In the agentic era, the role of enterprise data is shifting from a passive resource to a dynamic System of Action. That’s why we announced the Agentic Data Cloud — a unified architecture that integrates models, analytics, and operational databases into a single, AI-native system. This allows organizations to unlock the full potential of their data, ensuring every AI application and agent is grounded in truth and is capable of taking real-time action without costs spiraling.
Today, we’re moving the needle by:
Embedding AI across the entire data stack for an unmatched, optimized developer experience.
Unifying transactional and analytical worlds, to eliminate the friction of manual integration.
Simplifying enterprise-grade deployments with industry-leading, open databases that are easier to manage and faster to deploy.
Embed AI into every layer of the data stack
We continue to enable smooth connections between AI agents and enterprise databases, making your database experience effortless and agent-driven. From vibe coding tools for citizen developers, to high-scale vector search for the enterprise, we know that databases are one of the primary components of agentic workflows.
Key launches include:
Vibe coding integrations for databases with AI Studio: To enable creators, we are launching an integration with our leading vibe coding platform Google AI Studio. This agent-led, automated workflow allows you to create a live application from a simple text prompt in seconds, that connects the application to trusted database services like Firestore, with support for Cloud SQL for PostgreSQL coming soon.
Tools for Data Agents (Preview): These are the modular building blocks developers use to give their custom AI agents direct, secure access to database capabilities. Available for AlloyDB, Cloud SQL, and Spanner, these tools provide ready-to-use functions — like the QueryData tool for high-accuracy text-to-SQL— that allow any agent to query, understand, and interact with your data reliably. These tools for Data Agents in AlloyDB, Spanner, and Cloud SQL provide near-100% text-to-SQL accuracy.
Database onboarding and observability agents (Preview): The new Database Onboarding agent takes the guesswork out of database selection and deployment. By evaluating your stated requirements — from simple use case descriptions, to complex enterprise needs — it recommends the ideal Google Cloud database and guides you through provisioning. Additionally, the new AI-powered Database Observability Agent proactively monitors performance and health across AlloyDB, Bigtable, Cloud SQL, and Spanner. It identifies root causes of potential issues and delivers accurate remediations for troubleshooting.
AlloyDB AI-powered search at scale (Preview): We are introducing unprecedented scale and performance for enterprise search. AlloyDB can now scale to 10 billion vectors using Google’s ScaNN index, and offers up to 6 times faster vector queries than the HNSW index in standard PostgreSQL. Additionally, our HNSW index, accelerated by the Columnar Engine, now outperforms standard PostgreSQL by 4 times. We’re empowering developers to build the ultimate hybrid search engine by pairing high-speed vector retrieval with industry-standard full-text search, optimized by native BM25 support coming soon.
AlloyDB vector search innovations
New and optimized AI functions for AlloyDB (Preview): AlloyDB AI functions allow you to integrate enterprise data with the power of LLMs like Gemini. For example, the ai.if function can replace custom logic to determine if user transactions are fraudulent. Joining our existing suite, ai.if, ai.rank, ai.generate, and ai.forecast, are two new additions: ai.analyze_sentiment and ai.summarize. We also now have an optimized mode for ai.if, which delivers superior cost and performance boosts.
Managed, remote MCP servers for databases (GA): We’re announcing managed, remote Model Context Protocol (MCP) servers for AlloyDB, Bigtable, Cloud SQL, Firestore, and Spanner, alongside a preview for Memorystore, Database Migration Service, Datastream, Database Center, and Oracle Database@Google Cloud to achieve MCP coverage across our portfolio. By fully managing the infrastructure required to connect AI models securely to your data, we are eliminating the operational burden of hosting, securing, and scaling MCP servers yourself. This ensures your AI models can reason and act upon your most up-to-date enterprise data with production-grade reliability.
Open-source MCP Toolbox for Databases (1.0): Supporting over 40 databases with contributions from 10 vendors, the open source, industry-standard MCP Toolbox for Databases has achieved a major stability milestone with this release. By guaranteeing the API will not break without a major version bump, we are giving you the confidence to build production applications, delivering a highly reliable foundation for both developers and autonomous AI agents.
How UKG uses AlloyDB to scale its People Fabric platform
Break down walled gardens with lakehouse integrations
To build real-time AI applications, developers need to combine live transactional context with massive historical insights at sub-millisecond latencies. Today, 98% of our largest data cloud customers are running their operational and analytical workloads on Google’s Agentic Data Cloud.
We’re announcing:
Lakehouse federation for AlloyDB (Preview): AlloyDB customers can now access live data from Iceberg and BigQuery from the PostgreSQL data plane. Users can discover and search for any BigQuery or Lakehouse for Apache Iceberg table directly from the AlloyDB Studio UI and begin querying immediately, with pushdown to BigQuery for filters and aggregations. This allows live joins between AlloyDB’s transactional data and historical insights in BigQuery or Iceberg without any data movement.
Reverse ETL for BigQuery (Preview): Many customers use BigQuery and Lakehouse to combine structured warehouse data with rich context from the data lake using tools like Apache Spark or BigQuery ML. Now, with our new one-click Reverse ETL, customers can activate these insights and sync data from the lakehouse to AlloyDB with a single click. AlloyDB acts as a high-concurrency, low-latency serving layer, optimized to meet the needs of real-time and agentic applications. Excellent serving performance is powered by AlloyDB’s unique columnar engine and ultra-fast cache, available at no additional cost.
Seamless ongoing replication with Datastream (GA): To close the data loop, Datastream now allows users to replicate data continuously to BigQuery and, importantly, to Iceberg tables directly from AlloyDB. This ensures operational changes are quickly reflected in the analytical environment, crucial for real-time ML feature engineering. Datastream is simple to set up, fully serverless, and offers a free tier for AlloyDB to BigQuery streams.
Knowledge Catalog, formerly Dataplex (Preview): The Knowledge Catalog is a universal context engine which maps and infers business meaning across your entire data estate, using a rigorous framework of aggregation, continuous enrichment, and search. It works by aggregating native context across your Google and partner data platforms, semantic models, and third-party catalogs, unifying them into a single, governed source of truth.Governance is consistently applied, access controls are managed centrally, and full visibility and security are maintained across the entire data lifecycle.
Spanner Columnar Engine (GA): This capability accelerates analytical queries by speeding up scans up to 200 times on live operational data. By storing data in a columnar format alongside traditional row-based storage, Spanner can execute complex queries automatically using vectorized execution — processing batches of data at once rather than row-by-row. Spanner also now supports Iceberg tables, continuous reverse ETL from BigQuery, and accelerated federated queries — all using the Spanner columnar engine.
Database Center with BigQuery (Preview): Database Center now supports BigQuery alongside operational databases and self-managed databases on Google Compute Engine. You can now monitor your entire data estate — from the databases running your critical applications to the warehouses analyzing your business — from a single, intelligent management plane. Additionally, Gemini-powered fleet analytics now proactively surfaces performance optimization opportunities across your environment. And with our new API and managed MCP support, you gain the freedom to stream these rich fleet metrics directly into your preferred third-party tools and custom dashboards.
“Our partnership with Google Cloud is focused on building a smarter, faster exchange. Google’s Agentic Data Cloud allows us to dismantle the legacy silos and technical debt that once slowed us down. By integrating the operational reliability of Cloud SQL with the deep reasoning of BigQuery, we’ve created a data ecosystem where our developers and AI agents can validate, optimize, and innovate in real time.” —Kristofer Shane Sikora, Executive Director, Cloud Data Engineering, CME Group
Commitment to open data and multi-cloud flexibility
We recognize how critical databases are for the day-to-day operations of all your applications. We continue to innovate on industry-leading reliability, price/performance, and scale, while maintaining a focus on open source and open data formats like Iceberg, MySQL, PostgreSQL, and Valkey.
To simplify enterprise-grade deployments, we’re announcing:
Spanner Omni (Preview): A new downloadable edition of Spanner that extends the industry’s leading distributed database beyond Google Cloud. Spanner Omni empowers organizations to use Spanner’s unparalleled scalability, high availability, strong consistency, enterprise-grade security and fully interoperable, multi-model capabilities for AI-enabled applications in their own data centers, across clouds or even at the edge.
What makes Spanner Omni unique
Bigtable in-memory (Preview): Bigtable is evolving its architecture and now delivers sub-millisecond read latency with the new memory tier. Available as part of Bigtable’s new Enterprise Plus edition, the new nodes with hybrid storage architecture spanning RAM, SSD and HDD, integrate frequently accessed data in-memory with long-term archival storage — all within a single, fully-managed service. Complete this form to express interest.
Memorystore for Valkey 9.0 (GA): Demonstrating our strong commitment to open source, this milestone anchors our focus on simplifying adoption and delivering enterprise-grade performance. We are introducing a managed migration path from self-managed Redis and Valkey to Memorystore, making it easier than ever to modernize. We’re also adding new smaller and larger node sizes for better price-performance, support for the Valkey-bloom and Valkey-json modules, and enterprise-grade security featuring Access Control Lists (ACLs), token-based authentication, and Flexible Certificate Authorities.
Oracle Database@Google Cloud: We are expanding to 20 regions, and are adding Oracle GoldenGate Service support, near real-time data replication to BigQuery, and integration with Knowledge Catalog (formerly Dataplex) and Database Center.
Compute Engine-to-managed migrations (Preview): This is a new experience designed to eliminate the complexity of moving databases. Natively integrated into Cloud SQL, AlloyDB, and Database Center, it provides automated networking and replication for PostgreSQL, allowing you to move your Compute Engine workloads to our powerful managed services with minimal effort and downtime.
Firestore full-text and geospatial search (Preview): Google’s industry-leading search technology is now natively integrated into Firestore, creating a differentiated, unified serverless database and search offering. This offering allows developers to execute high-relevance keyword, phrase, and geospatial queries, while ensuring that search results remain consistent with the underlying database data to significantly reduce operational overhead.
Mercado Libre: “Mercado Libre built Fury, our in-house developer gateway. Through Fury, we offer a NewSQL service powered by Spanner, leveraging its unmatched scalability, resilience, and consistency. While we are a cloud-native company, we have long remained vigilant regarding risks like insider threats, ransomware, and cloud outages. Spanner Omni enables true cross-cloud resilience, providing a robust and differentiated strategy that would be significantly more complex to implement with other cloud providers.” – Diego Oscar Narducci, Sr Technical Manager, Mercado Libre
Get started today
To learn more about our full portfolio please visit https://cloud.google.com/products/databases. To get started with migration and modernization initiatives, get started with our RAMP offer.