Azure is one of the clouds showing strong Arm adoption, as shown by Liftr Insights data
32.9% of new CPUs deployed on Azure in the fourth quarter of 2024 were Arm-based chips. Specifically, these Arm-based processors were Azure's new Cobalt 100 chipset.
1 out of 3 chips being Arm-based, as shown in Liftr data, is an impressive launch with strategic implications.
Microsoft is making this play to draw businesses interested in lower costs, lower carbon footprints, and higher output for each dollar spent in the cloud. Liftr helps their customers and prospects calculate and forecast these savings with data.
Microsoft made big waves with this first quarter of general availability of Cobalt. 1 out of 3 chips being Arm-based, as shown in Liftr data, is an impressive launch with strategic implications. It's more than what the data showed for the cloud as a whole, which was still a large number.
What is Cobalt?
Microsoft Cobalt 100 is Azure's first-generation 64-bit Arm-based processor, fully designed in-house by Microsoft. It leverages the Arm Neoverse N2 architecture, which has been deployed in a wide range of solutions from other vendors such as Google and Nvidia. However, this variant of Arm has been tuned to perform better on Azure, and Microsoft testimonials claim Cobalt has ~40% better performance than other Arm-based solutions. That performance gain places Cobalt in the same competitive space as Intel's Sapphire Rapids and AMD 4th Gen EPYC 9004 processors.
Microsoft testimonials claim Cobalt has ~40% better performance than other Arm-based solutions.
What workloads is Cobalt targeting?
Cobalt is targeting two types of workloads with this initial offering. First is the common General Purpose computing workload we have seen in other Arm deployments. This includes small to medium databases, web servers, and application servers. Second, Microsoft has also released Memory-Optimized instance types for caching, analytics, and large database workloads. These types of instances open Cobalt to a significant portion of Azure deployments. According to Liftr data, Memory Optimized and General Purpose workloads comprise over 72% of the total workloads deployed across all clouds at the end of 2024.
According to Liftr data, Memory Optimized and General Purpose workloads comprise over 72% of the total workloads deployed across all clouds at the end of 2024.
This is not unlike other cloud providers who are targeting Arm as a low-cost option for a wide range of common workloads. By taking this broad approach, Microsoft is offering customers two migration opportunities.
Opportunity: Migration from Intel and AMD-based instances
The most obvious opportunity is providing existing on premise or existing Azure customers a means to lower their spend on existing Intel or AMD deployments. For example, here's a comparison of similar Cobalt vs. Intel and AMD instances.
This price is based on regions in the United States, but more detailed slices of price can be obtained through Liftr data.
In a Northern European region of Azure, for example, the cost differences are even greater.
The pricing advantage is also important for customers considering migrating from their existing cloud provider. Cobalt makes Azure more price competitive since some Azure instances are priced at a premium versus other clouds. For example, the Azure DDv5 instance above is about 12% more expensive than a comparable Intel-based AWS instance.
The Azure DDv5 instance above is about 12% more expensive than a comparable Intel-based AWS instance.
A comparable pricing story across a wider range of workloads enables Microsoft to offer a compelling cloud consolidation play to enterprises. However, it should also be noted that many of those competitive cloud providers also have their own Arm offering, such as AWS' Graviton and Google's Axion.
Going beyond price performance
While lower cost instances are a high priority for Azure customers, there is another reason to consider Cobalt. The reported energy savings associated with Arm processors can greatly help enterprises as they attempt to meet sustainability goals. This will be appreciated by companies focused on annual ESG metrics.
Why should enterprise customers care?
- Compatibility: Cobalt supports the same operating systems as its Intel and AMD competitors and is also container-friendly. These capabilities mean that migrating to a cheaper Cobalt VM should be straightforward.
- Cost: As noted above in the table, Microsoft is positioning Cobalt as a much lower-cost option than the competition. This has held true for multiple instance configurations evaluated within the Liftr historical dataset.
- Support: Cobalt offers more premium support features for its instances than for the Intel instances.
- Vendor Consolidation: As noted above, for general purpose applications, Azure now has an opportunity to take workloads from other cloud providers. Consolidating more workloads to fewer cloud vendors can go beyond the cost of the instance and extend into IT management overhead or drive bigger volume discounts.
- Sustainability: Arm processors are well known for their reduced energy consumption. In a cloud environment, those savings should be reflected in the price of the VM. Selecting low-energy CPUs may assist some enterprises that have KPIs related to energy savings or other environmental measurements.
What about Ampere instances at Azure?
Cobalt has replaced Ampere, offering more cores and memory options at a 5 to 10% lower price (in the same core configurations).
Microsoft has been deploying processors from Arm CPU manufacturer Ampere since 2022. Those CPUs were available in four series of the generation 5 Dps instances. As of December 2024, Ampere represented less than 5% of the total Azure fleet. However, with the release of the generation 6 Dps instances in late 2024, Cobalt has replaced Ampere, offering more cores and memory options at a 5 to 10% lower price (in the same core configurations). This means Cobalt is the upgrade path for the existing Ampere install base and Ampere doesn't seem to be an option for those wanting the latest generation.
What should corporations be doing with this information?
Corporations should be forecasting their spending and running analyses for the cost savings of switching to equivalent Arm-based instances. It can be difficult to track these comparisons across cloud providers (or even regions within the same cloud provider) and across x86 versus Arm or AMD vs Intel or CPU generations. Liftr data can help provide their comparables, along with pricing by region.
Conclusion
During the past 18 months, Microsoft has made multiple announcements of internally designed semiconductor offerings, including Cobalt. In its first quarter of availability, Cobalt is aggressively priced and positioned.
Another company that is watching Cobalt closely is Ampere. Ampere was the first Arm vendor deployed on Azure and has had success. Part 3 of this series will discuss Ampere and the impact on them from private-labeled Arm processors like Cobalt.
Corporations and organizations relying on compute instances and managing their bottom line and sustainability goals, can keep up-to-date with Liftr data for their forecasting and planning needs. A simple solution is one of Liftr's Supporting Data for Global AI & Cloud reports.
Feel current and informed with the first notice of the latest reports like these; subscribe to the Liftr newsletter.