Meta Platforms (NASDAQ:META) has signed a major agreement to deploy tens of millions of Amazon Web Services (AWS) Graviton processor cores, marking a significant strategic pivot in how the social media giant builds its next generation of artificial intelligence infrastructure.
The expansion of the long-standing partnership between Meta and AWS reflects an evolving architecture for AI.
While high-end graphics processing units (GPUs) remain the industry standard for training massive large language models, the industry is seeing a surge in demand for high-performance central processing units (CPUs) to handle "agentic AI."
These workloads include real-time reasoning, code generation, autonomous search, and the orchestration of complex, multi-step tasks that require high efficiency and low latency.
The deployment will center on AWS Graviton5, the latest generation of ARM-based processors purpose-built for cloud-native workloads.
Graviton5 provides the specific processing power required to coordinate the billions of interactions occurring across Meta’s platforms, particularly as the company moves toward more sophisticated AI agents that must manage multi-step workflows.
Industry analysts view the deal as a recognition of the shifting compute requirements for AI inference and orchestration.
As AI moves from static chat interfaces to proactive agents that can execute tasks and reason in real-time, the bottleneck often shifts from raw floating-point math to the general-purpose compute efficiency where Graviton excels.
Meta’s initial deployment of tens of millions of cores includes the flexibility to scale further as its AI initiatives, including the upcoming Llama 4 ecosystem, continue to expand.