Meta Platforms announced on April 24, 2026, a multi-year, multi-billion dollar agreement with Amazon Web Services (AWS) to integrate Graviton5 central processing unit (CPU) chips into its global infrastructure. The deal represents a significant expansion of the existing relationship between the social media giant and the cloud computing leader. Under the terms of the agreement, Meta will deploy Graviton5-based instances to support a variety of workloads, specifically focusing on the computational demands of its artificial intelligence initiatives, including the training and inference of its Llama large language models.
The move is part of Meta’s broader strategy to diversify its compute sources. While Meta continues to invest in its own custom silicon, such as the Meta Training and Inference Accelerator (MTIA), and maintains a heavy reliance on third-party GPU providers, the adoption of AWS Graviton5 chips provides a specialized ARM-based alternative for general-purpose and AI-adjacent tasks. Meta Chief Financial Officer Susan Li stated that the partnership allows the company to optimize performance-per-watt across its fleet, which is critical for managing the escalating energy costs associated with AI development.
AWS Graviton5, the latest generation of Amazon’s custom-designed ARM-based processors, was designed to offer enhanced performance for compute-intensive applications. According to technical specifications released by AWS, Graviton5 provides up to 30% better performance and 40% better energy efficiency compared to its predecessor, Graviton4. For Meta, these chips will be utilized to power recommendation engines, ad ranking systems, and the backend infrastructure for its family of apps, including Facebook, Instagram, and WhatsApp. AWS CEO Matt Garman noted that the collaboration underscores the growing demand for custom silicon tailored to the specific requirements of hyperscale cloud users.
The financial commitment, valued in the billions of dollars over several years, highlights the scale of Meta’s infrastructure spending. In its most recent quarterly earnings report, Meta projected capital expenditures for 2026 to remain elevated as it builds out the physical hardware necessary for its AI-centric roadmap. By leveraging AWS’s managed infrastructure and Graviton5 chips, Meta aims to reduce the latency of its AI services while maintaining a flexible supply chain. The deployment is expected to begin in the third quarter of 2026, with a phased rollout across AWS regions globally.
This agreement marks one of the largest external silicon procurement deals in Meta’s history. Official statements from both companies emphasized the collaborative nature of the deal, which includes joint engineering efforts to optimize Meta’s software stack for the Graviton architecture. As Meta continues to integrate generative AI features across its platforms, the partnership with AWS provides the necessary scale to meet increasing user demand without relying solely on a single hardware architecture.