SUNDAY, NOVEMBER 22, 2026 Tech news, minus the noise.
N News Cast
LATEST
Home  /  Cloud
Cloud

Azure Cobalt 200 Is Generally Available, and Arm Is Now the Center of Microsoft's Cloud

Microsoft announced general availability for Azure Cobalt 200 VMs on November 17. Its second-generation in-house Arm processor is up to 50 percent faster than Cobalt 100.

Azure Cobalt 200 Is Generally Available, and Arm Is Now the Center of Microsoft's Cloud
Source: Ioan Sameli (BY-SA), via Openverse

Microsoft announced general availability for Azure Cobalt 200 VMs on November 17, and the number that matters is 50 percent. The second-generation in-house Arm processor is up to 50 percent faster than Cobalt 100, the chip it replaces, and Microsoft is positioning it squarely at agentic cloud compute: the always-on, always-reasoning workloads that now dominate its data centers.

Cobalt 200 is Microsoft's own silicon, designed in-house and deployed across Azure, in the same way Amazon has Graviton and Google has Axion. The pattern is now unmistakable. Every hyperscaler has decided that the most important infrastructure in its data centers is the processor itself, and that renting that decision from Intel or AMD is no longer the default. Owning the chip means owning the performance curve, the power curve, and the cost curve, and those three curves decide who wins the cloud price war.

Why Arm, and Why Now

The shift to Arm is not about raw speed. It is about efficiency per watt, and efficiency per watt is the metric that decides whether a data center full of agents is affordable. AI agents do not work in bursts like a web request. They run long, looping workflows: reading, reasoning, calling tools, waiting, reasoning again. That pattern burns power continuously, and a processor that does the same work with less electricity is worth more than a processor that does it slightly faster.

That is the bet behind Cobalt 200. Microsoft says the chip is built for agentic compute, which is a way of saying it is built for the workloads that will fill Azure's growth for the next five years. The 50 percent figure is a peak claim, and real-world gains will vary by workload, but even half of it would be a meaningful improvement for customers running agent fleets, data processing, and the web services that sit in front of them.

There is also a strategic layer that has nothing to do with benchmarks. When Microsoft designs its own chip, it controls the roadmap. It can tune the processor for the exact workloads Azure runs most, add accelerators for the operations its AI services need, and ship improvements on its own schedule instead of waiting for a vendor's product cycle. That control is the real prize, and it is the same prize Amazon and Google are chasing with their own silicon.

What It Means for Customers

For teams running on Azure, the practical question is whether to move workloads to Cobalt 200 VMs. The honest answer is to test first, on the workloads that actually dominate your bill. Arm-based instances have been cheaper per unit of work for years in the right workloads, and more expensive in the wrong ones, because not every application is compiled and tuned for Arm. Microsoft has been closing that gap with better tooling and broader support, but the gap is not zero, and a migration that nobody benchmarks is a migration that can quietly cost more.

The bigger picture is the direction of the market. AWS has Graviton, Google has Axion, and now Microsoft has a second-generation Cobalt that it is confident enough to make generally available and promote for its most important workloads. The x86 era of the cloud is not over, but it is no longer the only story, and for new workloads the default question has flipped. It used to be why Arm. Now it is why not.

Cobalt 200 will not change anyone's cloud bill tomorrow. It is a general availability announcement, not a price cut. But it is a clear signal about where Microsoft is putting its silicon investment, and that signal is Arm, agents, and efficiency. Teams planning their next architecture should take the hint and at least benchmark the option. The cheapest cloud is increasingly the one that wastes the least power, and that is a race Microsoft now intends to win with its own chips.

AS
News Cast Editorial Team

We cover AI, gadgets, startups and security — original reporting and explainers, written by humans, for humans. Got a tip? Reach us via the contact page.