Hot AI News

Azure Catalyst sends AI agents to modernize old apps and databases together

Microsoft's Azure Catalyst uses AI agents to discover, plan and migrate legacy apps, while new AKS features aim to run those agents in isolated sandboxes.

Azure Catalyst sends AI agents to modernize old apps and databases together
Image: Microsoft

Microsoft has introduced Azure Catalyst, an agent-driven service for modernizing applications, with new previews that handle apps and their databases together and move BizTalk workloads to Azure Logic Apps. It was announced at Ignite on November 17.

What Catalyst does, and what is still unspecified

According to the post, written by Jeremy Winter, who leads product for the Azure platform, Catalyst covers estate insights, AI-powered application discovery, modernization strategy with cost comparisons, and AI-generated migration waves. Microsoft says it works across VMware, Hyper-V and physical servers, supports governance rulebooks and bring-your-own-model setups, and lets teams save reusable runbooks and skills.

Azure Catalyst sends AI agents to modernize old apps and databases together
Image: Microsoft

It also connects to other Microsoft products. Microsoft Security scans the estate and ranks risk, and GitHub Copilot is used to fix vulnerabilities. Microsoft says it already covers Java and .NET modernization.

The post claims results “in a fraction of the time and cost of traditional methods” but gives no price. It also names no regions and does not label Catalyst as a whole as preview or generally available. Only the two new previews are marked as such. Anyone planning a migration will need to check the linked Catalyst page for terms.

Microsoft’s own numbers come from internal teams. It says 1.4 million applications have been modernized, though the post does not make clear how many used Catalyst. Microsoft Quantum and Xbox reportedly reached 80% faster modernization, and the Intune team upgraded a key library 75% faster than expected. GitHub, the post says, ported 800,000 lines of its Copilot agent harness from TypeScript to Rust. None of these figures is independently verified.

Advertisements

Sandboxes, agent networking and GPU placement

The second half of the post covers infrastructure for running agents rather than modernizing code. For Azure Kubernetes Service, Microsoft describes an Agent Sandbox extension offering VM-level isolation per session, pre-warmed pools for fast startup and a serverless option. Agent Grid handles agent networking with authenticated, encrypted connections and default-deny isolation.

Context caching reuses already-processed prompt context instead of recomputing it on the GPU, and Microsoft says the cached data stays in the customer’s Azure tenant. An AKS AI Manager is meant to place each deployment where the fleet has the best GPU capacity. Flex nodes for AKS are in preview, and Windows Server 2028 is in public preview with a focus on running AI inference next to business-critical applications. The post gives no status or availability for the sandbox, Agent Grid or AI Manager features.

Rockwell Automation’s Justin Litherland is the only customer quoted: “Running our factory-level AI applications on Azure gives us one foundation for different AI workloads in production.”

Microsoft is selling the migration and the place agents run

The analysis angle here is strategy. Catalyst is a way to pull legacy estates onto Azure, and the infrastructure items are where the resulting agent workloads would run. Microsoft is closing that loop with hardware as well. In a separate press release from Microsoft and Intel, dated November 17, Microsoft says it is the first company to announce using Intel Foundry’s EMIB advanced packaging for its own Azure silicon. The release also confirms multi-year Xeon supply capacity, including the Diamond Rapids generation.

That release makes performance claims with footnoted test setups. It cites up to 2.3x more performance and memory flexibility on Azure M-series v4 VMs using CXL memory, and up to 2.5x faster SQL backups on the first Intel QuickAssist-enabled VMs. Both are Microsoft and Intel figures from their own tests, and the release itself says results may vary.

The migration push also builds on earlier work. Microsoft recently previewed a tool that converts Oracle application code to PostgreSQL inside a repository, which we covered on November 18. Catalyst’s joint app-and-database previews read as the broader version of the same idea. Our earlier Ignite report covered Cobalt 200, the Arm-based chip that Microsoft also promotes in this week’s infrastructure announcements.

Advertisements

Who can try it today

Right now, the clearest hands-on items are the two Catalyst previews, the AKS flex nodes preview and the Windows Server 2028 preview. Microsoft points readers to aka.ms/Azure-Catalyst and aka.ms/modernization-ignite26 for details. A Modernize CLI preview was also listed in Azure Updates, but the entry did not load for us, so its details are not covered here.

The announcement does not establish what Catalyst costs, how accurate its migration waves are on messy estates, or how the isolation and networking features perform at scale. Those answers will have to come from customers and Microsoft’s Ignite sessions.

Sign Up for the Daily AI Pulse

One email a day. All the stories that matter.

Alex CarterAI News & Big Tech Correspondent

Alex Carter writes for Aiholics, keeping readers updated on the fast-paced world of AI and Big Tech.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may also like