Ampere’s long history in the clouds is shown in the Lift Insights data, but there are dark clouds on the horizon.
SoftBank announced its intent to acquire Ampere Computing for $6.5 billion on March 19, 2025, stating "Ampere’s expertise in semiconductors and high-performance computing will help accelerate [SoftBank's] vision."
Ampere first appeared in Oracle Cloud among the major players. Liftr data show Ampere growth from that first appearance to its 24.0% share of all Arm in the cloud in mid-2024.
The cost savings of Ampere was a selling point to most enterprises. On average, the latest generation Ampere instances can save enterprises about 30% compared to equivalent instances by Intel and AMD.
Graviton was positioned as a low-cost, low-energy solution catering to certain workload classes, such as web servers and development environments.
AWS was the first public cloud provider to deliver Arm-based CPUs in the cloud with its earliest Graviton processors just before Liftr began tracking AWS in 2019. At the time, Graviton was a lower-end alternative to the x86 architecture. It was positioned as a low-cost, low-energy solution catering to certain workload classes, such as web servers and development environments.
Other cloud providers wanted to respond but did not necessarily have the skillset to develop their own silicon. Fortunately, Ampere had a competitive offering that was quickly adopted in many clouds, including Oracle at first, and then Alibaba, Tencent, Azure, and Google. Ampere instances have steadily grown at these cloud vendors over time and, according to Liftr data, represent 18.2% of the total Arm market in the cloud as of December 2024. This number is down from over 20% prior to the introduction of Azure Cobalt and Google Axion.
Arm Market Share in Public Cloud
NVIDIA is using Arm internally in a different way than the cloud providers.
Where does Ampere fit in the Arm ecosystem?
Ampere has a unique position in the Arm ecosystem, where it behaves like Intel or AMD. It is an independent semiconductor vendor that can sell its CPUs to cloud service providers like Microsoft Azure and Alibaba Cloud as well as to on-premise server providers., such as SuperMicro and HPE.
In contrast, as the cloud providers develop their own Arm solutions, they are not building the technology to sell externally. Rather, they are integrating them into their own offerings to their existing customer base.
Likewise, as we will discuss in a future article, NVIDIA is using Arm internally in a different way than the cloud providers.
Unfortunately, Ampere’s role as an independent vendor may have become a liability for Ampere in the cloud during the past 12 months.
Azure currently offers better performance and is 8.8% cheaper than Ampere.
What is the potential impact of Cobalt and other internally developed solutions on Ampere?
During the past two years, Alibaba, Google, and Microsoft have released their own Arm-based CPU instances. These new instances are being positioned as new versions of the previous Ampere-based instances. Here is a comparison of the new Cobalt-based instance vs. the previous Ampere-based instances. Additionally, Azure currently offers better performance and is 8.8% cheaper than Ampere.
Developing their own chips enables cloud vendors to pass on savings to the customer, which is good for customers. However, that does put Ampere at a disadvantage.
Liftr’s data show Ampere processors are currently deployed in five public clouds in December 2024. Three of these five clouds already have their own internally developed solutions. The two remaining solutions represent less than 2.0% of the total Ampere install base, so this competitive threat to Ampere is real.
What should Arm-curious cloud customers consider?
Concerns about vendor lock-in. While the Arm architecture and licensing model provide some degree of compatibility across vendors, some enterprises may be concerned that cloud providers are building their own proprietary hardware and software stacks. Over time, serving CPUs, running applications hosted on proprietary operating systems and other services (such as databases) may be harder to migrate. A lack of mobility may also give a cloud provider more negotiating leverage than is acceptable to some enterprises.
What is the impact on existing Ampere customers?
More options and long-term roadmaps. Three years ago, it was a leap of faith to deploy on Arm in the public cloud. The dominance of x86 and the relative lack of market proof point on Arm’s applicability in the server market were significant. However, innovators like Ampere have shown customers that Arm is viable for both cloud and on-premise servers. The additions seen in the Liftr data are measurable validations of that success. Its initial success has inspired new competition. This is good from a customer perspective, as they will continue to offer multiple Arm-based options.
Increased future testing. Despite these new competitive Arm instances, Liftr Insights analytics indicate that the deployment of Ampere-based instances has slowed, but has not declined. Like any cloud instance, this may change as customers look to upgrade instances over time. Also, like any cloud instance, there may be an increased level of testing at an upgrade due to the new CPU or associated operating system.
Liftr data helps analysts at enterprises easily compare similar instance types across cloud vendors. The data help them predicts costs and potential savings by region, provider, and architecture among many other criteria. Of course, one of the ways to compare is comparing instances for Ampere versus those by Intel, AMD, or the home-grown types.
Conclusion
While more Arm-based variants benefit the consumer, these variants will impact how Ampere goes to market in the public cloud. For example, they may deliver new innovations faster that stay one step ahead of the vendor’s branded CPUs or pursue different workload types. As the Liftr data show, Ampere was instrumental in proving that Arm is a viable strategic option in public clouds. The latest competition from Microsoft and Google is still fresh, so the data has not shown a detrimental impact for Ampere. But, the data will provide analysts the objectivity they need when and if that change does happen.
The next part of this series will discuss AWS, which has become a semiconductor powerhouse in its own right; the initial success with Arm and Graviton has led to a wide range of new products, including AI-based workloads.
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.
To stay current and informed and to be the first to know about the latest reports like these, subscribe to the Liftr Newsletter.