.
- Real-Time Interoperability: Agents can use data across AWS and Azure as if it were sitting locally in Google Cloud, enabling real-time analytics on “fresher” data than traditional ETL would allow.
This caching technology, paired with programs like Zero Egress Migration (0EM), effectively neutralizes the economic barriers to multi-cloud AI infrastructure. This means the freedom to place workloads in the most cost-effective region without being penalized for your data’s “origin” cloud.
Bi-Directional REST Catalog Federation
The borderless lakehouse further eliminates proprietary silos through lakehouse catalog federation. Powered by the Apache Iceberg REST Catalog, this feature enables bi-directional federation, allowing Google Cloud engines to read directly from:
- Databricks Unity Catalog on Amazon S3
- Snowflake Polaris
- AWS Glue Data Catalog
This bi-directional synchronization with Google’s new Knowledge Catalog ensures that governance policies and access controls apply instantly across the entire borderless environment.* By automating data reliability, it eliminates the manual effort once required to manage fragmented systems—a persistent obstacle for organizations prior to their cloud journey.
The Universal Context Engine: Powering the System of Action
Data without context is the leading cause of AI hallucinations and costly errors in the agentic era. McKinsey’s State of AI trust in 2026 survey indicates that nearly two-thirds of organizations cite security and risk concerns—specifically inaccuracy—as the top barrier to scaling agentic AI.
Knowledge Catalog: Always-On Universal Context
Google’s Knowledge Catalog serves as a universal context engine, providing the foundation for agents to execute tasks accurately. It automates metadata harvesting, data profiling, and lineage tracking, allowing agents to instantly find and use the specific data they need, whether it is stored in structured tables or hidden in unstructured formats like PDFs and spreadsheets.
The Knowledge Catalog also powers the Deep Research Agent, which can autonomously conduct multi-layered investigations across internal documents, analytical platforms (BigQuery), and the web. This represents a shift from “human scale” research—which could take weeks—to “agent scale” research that delivers precise answers with citations in minutes.
Model Context Protocol (MCP) and Agent Gateway
To ensure these agents can interact seamlessly with various tools and runtimes, Google has fully embraced the Model Context Protocol (MCP). MCP provides a secure, universal interface that allows any agent to discover and use data assets across BigQuery, Spanner, and AlloyDB.
The Agent Gateway acts as the “air traffic control” for this ecosystem, enforcing consistent security policies and implementing Model Armor protections against risks like prompt injection and data leakage. This centralized governance is critical for adhering to the “sovereignty-by-default” principles that are becoming standard.
High-Performance Foundations: TPU 8 and BeyondThe Agentic Data Cloud is powered by Google’s eighth-generation Tensor Processing Units (TPUs), which have been specifically reengineered to meet the demands of 2026. TPU 8t vs. TPU 8i: Specialization for ScaleFor the first time, Google has split its TPU architecture to optimize for different phases of the AI lifecycle:
This hardware acceleration, combined with Native PyTorch support for TPU (TorchTPU), allows EMEA organizations to run their preferred models with maximum efficiency. Spanner Omni and the Borderless DatabaseSpanner Omni (Preview) brings Google’s globally consistent, multi-model database to any environment. For the first time, enterprises can run the Spanner engine across multiple clouds, on-prem, or even locally, ensuring a consistent database experience regardless of infrastructure. This “database without boundaries” is critical for organizations that must manage data across a sprawling global footprint. |
Economic Efficiency: ROI and FinOps in 2026
The maturity of AI in 2026 is also defined by a shift from “exploratory” spending to disciplined financial gains. PwC 2026 Global CEO Survey shows that companies applying AI widely to products and services achieved nearly 4% higher profit margins than those that did not.
BigQuery Fluid Scaling
One of the most directly actionable cost reduction opportunities announced in 2026 is BigQuery Fluid Scaling.
- Efficiency: It reduces costs by up to 34% on average for autoscaling workloads by dynamically right-sizing compute allocation to match query demand.
- Operational Simplicity: For teams already running BigQuery with autoscaling, these savings apply with no code or configuration changes.
- Agentic Impact: As agents act on data, fluid scaling ensures that resources scale up instantly during action and scale back down when idle, preventing the “cost spiral” typically associated with high-frequency autonomous queries.
Lightning Engine for Apache Spark
For data science teams, the Lightning Engine for Apache Spark delivers up to 2x the price-performance over proprietary market alternatives. It provides industry-leading performance on Iceberg, Parquet, and Delta formats without requiring code changes, allowing organizations to run their existing Spark workloads faster and more cost-effectively.
Lead the Shift: Build Your Agentic Foundation
The borderless lakehouse represents a fundamental shift in how organizations perceive and manage their most valuable asset. By eliminating the “walled garden” approach, Google Cloud is enabling a future where data is liberated, agents are empowered with universal context, and multi-cloud complexity is replaced by a unified “system of action.”
To succeed in 2026, organizations must:
- Break Down Silos: Adopting the Iceberg REST Catalog for bi-directional federation across AWS and Azure
- Build Trust Through Context: Leveraging the Knowledge Catalog to ground AI agents in accurate, governed business meaning
- Optimize for Outcome: Utilizing BigQuery Fluid Scaling and Cross-Cloud Caching to achieve strategic differentiation with a lasting competitive edge
The era of simple prompts is over; we are now in the age of the “agent leap”. Businesses that transition from isolated initiatives to comprehensive technical and structural overhauls are set to gain significant advantages in the coming years.
Accelerate Your Transformation with Kartaca
Navigating the complexities of a borderless multi-cloud strategy requires more than just technology. It requires an expert partner who understands the regulatory landscape, the physics of high-performance networking, and the nuances of the Agentic Data Cloud.
As a Premier Google Cloud and Google Workspace Partner, Kartaca is uniquely positioned to help your organization dismantle its data silos and build a scalable, AI-ready foundation.
Contact us today to explore how the borderless lakehouse can transform your enterprise data into a dynamic reasoning engine, empowering your business to lead the shift into the agentic era.
Author: Gizem Terzi Türkoğlu
Published on: Jun 16, 2026