How do AWS Graviton5 chips enhance AI workloads?
AWS says Meta committed to deploying millions of Graviton cores for its agentic AI efforts, while Uber and Snowflake are also adopting them for similar workloads.
On 10 June, 2026, Amazon announced the launch of AWS Graviton5, claiming to deliver 25% better compute performance than the previous generation for agentic AI workloads.
The e-commerce giant said that its new Graviton5 processor features 192 cores and 33% lower inter-core latency, optimizing it for real-time reasoning, coding, and multi-step tasks. The company also said that its AWS Graviton5-powered EC2 M9g and M9gd instances, which debuted at re:Invent 2025, are now generally available through standard adoption paths.
Amazon also said Meta has committed to deploying millions of Graviton cores for its agentic AI efforts, while Uber and Snowflake are also adopting them for similar workloads, joining over 120,000 existing customers on the platform.
According to the company, compared to the previous generation, M9g instances deliver 35% faster web applications and machine learning inference, as well as 30% faster database performance. Meanwhile, M9gd instances add up to 11.4 TB of NVMe SSD storage and a 30% boost in IOPS.
It also said that both are built on the sixth-generation AWS Nitro System, which features the Nitro Isolation Engine for mathematically proven virtual machine isolation. Additionally, the underlying chip supports DDR5-8800 memory and PCIe Gen 6.
Proven Performance Results
According to the company, ClickHouse, an open-source column-oriented database, saw a 36% performance boost compared to M8g with zero code changes.
Honeycomb, a software observability platform for monitoring complex cloud applications, achieved 36% higher throughput per core than Graviton4 in a 6-month A/B test of production workloads.
Additionally, HubSpot, a customer relationship management (CRM) software company, deployed M9g for its MySQL databases and saw query durations drop by up to 60%.
Amazon told customers they can use Graviton for a wide variety of workloads, including web apps, microservices, databases, and AI inference.
To meet growing demands for more processing power and bandwidth, the company designed these new instances to efficiently package compute, memory, and I/O to maximize energy utilization.
Amazon claims that the new chip solidifies AWS's dominance in custom silicon, providing enterprises with a rare advantage: significantly improved power for next-generation AI tasks combined with a reduced carbon footprint and lower costs.
Key Takeaways
- AWS Graviton5 chips deliver 25% improved performance for agentic AI workloads over the previous generation.
- Meta, Uber, and Snowflake are adopting Graviton5 cores, enhancing their AI capabilities.
- M9g instances provide 35% faster web applications and 30% improved database performance compared to M8g.
- Graviton5 supports DDR5-8800 memory and PCIe Gen 6, optimizing data transfer and processing speeds.
- Real-world tests show performance boosts of up to 60% in specific database queries using M9g instances.