Inside Meta and AWS' Groundbreaking Graviton Chips Deal

Meta is diversifying its compute portfolio, bringing tens of millions of Graviton cores with its new AWS deal – making it one of the largest Graviton customers in the world.

The agreement expands on the long-standing partnership between the two giants to include Amazon’s custom silicon, supporting Meta’s broader goal of building the next generation of AI. 

The first deployment will focus on Graviton5 cores that are purpose-built for CPU-intensive workloads, with the flexibility to expand as AI capabilities grow at Meta. The partnership also includes the use of Amazon Bedrock at scale to support Meta's future AI goals.

Faster bandwidth for data processing

AWS Graviton is an ARM-based CPU, which is a central processing unit with a chip that handles general computing tasks. GPUs, which are graphical processing units on the other hand, remain the mostly used ones for training large models.

However, once these models are trained, AI agents built on top of them, creating compute-intensive workloads like real-time reasoning, writing code, search and the coordination involved in managing agents through multi-step tasks. This has caused a shift in the type of chip needed as GPU chips fail to do the task efficiently.

The Graviton chips developed by AWS are designed specifically to handle these AI-related compute needs. This will power various workloads at Meta, which requires infrastructure that can handle billions of interactions while coordinating complex, multi-step agent workflows. 

Nafea Bshara, Vice President and Distinguished Engineer at Amazon, explains: “This is not just about chips; it is about giving customers the infrastructure foundation, as well as data and inference services, to build AI that understands, anticipates and scales efficiently to billions of people worldwide.”

The Graviton5 chip features 192 cores and a cache that is five times larger than the previous generation. This reduces delays in core communication by up to 33%, providing the bandwidth for faster data processing which is key for agentic AI systems that must continuously coordinate multi-step workflows.

Ticking the energy-efficiency box

The AWS Graviton5 is built on 3-nanometer chip technology, a manufacturing process that produces smaller and more efficient processors.

AWS can also optimise performance and energy efficiency in ways that off-the-shelf processors cannot match as it designs its chips from the ground up. 

Graviton5 delivers up to 25% better performance than the previous generation. This helps  Meta pursue ambitious AI goals and deliver strong performance while staying on track with its sustainability targets.

Additionally, the system is built on the AWS Nitro System which uses dedicated hardware and software for high security, performance and availability.

The range of Graviton5 instances also supports low-latency communication enabled through the Elastic Fabric Adapter (EFA). This is essential for Meta’s agentic AI workloads, where large-scale tasks need to be distributed across many processors working in coordination. 

The deal between Meta and AWS is forged on the back of a principle of requiring a diversified approach to infrastructure to build AI at Meta’s scale. The company invests in its own data centres and custom hardware while partnering with cloud providers to ensure the right compute is available for specific tasks.

Santosh Janardhan, Head of Infrastructure at Meta, says: “As we scale the infrastructure behind Meta’s AI ambitions, diversifying our compute sources is a strategic imperative. AWS has been a trusted cloud partner for years, and expanding to Graviton allows us to run the CPU-intensive workloads behind agentic AI with the performance and efficiency we need at our scale.”

By adding AWS Graviton cores, Meta ensures its infrastructure can handle billions of interactions while coordinating complex agentic experiences for a global audience.

Signalling a new chapter in how large-scale AI infrastructure will be built, the new agreement is a step forward in how purpose-built chips like Graviton can help companies like Meta deliver the rising AI demands.