The countdown to innovation has begun: Day 1 of Microsoft Build 2025 bought several new announcements, demos, and live coding sessions to in-person and online audiences, showcasing the potential of cloud-native and AI-driven applications. Azure Cosmos DB is at the forefront of this innovation. From powering multi-agent applications and unlocking high-speed analytics to delivering new features designed for faster and easier app development, we highlighted once again why Azure Cosmos DB is the go-to database for developers building at the edge of what’s possible. 

Here’s a look at what we announced today and Day 1 session highlights. 

A database built for the agentic web 

In his opening keynote, Satya Nadella outlined Microsoft’s vision for an open agentic web—a world where intelligent agents work side by side with developers and end users. Azure Cosmos DB is playing a critical role in that future. 

Satya announced the integration of Azure Cosmos DB directly into Azure AI Foundry, in GA, allowing any AI agent to store and retrieve conversational history and account data. Whether you’re building multi-agent apps using Semantic Kernel or LangChain , Azure Cosmos DB offers low-latency persistence, stateful memory, and global scale—exactly what modern, autonomous AI agents need to function.  

Satya also highlighted leading customers, such as NFL, Carvana, and Sitecore, who are innovating with Azure Cosmos DB and AI to transform digital experiences for customers. XM Cloud, the AI-forward CMS from Sitecore, is delivering more relevant content 10 times faster with Azure AI Foundry and Azure Cosmos DB at its core, according to Roger Connolly, CPO of Sitecore, “enabling the delivery of millions of real-time, personalized experiences like never before.” 

The NFL Combine used Azure Cosmos DB and AI to surface insights for draft decisions. The result? Coaches and scouts can analyze players in seconds using real-time, AI-powered insights—at scale. 

On-site attendees got to try their hands defining functions, managing memory, and persisting multi-agent interactions in the first session of our lab, “Multi-agent apps with Semantic Kernel or LangChain & Azure Cosmos DB.” Missed the lab? We’re offering a second instance on Thursday, May 22, 8:30-9:45 am PST – RSVP to get experience building scalable and fast multi-agent AI systems for real-world apps. 

Real AI workloads in production 

In What’s New in Microsoft Databases: Empowering AI-Driven App Dev, VP Azure Databases Shireesh Thota and VP Azure Cosmos DB Kirill Gavrylyuk kicked off the section on Azure Cosmos DB with a special customer highlight: Carvana achieved a reduction in inbound calls per sale with their AI agents powered by Azure Cosmos DB. In a new video, Michael Graf, Associate Director, Engineering, at Carvana, shared why he loves using Azure Cosmos DB. 

Shireesh and Kirill then highlighted several new Azure Cosmos DB capabilities for AI app dev at scale: 

Azure AI Foundry Connection for Azure Cosmos DB and BYO Thread Storage in Azure AI Agent Service 

You can now persist agent memory, system messages, and tool outputs directly in Azure Cosmos DB with full control and security. This integration powers the Azure AI Agent Service’s Bring-Your-Own Thread Storage feature. 

Why it matters: Azure Cosmos DB becomes the default memory and state store for building secure, scalable multi-agent systems. 

Full-Text, Hybrid, and Vector Search (GA and Preview) 

Search capabilities are now GA and extended with: 

• BM25 full-text scoring 

• Hybrid search using Reciprocal Rank Fusion (RRF) 

• Weighted search (e.g., more relevance from vector or keyword scoring) 

• LINQ and Entity Framework Core integration for .NET devs 

 Preview features: 

• Multi-language full-text search (French, German, Spanish) 

• Fuzzy search for typos and near matches 

• Filtered vector search with DiskANN (automatic plan optimization) 

• Document Indexer integration with Azure Logic Apps for parsing PDFs, images, markdown, etc. 

Why it matters: These features turn Azure Cosmos DB into a fully AI-ready document store for building RAG systems, conversational agents, and semantic search—all with massive scale and speed. 

Global Secondary Indexes (Preview) 

GSIs let you define alternate partition keys and indexing policies in new containers automatically synced from your primary container. 

Why it matters: You can optimize query performance and isolate workloads—without remodeling data. This is especially useful for high-volume apps and evolving microservices where access patterns change over time. 

Per Partition Automatic Failover (Preview) 

Traditional failover affected entire accounts. With Per Partition Automatic Failover (PPAF), only the affected partition fails over—automatically. It boosts availability and resilience without requiring complex logic in your app. Write latency is reduced during regional outages with P99 recovery time under 2 minutes. 

Why it matters: This is useful for mission-critical apps that require both strong consistency and fast recovery, without the complexity of active-active multi-write conflict resolution. 

Accelerating MongoDB Development on Azure (GA + Preview) 

 If you build MongoDB-compatible apps, Azure Cosmos DB for MongoDB (vCore) just got a major boost: 

 DiskANN vector indexing (GA) 

• Change Streams (GA) 

• DocumentDB Local (Preview) 

• Azure Functions triggers (Preview) 

• VS Code extension (GA) 

• Entra ID integration (GA) 

Why it matters: Developers now have modern open-source tooling, local development support, and enterprise-grade scalability—all fully compatible with MongoDB APIs. 

Fabric Integration: Cosmos DB in Fabric (Preview) 

Azure Cosmos DB is now a native database in Microsoft Fabric—automatically integrated into OneLake, real-time analytics, Copilot-powered Power BI, and more. 

Why it matters: It eliminates the complexity of building data pipelines between your app database and your analytics platform. Cosmos DB data is now queryable and visualizable alongside all your other sources in Fabric. 

Build scalable multi-tenant SaaS apps with confidence 

Azure Cosmos DB PM leads Andrew Liu and Deborah Chen wrapped-up the day in a nearly-packed session, Design Scalable Data Layers for Multi-Tenant Apps with Azure Cosmos DB. Andrew and Deborah unveiled Fleet Management with RU pooling and Global Secondary Indexes—a powerful duo of features for SaaS developers. 

With these tools, Azure Cosmos DB now makes it easier than ever to: 

• Create isolated, performant environments for each tenant 

• Optimize for query speed and cost 

• Visualize usage at scale with fleet analytics 

Andrew and Deborah were joined on stage by Liz Nelson, Sr. Product Lead at Sitecore, who shared how her team is already leveraging many of these capabilities to deliver super-fast experiences across Sitecore’s SaaS Content Management System (CMS) offering. Liz discussed her experience designing for multi-tenancy, leveraging Azure Cosmos DB global distribution, keeping costs in check, and using new features like full-text and semantic search to drive more relevant search results when retrieving information. 

Straight from our customers 

In addition to novel Azure Cosmos DB capabilities announced today, we debuted many new customer stories, highlighting real-world applications and tangible business impact. Check out our latest stories: 

HEINEKEN unites data and global teams with AI assistant fueled by Azure AI Foundry 

The ODP Corporation transforms HR, sales, and retail workflows with Azure AI app platform 

NFL Combine gains real-time insights for informed draft day decisions with Azure AI 

Carvana builds efficient, seamless customer experience with agent on Azure 

From idea to orbit: Terra Mater Studios builds Hera AI companion with Azure data, AI 

Assembly Software saves law firms up to 25 hours per case with Azure 

Still ahead this week 

Wednesday, May 21st   

🗓️ 9:00am PT: Scale and secure MongoDB-compatible apps with Azure Cosmos DB (BRK202) with Khelan Modi and Patty Chow from the Azure Cosmos DB team, and Hari Krishna Sunder, Principal Engineer, and Rohan Thomas, VP of Partner Sales, from Yugabyte. 

🗓️ 10:30am PT: Beyond the Game: How AI and Data Are Powering the Future of Sports (BRK143) with Indiana Pacers’ Manager of Data Engineering Jared Chavez, CTO of Southworks Johnny Halife, James Codella from Azure Cosmos DB, and Andy Beatman from Azure AI. 

🗓️ 10:30am PT: Building secure AI Agents with Azure Functions (BRK189) with Pim Zandee Global Business Intelligence Lead of Heineken. 

💻 2:20pm PT: Tips for fast vector and full-text search with Azure Cosmos DB (DEM567) with James Codella and Haiyang Xu. 

🗓️ 3:30pm PT: Architecting highly resilient applications in Azure: A technical guide for developers (BRK197) with Kirill Gavrylyuk, VP of Product, and Justine Cocchi, Sr. Program Manager, from the Azure Cosmos DB team. 

🗓️ 3:30pm PT: Earth’s Defense with Hera: AI Agents Battle Planet Extinction Threats (BRK186) with ImpactAI, who will share how they used Azure Container Apps, Azure Cosmos DB, and Azure OpenAI Service for space exploration. 

Thursday, May 22nd   

💻 8:30am PT: Don’t miss the second instance of our lab, Multi-agent apps with Semantic Kernel or LangChain & Azure Cosmos DB (LAB361 with Mark Brown and James Codella (live only)  

Our team will also be in the Expert Meet-Up area onsite, booth 810, to share more about what’s new in Azure Cosmos DB, demonstrate some of the latest capabilities, and hand out swag. Come on by!