AWS Graviton5 is now generally available, bringing faster cloud performance for AI, databases, and demanding workloads
Amazon has officially made AWS Graviton5 generally available, introducing what it calls its fastest and most efficient cloud CPU to date. Built for modern workloads such as artificial intelligence, machine learning inference, databases, analytics, real-time gaming, and large-scale enterprise applications, Graviton5 represents a major step forward in Amazon’s custom silicon strategy.
The new processor is now available through Amazon EC2 M9g and M9gd instances, giving customers access to higher performance, improved efficiency, faster memory, expanded cache, and next-generation I/O technologies including PCIe Gen6.
Graviton5 arrives with up to 25% better performance than Graviton4. Amazon says applications running on the new chip can see up to 35% faster performance, while AI and machine learning inference workloads can also improve by up to 35%. Database workloads are expected to gain up to 30% better performance, making the chip especially attractive for companies running large-scale cloud infrastructure.
A major part of that performance boost comes from the design of the processor itself. Graviton5 features 192 cores, doubling the core count compared with the previous generation. It is built using 3nm process technology and offers 33% lower inter-core latency, helping data move more quickly between cores and improving responsiveness in complex workloads.
Amazon’s Graviton journey began in 2018, when the company introduced its first custom cloud processor. That original chip showed that energy-efficient processor technology, similar in concept to what powers mobile devices, could also handle serious cloud computing tasks.
Graviton2 followed in 2019 and delivered a major leap in value, offering up to 40% better price performance than comparable processors. This made it easier for businesses to reduce infrastructure costs while maintaining strong application performance.
In 2021, Graviton3 pushed efficiency even further, using up to 60% less energy for the same level of performance as comparable Amazon EC2 instances. That made it appealing not only for cost savings but also for companies focused on reducing their environmental impact.
Graviton4 arrived in 2023 with a much larger core count and stronger performance for demanding workloads such as large databases, analytics, and high-throughput enterprise applications.
Now, Graviton5 continues that momentum with more cores, more cache, faster memory, and better connectivity.
One of the most important upgrades in Graviton5 is its larger cache. The chip includes a five-times larger L3 cache compared with the previous generation. Each core has access to 2.6 times more L3 cache than Graviton4, which helps keep frequently used data closer to the processor.
This matters because when a processor has faster access to data, applications can respond more quickly. Larger cache can reduce delays, improve throughput, and help with workloads that repeatedly access large amounts of information. This is especially valuable for AI inference, machine learning, databases, analytics, and high-performance cloud applications.
Memory performance has also been improved. Graviton5 supports faster DDR5 memory speeds of up to 8800, which Amazon describes as the fastest DDR5 memory available in the cloud. Faster memory allows customers to process larger datasets more efficiently and run memory-intensive applications with fewer bottlenecks.
The processor also supports PCIe Gen6, giving it access to the latest generation of high-speed I/O connectivity. This can benefit workloads that depend on rapid data movement between processors, accelerators, storage, and networking components.
Networking and storage performance are also getting a boost. Across instance sizes, Graviton5 offers up to 15% higher network bandwidth and 20% higher Amazon Elastic Block Store bandwidth on average. For the largest instances, network bandwidth can be up to twice as high as before.
These improvements can lead to faster data transfers, quicker backups, smoother distributed applications, and better performance for cloud services that rely heavily on network communication or storage access.
The first Graviton5-powered instances are Amazon EC2 M9g and M9gd. The M9g instances are designed for general-purpose workloads and provide up to 25% better compute performance than the previous generation.
The M9gd instances are aimed at businesses that need high-speed local SSD storage. These instances offer up to 11.4 TB of local storage capacity and up to 30% higher IOPS than the prior generation, making them useful for workloads that require fast access to local data.
Amazon says more than 120,000 customers are already building on Graviton processors, and adoption continues to grow. Major companies including Meta, Uber, and Snowflake are among the customers using or planning to use Graviton-based infrastructure. Meta has committed to deploying tens of millions of Graviton cores for agentic AI workloads, showing how important custom cloud CPUs are becoming for large-scale artificial intelligence systems.
The launch of Graviton5 strengthens Amazon’s position in custom cloud silicon. Instead of relying only on traditional server processors, AWS has spent years building its own chips to improve performance, efficiency, and cost control across its cloud platform.
For customers, the appeal is clear: better performance per dollar, improved energy efficiency, and access to hardware optimized specifically for cloud workloads. As AI, machine learning, analytics, gaming, and database applications become more demanding, cloud providers need processors that can deliver more compute power without dramatically increasing cost or energy use.
Graviton5 is designed to meet that challenge. With 192 cores, faster DDR5 memory, larger cache, PCIe Gen6 support, stronger network bandwidth, and improved storage performance, it is built for the next generation of cloud computing.
Now that AWS Graviton5 is generally available in M9g and M9gd instances, businesses can begin deploying workloads on Amazon’s most advanced custom CPU yet. For organizations looking to improve cloud performance, reduce infrastructure costs, and prepare for more demanding AI and data-heavy applications, Graviton5 could become one of the most important cloud processor options available today.
