Oracle Cloud Infrastructure (OCI) today announced the upcoming general availability of A4 compute shapes powered by AmpereOne® M, the latest generation of Ampere-based compute. Launched in December, AmpereOne® M is gaining momentum, with systems available and additional designs under development with lead OEMs and systems builders.

Continuing this expansion with A4, OCI becomes the first cloud provider to launch AmpereOne® M-based instances, delivering significant performance, efficiency, and cost benefits to customers worldwide. A4 builds on the success of the widely adopted A1 and A2 compute shapes—which have grown to serve more than 1,000 customers in over 65 regions. A4 shapes are expected to be generally available in November in Ashburn (IAD), Phoenix (PHX), Frankfurt (FRA), and London (LHR), with additional regions to follow.

Shapes and Performance Details

A4 will be available in both bare metal and virtual machine configurations. Instances scale up to 96 cores running at 3.6GHz, delivering a 20% clock speed increase over A1 and A2. With 100G networking and an expanded 12-channels of DDR5 memory bandwidth, A4 shapes are built to support demanding AI inference workloads, including large language models (LLMs).

“Customers choose OCI for choice and flexibility—broad compute options and flexible shapes from small VMs to large bare metal—so they can align each workload to the right balance of performance, efficiency, and cost,” said Kiran Edara, VP of Compute at Oracle Cloud Infrastructure. “Our upcoming ARM-based Ampere® A4 shape builds on what leaders like Uber and Oracle Red Bull Racing already achieve on OCI—stronger price‑performance, and meaningful power savings—and takes it further so teams can scale cloud‑native services across our global footprint, spend less, and meet their sustainability goals.”

Powered by AmpereOne® M, the A4 compute shapes on OCI deliver Ampere’s most advanced cloud architecture to date, built to provide consistent, efficient performance across a variety of workloads. The design innovations translate into up to 45% higher per-core performance on Cloud Native workloads than OCI A2, Ampere’s previous generation product. A4 is also expected to deliver 30% better price-performance compared to AMD EPYC-based OCI E6 shapes.

“AmpereOne® M was designed from the ground up for cloud and AI workloads, delivering predictable performance, efficiency, and scalability,” said Jeff Wittich, Chief Product Officer at Ampere. “The A4 launch at OCI gives customers access to the full potential of this latest processor, helping organizations accelerate their cloud and AI initiatives.”

Optimized for AI Inference

The rapid scaling of generative AI demands lower cost, energy-efficient compute for AI inferencing at-scale. With AmpereOne® M’s increased memory bandwidth for AI, A4 shapes are purpose-built to deliver on this challenge. Customers running small and mid-sized LLMs are already reporting improvements in Time-To-First-Token (TTFT) and Tokens-Per-Second (TPS), enabling cost-efficient CPU-based deployments for AI inference.

When running Llama 3.1 8B with publicly available software stacks, OCI A4 is expected to offer a substantial 83% price-performance advantage compared to alternatives like Nvidia A10. With A4, customers can benefit from leveraging highly granular, cost-effective resources that scale for better overall performance. This approach contrasts with large and expensive solutions that often require renting the entire unit, making them more expensive upfront and per unit of work, and less elastic.

To accelerate adoption of LLM workloads, Ampere has developed an AI Playground which is an easy entry point for customer adoption. Optimized software libraries and pre-built demos in Ampere’s AI Playground GitHub are helping developers quickly initiate proofs-of-concept and deploy inference-ready applications.

Industry Leaders Among First to Adopt

Several high-profile customers are already moving workloads to A4, with early adopters secured in the US and Europe:

• Uber, which already runs a large portion of its capacity at OCI on Ampere, will deploy additional workloads on A4 in U.S. regions. Uber has already increased price-performance and lowered power consumption with existing Ampere-based infrastructure. The company expects up to 15% more performance from A4, along with further price-performance benefits and a lower carbon footprint.
• Through its partnership with Oracle, Red Bull Racing is already leveraging Ampere instances at OCI to predict optimal race strategy using their Monte Carlo simulations across billions of scenarios and outcomes. The company will adopt A4 instances in London for this use case and other AI and LLM workloads. The team expects a 12% performance boost for its race strategy simulations with A4.

Oracle Accelerates Its Own Adoption

Beyond external customer adoption, Oracle continues to deepen its internal use of Ampere-based compute. Fusion Applications are currently deployed on A1 and are expected to move onto A4, enabling better SaaS performance, while Block Storage is expected to join the growing list of OCI services now powered by Ampere processors.

In addition, Oracle Database software development teams are actively implementing Ampere’s memory tagging capability, which detects memory safety violations to prevent potential exploits. They have reported strong results with almost no added overhead when deploying this feature. Delivering high performance with almost no memory capacity penalty, memory tagging is available on all AmpereOne® Family processors, the only data center processors to do so in production today. This work is another example of how Ampere and Oracle are working together to improve the performance, efficiency, and resilience of modern applications.

Setting the Pace for Cloud Performance

The A4 launch reflects the growing momentum of Ampere at Oracle. Over the last two years, the adoption of Ampere-based compute shapes at OCI has grown rapidly, as enterprises seek greater performance, efficiency, and sustainability. As the first cloud instance powered by AmpereOne® M, the OCI A4 shapes extend this momentum, bringing the latest generation of Ampere innovation to cloud customers.

Related: AmpereOne® M Product Brief

Related: AmpereOne® M in the Cloud: Redefining Scalable AI Infrastructure

Disclaimer

All data and information contained herein is for informational purposes only and Ampere reserves the right to change it without notice. This document may contain technical inaccuracies, omissions and typographical errors, and Ampere is under no obligation to update or correct this information.

Ampere makes no representations or warranties of any kind, including express or implied guarantees of noninfringement, merchantability, or fitness for a particular purpose, and assumes no liability of any kind. Where data or information is sourced from or provided by third-party partners, Ampere does not independently verify such data or information and makes no representations or warranties regarding its accuracy, completeness, or reliability, and assumes no liability of any kind related to such third-party content. All information is provided “AS IS.” This document is not an offer or a binding commitment by Ampere.

Price-performance calculations for A4 are based on an expectation of A2 instance pricing, as final A4 pricing remains unannounced. Actual A4 price-performance may vary. These calculations are illustrative and should not be relied upon as definitive. Ampere assumes no liability for discrepancies with final A4 pricing.

System configurations, components, software versions, and testing environments that differ from those used in Ampere’s tests may result in different measurements than those obtained by Ampere.

©2025 Ampere Computing LLC. All Rights Reserved. Ampere, Ampere Computing, AmpereOne and the Ampere logo are all registered trademarks or trademarks of Ampere Computing LLC or its affiliates. All other product names used in this publication are for identification purposes only and may be trademarks of their respective companies.