GCP – Google Cloud databases supercharge the AI developer experience

Generative AI continues to capture our imagination and promises to transform every industry. Its transformative potential hinges on integrating powerful models like Gemini, with the most contextually-relevant enterprise data.

Google Cloud is leading this transformation, not only by delivering the gen AI technologies themselves, but also by embedding those technologies directly into operational databases. For example, since introducing AlloyDB AI with vector search, we’ve expanded the capability across all our other database services — Bigtable, Cloud SQL, Firestore, Memorystore, and Spanner.

Today, operational databases are powering new AI agents and multimodal applications, and are enhancing existing applications with AI capabilities. That’s why we’re committed to delivering a truly innovative and unified data and AI foundation, helping customers redefine their industries, tackle massive data challenges, and drive groundbreaking innovation.

Today, at Google Cloud Next, we’re unveiling new database capabilities, including:

1. Generative AI capabilities in AlloyDB and agentic programming with MCP

2. MongoDB compatibility in Firestore

3. Expanded Oracle services, and a SQL Server modernization solution

New generative AI capabilities in AlloyDB and agentic programming with MCP

Agentic workflows are emerging as a primary architectural pattern for AI-based applications, and databases will be a primary component in this. The best agents will be the ones that can use real-time data to enable high quality, agentic decisions and actions — bolstering the quality of reasoning. To support this, we’re investing in AlloyDB AI capabilities to make it easier for developers to build intelligent agents and applications. 

First, we’re enabling Google Agentspace to search structured data in AlloyDB. Agentspace brings together Gemini’s advanced reasoning with Google-quality search and enterprise data. Now you can activate all your data stored in AlloyDB, enabling you to combine real-time, structured, and unstructured data in creative ways.

Second, we know that developers are looking to enable more flexible interfaces to their databases, while staying secure and accurate. That’s why last year we announced natural language support in AlloyDB to help developers build applications which accurately query data with natural language, just like they do with SQL.

We’ve been iterating on this, and today, we’re launching the next-generation of AlloyDB natural language. This technology lets you query structured data in AlloyDB securely and accurately, enabling natural language text modality in apps. 

AlloyDB AI natural language goes beyond interpreting database metadata, using supplied context and interactive intent clarification when querying the database. You can define which data can be accessed by natural language queries by using AlloyDB’s parameterized secure views, which provide an extra layer of security for agents and gen AI apps.

Third, building intelligent apps requires powerful vector search. We’ve seen AlloyDB’s vector search adoption increase nearly seven times since the launch of the state-of-the-art Scalable Nearest Neighbor (ScaNN) for AlloyDB index in 2024. Along with enhancements to ScaNN indexing in AlloyDB, we now offer optimized SQL functionality spanning vector search and structured filters and joins. 

With these innovations, AlloyDB’s ScaNN index offers up to 10 times faster filtered vector search queries compared to the hierarchical navigable small world (HNSW) index in standard PostgreSQL.

Fourth, in partnership with Vertex AI and Google DeepMind, we are introducing three new models in AlloyDB AI — one that improves the relevance of vector search results using cross attention reranking; a multimodal embeddings model that supports text, images, and videos’ and a new state-of-the-art Gemini Embedding text model. As a result you can easily add intelligence into your apps across multiple modalities such as images and text, with high efficiency and accuracy.

Fifth, to further infuse AI experiences into apps, we’re introducing AlloyDB AI query engine. AI query engine enables developers to freely use natural language expressions and constructs right within their SQL queries in a natural way. Developers can now use free text questions like “find family-friendly hotels in Orlando,” that require real-world data, including images and descriptions, and directly embed them in their SQL queries. 

At the heart of this are industry-leading foundation model-powered semantic operators that run side-by-side with traditional relational operators in the AlloyDB AI query engine.

Many of these AlloyDB capabilities are available in preview, and you can sign up to use them here today.

“At Target, we used AlloyDB to improve our online search experience. We used the ability to combine our structured and unstructured data to enhance the accuracy of natural language search queries by 20%!” Visagan Subburayalu, VP of Infrastructure & Cybersecurity, Target

Finally, we are announcing that MCP Toolbox for Databases (formerly Gen AI Toolbox for Databases) now supports Model Context Protocol (MCP). MCP enables seamless connections between AI agents and enterprise databases, replacing custom code that would otherwise be required. MCP Toolbox for Databases is an open source (Apache 2.0) server that supports multiple databases including PostgreSQL, MySQL, AlloyDB, Spanner, Cloud SQL (for PostgreSQL, MySQL, and SQL Server), Neo4j and Dgraph. It offers simplified development with reduced boilerplate code, enhanced security through OAuth2 and OIDC, and end-to-end observability with OpenTelemetry integration. We have open-sourced the code, so you are free to add whatever databases you choose.

Big step forward for document databases with MongoDB compatibility in Firestore

Developers love the agility of the popular MongoDB API and query language to store and query semi-structured JSON data, including the ability to use the open-source ecosystem of MongoDB integrations. Consequently, customers are looking for increased choice in how to build and deploy these workloads.

Today, we’re announcing the preview of Firestore with MongoDB compatibility. Built from the ground up by Google Cloud, it provides developers with additional choice for their demanding document database workloads. MongoDB API compatibility has been a highly-requested capability from Firestore’s existing community of over 600,000 monthly active developers. 

With this launch, Firestore developers can now take advantage of MongoDB’s API portability along with some of the Firestore capabilities they have come to depend on. These capabilities include multi-region replication with strong consistency, virtually unlimited scalability, industry-leading availability of up to 99.999% SLA, and single-digit milliseconds read latency performance, without having to worry about managing the underlying database infrastructure. 

In addition, developers can use their existing MongoDB application code, drivers, and integrations with the Firestore service. Firestore with MongoDB compatibility offers a customer-friendly serverless pricing model, with no up-front commitments required, and customers only pay for what they use. 

We’re committed to our partner ecosystem and will continue to support MongoDB Atlas in our Google Cloud Marketplace. This launch simply offers developers an additional choice for building their applications. You can get started here today.

“After migrating to Firestore, we improved developer productivity by 55%, observed better service reliability, and have been able to seamlessly scale to over 250,000 requests per second and 30 billion documents. Because Firestore is completely serverless and provides virtually unlimited scalability, we no longer have to worry about managing our underlying database infrastructure – liberating us from database DevOps. This has enabled us to focus on product innovations that matter to our customers,” said Karan Agarwal, director of engineering, HighLevel.

Expanded range of Oracle services, and a SQL Server modernization solution

Last year we announced Oracle Database@Google Cloud with availability in four global regions, enabling customers to migrate Oracle workloads, and modernize them with Google’s industry-leading data and AI capabilities such as BigQuery, Vertex AI, and Gemini foundation models. 

Today we’re announcing support for Oracle Base Database Service, offering a flexible and controllable way to run Oracle Databases in the cloud. We’re also announcing general availability of Oracle Exadata X11M, bringing the latest generation of the Oracle Exadata platform to Google Cloud and offering additional enterprise-ready capabilities, including customer managed encryption keys (CMEK). 

We’re continuing to invest in global infrastructure for Oracle, and these services are being deployed natively in 20 Google Cloud locations.

Building on this foundation, customers can now accelerate the development of cutting-edge agentic applications by integrating the power of the Google ecosystem with their business data residing in Oracle Database@Google Cloud. Learn more in this partner blog.

“Banco Actinver is committed to providing innovative financial solutions to our clients. By combining the security and performance of Oracle Database with Google Cloud’s data analytics and AI tools, we’re gaining deeper insights into market trends, enhancing our services, and delivering personalized experiences to our customers.” Jorge Fernandez, CIO, Banco Actinver

We know that many organizations are looking to reduce their dependence on SQL Server in order to eliminate its high costs and unfriendly licensing. Customers want automated ways to move their workloads to modern, AI-enabled databases.

Today, we’re announcing that Database Migration Service (DMS) now supports SQL Server to PostgreSQL migrations for Cloud SQL and AlloyDB. 

These new capabilities support the migration of both self-managed and cloud-managed SQL Server offerings, as well as a wide range of SQL Server editions and versions to allow you to fully execute on your database modernization strategy.

DMS offers online data migration and features a schema and code conversion engine that utilizes a combination of algorithmic and a specially-trained Gemini model, all designed to automate the most difficult migration steps, like converting Transact-SQL code and SQL Server-specific data types like DATETIME to their PostgreSQL equivalents.

But that’s not all, additional databases announcements at Next include:

1. Cloud SQL and AlloyDB are available on C4A instances. C4A instances are based on Google Axion Processors, our first Arm-based CPUs, custom-built for the cloud. These instances deliver nearly 50% better price-performance compared to N series machines, and up to two times better throughput than Amazon’s equivalent Graviton4-based offerings. Learn more here. 

2. Database Center, our AI-powered unified fleet management solution, is generally available and supports every database in our portfolio. This release delivers richer metrics and actionable recommendations, empowering users to optimize performance and reliability for their database fleet.

3. Spanner vector search is now generally available, designed to work with our SQL, Graph, Key-Value, and Full-Text Search modalities to address the most demanding AI workloads at virtually unlimited scale. 

4. Graph Visualization for Spanner, is now generally available. It allows users to visually explore valuable information from graph data. You can drill into specific nodes and relationships, filtering and highlighting relevant data subsets, and exploring the graph through intuitive navigation.

5. Aiven for AlloyDB Omni, is now generally available. This is a fully-managed AlloyDB Omni service, from our partner Aiven, running on AWS, Azure, and Google Cloud. 

6. Bigtable continuous materialized views, in preview, delivers an enhanced developer experience that simplifies real-time updates for modern applications that rely on immediate reporting and insights.

7. Memorystore for Valkey is now generally available, it supports 7.2 and 8.0 engine versions. Memorystore for Valkey 8.0 achieves up to 2x Queries Per Second (QPS) at microseconds latency when compared to Memorystore for Redis Cluster and provides optimized memory efficiency and better reliability for customers.

The future of data with Google Cloud

We’re excited to keep innovating on your behalf. At Google Cloud, we offer an intelligent, unified, and open data platform to support your generative AI journey every step of the way.

Learn more about Google Cloud databases and start a free trial for Cloud SQL, AlloyDB, and Spanner.