Build agentic Al applications with Azure Cobalt 200 | ODSP2181
Sameer Nori presents how Arm and Microsoft’s collaboration supports cloud and AI workloads, with a focus on Azure Cobalt 200. The session combines customer experiences, performance comparisons, and a demonstration of a customer-service agentic AI workflow.
Azure Cobalt 200 and the Arm ecosystem
The session introduces the Arm and Microsoft partnership and examines Cobalt 200 performance improvements. It covers Arm software ecosystem support for cloud-native, AI, and enterprise workloads, alongside customer adoption of Cobalt infrastructure.
The technical overview includes Cobalt 200 architecture, its VM portfolio, and performance benchmark comparisons. A central topic is the role of CPUs in agentic AI workloads and how Arm-based infrastructure supports those workloads.
Customer experiences: Elasticsearch and GitHub runners
Two customer segments explore adoption and workload performance:
- Elastic: Elasticsearch, vector search, and AI workload gains on Cobalt 200.
- GitHub: Arm runner adoption, CI/CD benefits, and Cobalt 200 performance improvements.
These segments connect the infrastructure discussion to search, AI, and development workflows.
Agentic AI workflow demonstration
The session examines Cobalt 200’s agentic compute advantages, including sandbox speed, density, and scalability. A demonstration then shows a customer-service agentic AI workflow running on Cobalt 200 infrastructure.
Migration and optimization resources
The closing segment covers tools, resources, and best practices for migrating and optimizing applications for Arm-based cloud environments. It includes the Arm MCP Server and Performix optimization tools.
Session chapters
- 00:00: Introduction to the Arm and Microsoft partnership for agentic AI on Azure Cobalt 200
- 01:15: Arm ecosystem overview, Microsoft collaboration, and Cobalt 200 performance improvements
- 02:05: Arm software ecosystem support for cloud-native, AI, and enterprise workloads
- 03:10: Customer adoption of Cobalt and the rise of CPU-driven agentic AI workloads
- 04:34: Cobalt 200 architecture, VM portfolio, and performance benchmark comparisons
- 06:50: Elastic on Elasticsearch, vector search, and AI workload gains
- 09:17: GitHub on Arm runners, CI/CD benefits, and performance improvements
- 12:16: Agentic compute advantages: sandbox speed, density, and scalability
- 14:05: Customer-service agentic AI workflow demonstration
- 15:53: Migration resources, Arm MCP Server, Performix optimization tools, and closing guidance
This pre-recorded session is part of Microsoft Ignite 2026, listed as ODSP2181, Cloud & AI, at the 200 — Intermediate level, in English (US). Event information is available at Microsoft Ignite.