Microsoft has unveiled its next-generation Cobalt CPU, which delivers a 50% performance boost over its predecessor, alongside new Azure AI tools for building and deploying generative and agentic AI applications. The announcement came today during the Microsoft Ignite conference in San Francisco.
The new processor, Azure Cobalt 200, is a custom CPU optimized for cloud-native applications on Azure and is now in preview. The chip features more cores, a larger cache, and faster memory than Cobalt 100, which debuted in 2023 and became generally available last year.
Microsoft designs Cobalt in-house, tailoring it specifically for Azure’s data center infrastructure. Cobalt 100 currently powers Microsoft services like Teams and Defender for Endpoint security, while Azure customers can access it as virtual machines for their own workloads.
“A 50% performance increase from one generation to the next is huge, and the fact that they’re using it for their own applications and that their clients can port to it is a huge benefit for Microsoft,” said Jim McGregor, principal analyst at Tirias Research, in an interview with Data Center Knowledge.
“This isn’t just a minor tweak or change in process technology,” McGregor added. “They’re investing heavily in it and putting a lot into this to make it an optimal architecture.”
While major cloud providers – Azure, AWS, and Google Cloud – partner with chipmakers like Nvidia, AMD, and Intel, they also find it advantageous and cost-effective to develop custom chips that optimize their data center operations and cloud services.
Analysts expect strong adoption of the second-generation Cobalt chip among Microsoft’s customers. Although Cobalt 100 was decent, it was a first-generation chip, according to analyst Matt Kimball of Moor Insights & Strategy. Cobalt 200 delivers substantial improvements in both power and performance. The chip’s foundation on the Arm Neoverse architecture, its 3nm process node, and Microsoft’s engineering expertise drive these advancements.
“While the first-generation Cobalt chip had good success, I imagine Microsoft will see considerable pickup with Cobalt 200,” Kimball told Data Center Knowledge. “The average Azure customer is an enterprise customer that likes to let newer technology prove out before deploying.”
Scott Guthrie, Microsoft Executive Vice President of Cloud and AI, discusses the Cobalt 200 chip at Microsoft Ignite.
Azure AI Platform Enhancements
Microsoft also introduced advancements to Microsoft Foundry (formerly Azure AI Foundry), a platform-as-a-service offering for enterprises that develop, deploy, and manage generative AI and agentic AI applications.
One key update is support for Model Context Protocol (MCP), an open-source protocol that enables AI applications to connect data sources and tools. With MCP integration, developers can use Microsoft Foundry to discover, connect, and manage public and private MCP tools for agents from a single, secure interface.
Microsoft also announced that more than 1,400 business systems, including SAP, Salesforce and Workday, are now accessible as MCP tools through Azure Logic Apps connectors. This feature is now in public preview.
Kimball said this connectivity is important for agentic AI to reach its full potential. “Think about a manufacturing company trying to automate its supply chain management. The number of suppliers will be dozens to hundreds of vendors,” he said. “Automating this connectivity, authentication, and interaction helps enterprise developers enable this supply chain optimization so much quicker and securely.”
Microsoft Foundry now includes a model router, which is generally available. The model router automatically selects the most suitable model for a given task, helping developers optimize cost, performance, and complexity, Microsoft said.
Additional Azure Innovations
Microsoft announced several more innovations for its Azure platform:
Fabric IQ (now in preview) introduces a semantic intelligence layer that transforms Microsoft Fabric from a unified data platform to a unified intelligence platform, according to the company. The cloud-based data platform now provides holistic, live, and context-rich insights for smarter decision-making and enhanced business outcomes.
Foundry IQ (now in preview) is the next generation of retrieval-augmented generation (RAG), providing improved agent-data connectivity from a single knowledge base, the company said.
Azure Copilot (currently in private preview) now integrates agents directly into the Azure portal, PowerShell, and command-line interface, offering capabilities in six areas: deployment, cloud migration, usage optimization, observability, resiliency, and troubleshooting.
Microsoft Fabric databases (now generally available) unify SQL database and Cosmos DB in a single software-as-a-service offering.
Microsoft’s messaging regarding Microsoft Fabric databases and the platform overall is clear: Bring us any database, and we will not only host it but also unify it and make it ready for AI, said Forrester analyst Indranil Bandyopadhyay.
“Microsoft is aggressively positioning Azure as the single, preeminent platform for running all data workloads, specifically to win the next frontier: AI,” Bandyopadhyay told Data Center Knowledge.