Microsoft has revealed its latest generation Cobalt CPU, as well as the latest version of its Azure Boost system.

The new CPU – dubbed the Azure Cobalt 200 – is an Arm-based CPU designed for cloud native workloads, and was detailed in a blog post.

According to Microsoft, the Cobalt 200 delivers a 50 percent increase in performance compared to the previous generation, Cobalt 100, and is the result of an evaluation of more than 350,000 configuration candidates.

The Cobalt 200 SoC is developed using the TSMC 3nm process and is based on the Arm Neoverse Compute Subsystems V3 (CSS V3).

Each Cobalt 200 SoC includes 132 active cores with 3MB of L2 cache per core and 192MB of L3 system cache. According to Microsoft, the Cobalt CPUs also offer individual per-core dynamic voltage and frequency scaling, which enables each of the 132 cores to run at a different performance level.

The first production Cobalt 200 servers are live in Microsoft data centers, and a wider rollout and general availability to customers is expected for 2026. The previous generation – the Cobalt 100 – was made generally available in November 2024.

In addition to the new CPU, Microsoft has also unveiled the next generation of Azure Boost, a system that offloads the virtualization process from the host to purpose-built hardware and software.

Detailed in a separate blog post, the new generation of Azure Boost provides up to 1 million IOPS and 20GBps of throughput for remote storage, which enables faster data access and greater reliability for demanding workloads.

The new generation also offers Remote Direct Memory Access (RDMA), which moves data between servers at high speed with very low delays, which are ideal for AI and HPC workloads, and can run across regions.

The new generation is currently available in preview on v7-series VMs and will be more widely rolled out in 2026.

While Microsoft is continuing to innovate internally, during a recent podcast interview, CEO Satya Nadella said that as part of its deal with OpenAI, the company has rights to OpenAI's IP from its research efforts, excluding consumer hardware. OpenAI is currently working on a customer AI semiconductor with Broadcom.