Meta Expands AI Compute Power with Massive AWS Graviton5 Deployment
100Meta is stepping up its efforts to boost AI capabilities by forming a new partnership with Amazon Web Services (AWS). The deal will see Meta deploying tens of millions of AWS Graviton5 cores, a type of CPU designed for high-performance workloads. Each Graviton5 chip packs 192 cores, and Meta plans to expand this deployment as its AI projects grow. This makes Meta one of the largest users of Graviton chips globally, highlighting its focus on controlling its AI infrastructure.
Meta’s Hardware Strategy and Partnerships
Meta’s move continues its pattern of working with a variety of chip and compute providers. The company already has partnerships with Nvidia, Arm, and AMD, and has developed its own custom chips for training and running AI models. This diverse approach allows Meta to optimize its infrastructure for different workloads and avoid relying on a single technology. The company recently announced four new generations of its own accelerator chips for AI training and inference, underlining its focus on in-house hardware development.
Meta also signed a significant deal with AMD to access over 6 gigawatts of CPUs and AI accelerators. Additionally, it partnered with Nvidia to incorporate millions of GPUs from its Blackwell and Rubin series, along with Nvidia Spectrum-X Ethernet switches. These collaborations show Meta’s broad strategy to build a robust and flexible AI infrastructure, capable of handling a variety of tasks and workloads.
The Role of CPUs in Next-Gen AI Workloads
The new AWS Graviton5 CPUs are designed to support advanced, agentic AI systems that require more than just raw processing power. Unlike traditional GPUs, these CPUs can handle complex tasks like real-time reasoning, multi-step problem solving, and managing large-scale AI models. They are built on AWS’s Nitro System, which ensures high performance, security, and availability.
Experts say that as AI systems become more persistent and autonomous, the CPU’s role becomes even more important. These processors serve as the control plane, orchestrating tasks, managing memory, and scheduling processes across different hardware accelerators. This shift means controlling the infrastructure and ensuring ample resources will be critical for AI developers.
Meta’s deployment of Graviton5 cores highlights a broader industry trend. As AI workloads grow in complexity and size, control and efficiency become key. CPUs like Graviton5 are stepping into a new role, supporting not just basic operations but enabling AI systems to operate more like autonomous agents, capable of handling multi-stage, stateful tasks in real time.
Overall, Meta’s aggressive hardware investments reflect its desire to stay at the forefront of AI development. By securing massive compute resources and working with multiple chipmakers, Meta aims to maintain control over its AI systems and accelerate innovation across its platforms.
What do you think?
It is nice to know your opinion. Leave a comment.