Amazon Web Services (AWS) today previewed additional instances for its cloud services based on an AWS Graviton5 processor that delivers 25% better performance than previous versions of the Arm-based chip.

Announced at the AWS re:Invent 2025 conference, the AWS Graviton5 processor provides access to 192 cores per chip and a five times larger cache via an Amazon EC2 M9g instance. To achieve that goal, AWS has reduced the distance data must travel between cores to cut communication latency by up to 33% while increasing bandwidth.

Based on 3nm chips and scheduled to be generally available in 2026, the AWS Graviton5 also provides access to a five times larger L3 cache, with each core having access to 2.6x more L3 cache than the Graviton4 processor without substantially increasing energy consumption. Memory performance has also improved to enable the Amazon EC2 M9g instance to run workloads such as real-time gaming, high-performance databases, big data analytics, application servers, and electronic design automation (EDA) applications.

Network and storage bandwidth have also been increased to support these applications, with on average up to 15% higher network bandwidth and 20% higher Amazon Elastic Block Store (EBS) bandwidth across instance sizes, and up to twice the network bandwidth for the largest instances. Like other AWS instances, Amazon EC2 M9g makes use of AWS Nitro cards to offload virtualization, storage, and networking functions to dedicated hardware.

AWS is also readying C9g instances for compute-intensive workloads and R9g instances for memory-intensive workloads that will be added in 2026.

In the meantime, the number of workloads deployed on Graviton processors continues to grow. AWS revealed today that for the third year in a row, more than half of new CPU capacity added to AWS is powered by Graviton, and that 98% of the top 1,000 EC2 customers, including now Apple along with Adobe, Airbnb, Atlassian, Epic Games, Formula 1, Pinterest, SAP, Siemens, Snowflake, and Synopsys, are running workloads on Graviton processors.

Ali Saidi, vice president and distinguished engineer at AWS, said in total there are 90,000 organizations running workloads on Graviton processors. Additionally, Graviton and other processors that AWS has designed are the default option for running all the managed services that AWS provides.

Most Java applications can run on Arm processors, with applications written in C and Rust requiring only minor changes, noted Saidi. While there are some applications running on databases from Microsoft and Oracle that still require x86 processors, the number of applications that can run on Graviton processors, which are being upgraded every two years, only continues to increase, especially as FinOps teams apply more rigor to reining in cloud computing costs, he added.

Daniel Newman, CEO of the Futurum Group, said it’s clear AWS is relying more on custom silicon to run more applications at scale in a way that improves its profit margins. Providers of alternative processors will need to make a stronger case for using AWS instances based on their processors, he added.

AWS will continue to provide access to x86 processors, including 5th Gen AMD EPYC processors that are now generally available on the Amazon Elastic Compute Cloud (Amazon EC2) service, but the primary focus will increasingly be on raising the percentage of workloads running on custom AWS silicon.