Azure launches VM series powered by Cobalt 100 ARM processors

Azure launches VM series powered by Cobalt 100 ARM processors

Microsoft Azure has launched VM series powered by its custom Cobalt 100 ARM processors, targeting general-purpose and compute-optimized workloads. Preview benchmarks show up to 40% better price-performance compared to equivalent x86 instances for web servers, containerized applications, and open-source databases.

Cobalt 100 specs

Cobalt 100 is based on the ARM Neoverse N2 architecture, offering 128 cores per processor with 12-channel DDR5 memory and PCIe 5.0 support. Azure's initial offerings include the general-purpose Dpsv6 and compute-optimized Fpsv6 series, covering web services, microservices, API gateways, and batch processing.

ARM ecosystem maturity

With rapid ARM ecosystem growth, mainstream Linux distributions, Node.js, Python, Java, Go, and .NET all provide excellent ARM64 support. For new projects or containerized applications, ARM instances require minimal adaptation effort.

16IDC Takeaway

All three major cloud providers now offer custom ARM instances — AWS Graviton, Azure Cobalt, and Google Axion. For site owners and SaaS teams, ARM should be a default evaluation option in the next instance selection round.

Background: Microsoft's Custom Silicon Strategy Takes Shape

Cobalt 100 is Microsoft's custom ARM server processor, based on the ARM Neoverse N2 architecture. Its launch marks Microsoft's official entry into the "cloud provider custom ARM processor" club — directly competing with AWS Graviton and Google Axion.

Microsoft's chip strategy has been building for years — from Xbox custom silicon, to Surface SQ processors, to Azure's Maia 100 AI accelerator. Cobalt 100 targets general-purpose computing, aiming to reduce Azure's operational costs while offering customers higher value compute options.

Notably, Cobalt 100 uses a 128-core design — the highest core count among the three cloud ARM processors. This provides a clear advantage for multi-threaded and parallel general-purpose workloads.

The naming reveals Microsoft's product thinking too: the D in Dpsv6 stands for general purpose and ps for ARM processor, mirroring the Intel-based Dsv6 one-to-one so users can compare within the same series. For teams already on Azure, this means migration is not "a new thing" but "the same series on a different architecture" — the API, VNet, disks, and monitoring stack all stay the same.

Practical Impact for Site Builders

Cobalt 100 vs x86 Comparison

Creating a Cobalt 100 Instance

Provisioning follows the same flow as any Azure VM. A single CLI command is enough — the key is picking the right series and SKU:

az vm create \
  --resource-group myrg \
  --name my-cobalt \
  --image Ubuntu2204 \
  --size Standard_D2ps_v6 \
  --admin-username azureuser \
  --generate-ssh-keys

Note the ps suffix in the SKU means ARM architecture: Dpsv6 is general purpose, Fpsv6 is compute optimized, and the number is the vCPU count (D2ps_v6 is 2 vCPUs). When choosing an image, prefer Ubuntu, Debian, or AlmaLinux builds labeled ARM64 or arm64; Windows Server ARM preview is only worth testing with.

Best Candidates for Migration

1. Web servers and API gateways: Nginx, Apache, Caddy — excellent ARM support
2. Containerized apps: Docker/Kubernetes ARM64 support is mature; just swap base images
3. Open-source databases: MySQL, PostgreSQL, Redis perform well on ARM
4. CI/CD pipelines: Build and test tasks run well on ARM at lower cost

A Typical Migration Case

Say you run a WordPress site on a Standard_D4s_v4 (Intel Xeon) instance at about $80/month. Migrating to an equivalent Standard_D4ps_v6 usually happens in three steps: first containerize the app and build a linux/arm64 image; second, run a week of load testing in a staging environment comparing QPS and P95 latency; third, cut over gradually with a 24-hour rollback window. In practice, sites with a high share of static content often see QPS stay flat or improve while monthly cost drops about 30% — and the whole migration takes one weekend.

Not Yet Suitable

• Windows Server workloads (ARM Windows support still maturing)
• Closed-source commercial software without ARM versions
• Tasks with deep x86 instruction dependencies

Actionable Recommendations

1. Start with new projects: Avoid migration costs by choosing Cobalt 100 for new launches
2. Containerize first: If not yet containerized, do that before migrating to ARM to abstract architecture differences
3. Use Azure migration assessment: Azure Migrate offers ARM compatibility scanning tools
4. Compare TCO: Run real-world cost comparisons between Cobalt 100 and Intel/AMD instances
5. Check regional availability: Cobalt 100 may launch in select regions first

Deeper Perspective

The three-way ARM competition — AWS Graviton, Azure Cobalt, Google Axion — is reshaping cloud cost structures. For users, this is extremely favorable: more choices mean stronger bargaining power and faster innovation.

However, ARM migration is no silver bullet. While price-performance improvements are attractive, migration costs, compatibility testing, and team learning curves must factor into TCO calculations. Take a "gradual migration" approach — start with non-critical workloads and expand ARM usage over time.

Source: Microsoft Azure Blog

Reference: Azure Cobalt 100 VM docs: https://learn.microsoft.com/en-us/azure/virtual-machines/dpsv6-dpsdsv6-series