Microsoft’s ARM-based Cobalt 200 chips go GA. Why it matters for AI agents.

The New Stack (AI)
Read full story· 5 min
Microsoft has launched its Cobalt 200 virtual machines with second-generation Arm-based CPUs, offering up to 50% better performance than the previous generation. These VMs, available in 14 Azure regions, improve AI agent tool execution for tasks like code compilation and sandbox operations.

What happened

Microsoft has made its Cobalt 200 virtual machines, powered by second-generation Arm-based CPUs, generally available in 14 Azure regions. The announcement was made at the Ignite conference in San Francisco. These VMs offer up to 50% better performance than the previous Cobalt 100 generation and are designed to improve execution for AI agent tools, such as code compilation and sandbox operations. Microsoft plans to expand availability to 23 regions by the end of the year.

Key details

  • Each VM can scale up to 128 vCPUs, each with 3 MB of private L2 cache.
  • Two processor modes are available: deterministic (fixed 3.2 GHz) and opportunistic (up to 3.6 GHz).
  • Cobalt 200 VMs include Azure Boost for improved networking and storage, with up to 15% higher network bandwidth and 20% higher remote storage IOPS compared to Cobalt 100.
  • Additional VM families launched: high-memory Mpsv4 (16 GiB/vCPU) and storage-optimized Lpsv5 (up to 23 TB local NVMe storage).

Why it matters

Developers and businesses using Azure can now access faster and more efficient virtual machines for workloads that support AI agents, such as code compilation and sandboxed tasks. This enables improved performance for services relying on these operations, potentially reducing execution times and infrastructure costs. Organizations running large in-memory databases, enterprise workloads, or requiring high storage and network throughput can benefit from the new VM options and enhanced capabilities.


More in Agents