The countdown to innovation has begun: Day 1 of Microsoft Construct 2025 purchased a number of new bulletins, demos, and dwell coding periods to in-person and on-line audiences, showcasing the potential of cloud-native and AI-driven purposes. Azure Cosmos BD is on the forefront of this innovation. From powering multi-agent purposes and unlocking high-speed analytics to delivering new options designed for sooner and simpler app growth, we highlighted as soon as once more why Azure Cosmos DB is the go-to database for builders constructing on the fringe of what’s doable.
Right here’s a take a look at what we introduced in the present day and Day 1 session highlights.
A database constructed for the agentic internet
In his opening keynote, Satya Nadella outlined Microsoft’s imaginative and prescient for an open agentic internet—a world the place clever brokers work facet by facet with builders and finish customers. Azure Cosmos DB is enjoying a crucial function in that future.
Satya introduced the combination of Azure Cosmos DB immediately into Azure AI Foundry, in GA, permitting any AI agent to retailer and retrieve conversational historical past and account knowledge. Whether or not you’re constructing multi-agent apps utilizing Semantic Kernel or LangChain , Azure Cosmos DB provides low-latency persistence, stateful reminiscence, and world scale—precisely what fashionable, autonomous AI brokers have to perform.
Satya additionally highlighted main prospects, comparable to NFL, Carvana, and Sitecore, who’re innovating with Azure Cosmos DB and AI to remodel digital experiences for patrons. XM Cloud, the AI-forward CMS from Sitecore, is delivering extra related content material 10 instances sooner with Azure AI Foundry and Azure Cosmos DB at its core, in response to Roger Connolly, CPO of Sitecore, “enabling the supply of tens of millions of real-time, personalised experiences like by no means earlier than.”
The NFL Mix used Azure Cosmos DB and AI to floor insights for draft selections. The outcome? Coaches and scouts can analyze gamers in seconds utilizing real-time, AI-powered insights—at scale.
On-site attendees acquired to strive their fingers defining capabilities, managing reminiscence, and persisting multi-agent interactions within the first session of our lab, “Multi-agent apps with Semantic Kernel or LangChain & Azure Cosmos DB.” Missed the lab? We’re providing a second occasion on Thursday, Could 22, 8:30-9:45 am PST – RSVP to get expertise constructing scalable and quick multi-agent AI methods for real-world apps.
Actual AI workloads in manufacturing
In What’s New in Microsoft Databases: Empowering AI-Pushed App Dev, VP Azure Databases Shireesh Thota and VP Azure Cosmos DB Kirill Gavrylyuk kicked off the part on Azure Cosmos DB with a particular buyer spotlight: Carvana achieved a discount in inbound calls per sale with their AI brokers powered by Azure Cosmos DB. In a brand new video, Michael Graf, Affiliate Director, Engineering, at Carvana, shared why he loves utilizing Azure Cosmos DB.
Shireesh and Kirill then highlighted a number of 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
Now you can persist agent reminiscence, system messages, and gear outputs immediately in Azure Cosmos DB with full management and safety. This integration powers the Azure AI Agent Service’s Convey-Your-Personal Thread Storage function.
Why it issues: Azure Cosmos DB turns into the default reminiscence and state retailer for constructing safe, scalable multi-agent methods.
Full-Textual content, Hybrid, and Vector Search (GA and Preview)
Search capabilities at the moment are GA and prolonged with:
- Hybrid search utilizing Reciprocal Rank Fusion (RRF)
- Weighted search (e.g., extra relevance from vector or key phrase scoring)
- LINQ and Entity Framework Core integration for .NET devs
Preview options:
- Multi-language full-text search (French, German, Spanish)
- Fuzzy seek for typos and close to matches
- Filtered vector search with DiskANN (automated plan optimization)
- Doc Indexer integration with Azure Logic Apps for parsing PDFs, pictures, markdown, and so on.
Why it issues: These options flip Azure Cosmos DB into a totally AI-ready doc retailer for constructing RAG methods, conversational brokers, and semantic search—all with large scale and pace.
World Secondary Indexes (Preview)
GSIs allow you to outline alternate partition keys and indexing insurance policies in new containers mechanically synced out of your main container.
Why it issues: You’ll be able to optimize question efficiency and isolate workloads—with out transforming knowledge. That is particularly helpful for high-volume apps and evolving microservices the place entry patterns change over time.
Per Partition Computerized Failover (Preview)
Conventional failover affected total accounts. With Per Partition Computerized Failover (PPAF), solely the affected partition fails over—mechanically. It boosts availability and resilience with out requiring advanced logic in your app. Write latency is lowered throughout regional outages with P99 restoration time beneath 2 minutes.
Why it issues: That is helpful for mission-critical apps that require each sturdy consistency and quick restoration, with out the complexity of active-active multi-write battle decision.
Accelerating MongoDB Improvement on Azure (GA + Preview)
In the event you construct MongoDB-compatible apps, Azure Cosmos DB for MongoDB (vCore) simply acquired a serious increase:
DiskANN vector indexing (GA)
- DocumentDB Native (Preview)
- Azure Features triggers (Preview)
- Entra ID integration (GA)
Why it issues: Builders now have fashionable open-source tooling, native growth help, and enterprise-grade scalability—all absolutely suitable with MongoDB APIs.
Material Integration: Cosmos DB in Material (Preview)
Azure Cosmos DB is now a local database in Microsoft Material—mechanically built-in into OneLake, real-time analytics, Copilot-powered Energy BI, and extra.
Why it issues: It eliminates the complexity of constructing knowledge pipelines between your app database and your analytics platform. Cosmos DB knowledge is now queryable and visualizable alongside all of your different sources in Material.
Construct 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 Knowledge Layers for Multi-Tenant Apps with Azure Cosmos DB. Andrew and Deborah unveiled Fleet Administration with RU pooling and World Secondary Indexes—a robust duo of options for SaaS builders.
With these instruments, Azure Cosmos DB now makes it simpler than ever to:
- Create remoted, performant environments for every tenant
- Optimize for question pace and price
- Visualize utilization at scale with fleet analytics
Andrew and Deborah had been joined on stage by Liz Nelson, Sr. Product Lead at Sitecore, who shared how her staff is already leveraging many of those capabilities to ship super-fast experiences throughout Sitecore’s SaaS Content material Administration System (CMS) providing. Liz mentioned her expertise designing for multi-tenancy, leveraging Azure Cosmos DB world distribution, holding prices in verify, and utilizing new options like full-text and semantic search to drive extra related search outcomes when retrieving info.
Straight from our prospects
Along with novel Azure Cosmos DB capabilities introduced in the present day, we debuted many new buyer tales, highlighting real-world purposes and tangible enterprise impression. Take a look at our newest tales:
HEINEKEN unites knowledge and world groups with AI assistant fueled by Azure AI Foundry
The ODP Company transforms HR, gross sales, and retail workflows with Azure AI app platform
NFL Mix features real-time insights for knowledgeable draft day selections with Azure AI
Carvana builds environment friendly, seamless buyer expertise with agent on Azure
From concept to orbit: Terra Mater Studios builds Hera AI companion with Azure knowledge, AI
Meeting Software program saves regulation corporations as much as 25 hours per case with Azure
Nonetheless forward this week
Wednesday, Could twenty first
🗓️ 9:00am PT: Scale and safe MongoDB-compatible apps with Azure Cosmos DB (BRK202) with Khelan Modi and Patty Chow from the Azure Cosmos DB staff, and Hari Krishna Sunder, Principal Engineer, and Rohan Thomas, VP of Accomplice Gross sales, from Yugabyte.
🗓️ 10:30am PT: Past the Sport: How AI and Knowledge Are Powering the Way forward for Sports activities (BRK143) with Indiana Pacers’ Supervisor of Knowledge Engineering Jared Chavez, CTO of Southworks Johnny Halife, James Codella from Azure Cosmos DB, and Andy Beatman from Azure AI.
🗓️ 10:30am PT: Constructing safe AI Brokers with Azure Features (BRK189) with Pim Zandee
World Enterprise Intelligence Lead of Heineken.
💻 2:20pm PT: Suggestions for quick vector and full-text search with Azure Cosmos DB (DEM567) with James Codella and Haiyang Xu.
🗓️ 3:30pm PT: Architecting extremely resilient purposes in Azure: A technical information for builders (BRK197) with Kirill Gavrylyuk, VP of Product, and Justine Cocchi, Sr. Program Supervisor, from the Azure Cosmos DB staff.
🗓️ 3:30pm PT: Earth’s Protection with Hera: AI Brokers Battle Planet Extinction Threats (BRK186) with ImpactAI, who will share how they used Azure Container Apps, Azure Cosmos DB, and Azure OpenAI Service for house exploration.
Thursday, Could twenty second
💻 8:30am PT: Don’t miss the second occasion of our lab, Multi-agent apps with Semantic Kernel or LangChain & Azure Cosmos DB (LAB361 with Mark Brown and James Codella (dwell solely)
Our staff may also be within the Professional Meet-Up space onsite, sales space 810, to share extra about what’s new in Azure Cosmos DB, show a few of the newest capabilities, and hand out swag. Come on by!
