Oracle Cloud Infrastructure (OCI) is extending its partnership with chipmaker Ampere by announcing the upcoming launch of its A4 compute instances, powered by AmpereOne M, the latest generation of Arm-based processors. The move positions Oracle as the first major cloud provider to deliver general availability of AmpereOne M-based instances, marking another milestone in the evolution of energy-efficient, high-performance computing for the cloud and AI workloads.
The A4 instances are scheduled for general availability in November 2025, starting in Oracle’s Ashburn (IAD), Phoenix (PHX), Frankfurt (FRA), and London (LHR) regions, with further global rollouts to follow. OCI’s new offering expands on the success of its A1 and A2 compute shapes, which have already attracted more than 1,000 customers across 65 regions worldwide. The A4 launch represents a strategic advance in OCI’s goal to provide enterprises with scalable, cost-efficient, and sustainable cloud infrastructure options.
The AmpereOne M-powered A4 instances will be available in both bare metal and virtual machine (VM) configurations, scaling up to 96 cores clocked at 3.6 GHz – a 20% increase in clock speed over previous generations. The instances also feature 12 channels of DDR5 memory and 100G networking, targeting demanding workloads such as AI inference and large language models (LLMs) that require high throughput and low latency.
VMs to Large Bare Metal
Oracle’s Vice President of Compute, Kiran Edara, said the new shapes reinforce OCI’s philosophy of offering customers broad flexibility across performance, cost, and sustainability. “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. “Our new Ampere A4 shape builds on what companies like Uber and Oracle Red Bull Racing already achieve on OCI, delivering stronger price-performance and measurable power savings while enabling them to meet global sustainability goals.”
The AmpereOne M architecture, introduced in late 2024, is designed from the ground up for cloud-native and AI workloads, offering consistent performance per core and higher efficiency compared to x86-based architectures.
Jeff Wittich, Chief Product Officer at Ampere Computing, emphasized that the A4 deployment demonstrates how modern Arm designs are redefining compute economics. “AmpereOne M was built to deliver predictable performance, efficiency, and scalability for the cloud,” said Mr. Wittich. “With the launch of A4 on OCI, customers can now fully leverage this technology to accelerate their cloud and AI initiatives.”
Cost-Effective Compute for AI Inference
AmpereOne M introduces a number of design improvements, including 45% higher per-core performance compared to OCI’s previous A2 shapes and up to 30% better price-performance than AMD EPYC-based OCI E6 instances. These gains position the A4 shapes as a cost-effective choice for AI inference workloads, where efficiency and predictability are critical.
The scaling of generative AI has created demand for CPU-based inference solutions that offer lower cost and energy consumption than GPU-heavy setups. The A4 instances are specifically optimized for this purpose, thanks to increased memory bandwidth and energy efficiency. Oracle reports that when running Llama 3.1 8B models with standard software stacks, the new A4 instances can deliver an 83% price-performance advantage over Nvidia A10-based alternatives. This makes them particularly suitable for small and mid-sized LLM deployments where elasticity, affordability, and predictable scaling are essential.
To make AI adoption easier, Ampere has introduced an AI Playground, a developer ecosystem with optimized libraries, pre-built demos, and reference models hosted on GitHub. This resource is designed to help organizations quickly prototype and deploy inference-ready applications using Ampere processors on OCI.
Several early adopters have already committed to using the new A4 shapes. Uber, which runs a significant portion of its workloads on Ampere-based OCI infrastructure, plans to extend its deployment to the new instances in U.S. regions, expecting up to 15% higher performance and reduced energy usage. Oracle Red Bull Racing is also set to migrate its AI-driven Monte Carlo simulation workloads to A4 instances in London, targeting a 12% performance boost for its race strategy computations.
Oracle itself is expanding internal use of Ampere-based compute across its own services. Its Fusion Applications currently run on A1 instances and will migrate to A4 for better SaaS performance. Other OCI services, including Block Storage and Oracle Database development, are also integrating Ampere processors. The database team has begun implementing Ampere’s memory tagging technology – unique in the industry – which detects and prevents memory safety violations with minimal performance impact.
The partnership between Oracle and Ampere has rapidly evolved from experimental deployment to strategic foundation. Over the past two years, adoption of Ampere-powered compute on OCI has grown sharply as enterprises seek stronger performance and energy efficiency to support AI and cloud-native transformation. The A4 shapes not only extend OCI’s Ampere portfolio but also reinforce Oracle’s role as a leading cloud provider offering sustainable, high-performance alternatives to x86 and GPU-dominated solutions.
For Ampere, the deployment signifies growing momentum among cloud hyperscalers embracing Arm-based architectures to balance performance with sustainability imperatives. For Oracle, it’s a decisive play in a competitive landscape where cloud efficiency, flexibility, and cost transparency are key differentiators.
