Meta and Amazon Web Services (AWS) officially announced a massive expansion of their partnership on Friday, April 24, 2026. Meta has signed an agreement to deploy tens of millions of AWS Graviton CPU cores, making it one of the largest Graviton customers in the world.
This move signals a strategic shift in AI infrastructure, moving away from a singular focus on GPUs to a more diversified stack optimized for “Agentic AI.”
1. The Deal: Scale and Scope
The agreement is described as a multi-year, multi-billion-dollar commitment that embeds AWS Graviton silicon at the heart of Meta’s next-generation AI efforts.
- Initial Deployment: Meta is starting with tens of millions of Graviton cores, with the flexibility to scale further as demand grows.
- The Hardware: The partnership focuses on Graviton5, Amazon’s latest custom Arm-based chip. It features 192 cores and a cache 5x larger than its predecessor.
- Strategic Rationale: Meta’s Head of Infrastructure, Santosh Janardhan, stated that “diversifying compute sources is a strategic imperative” as the company scales infrastructure beyond its massive Nvidia GPU clusters and in-house MTIA designs.
2. Why Millions of CPUs? (The “Agentic AI” Shift)
While GPUs are the “muscle” used for training large models, the rise of Agentic AI—autonomous systems that reason, plan, and execute multi-step tasks—requires an immense amount of “nervous system” work that is best handled by CPUs.
- Orchestration Logic: Agents must manage complex workflows, such as calling external tools, managing long-term memory state, and real-time reasoning. These are CPU-intensive tasks.
- Reduced Latency: The 33% reduction in inter-core latency in Graviton5 is critical for “chatter-intensive” agent workflows where different layers of the AI must communicate instantly.
- Llama 4 Optimization: Analysts noted that Llama 4’s Mixture-of-Experts (MoE) architecture and its massive 10-million-token context window require sophisticated Key-Value cache management, which Graviton is purpose-built to handle.
3. Market and Investor Impact
The announcement had immediate ripples across the tech industry and financial markets:
- Amazon (AMZN): Shares rose 2% following the news. Amazon CEO Andy Jassy revealed that the demand for AWS silicon is so high that the company’s annual chip revenue run rate has topped $20 billion.
- The “CPU is Cool Again”: The deal validates a growing industry consensus (echoed by Intel and AMD) that as AI shifts from training to inference and orchestration, CPUs are regaining their central importance in the data center.
- Infrastructure Strategy: The move allows Meta to deploy workloads on AWS immediately while it prepares to bring similar Arm-based “AGI CPUs” (designed in collaboration with Arm) into its own data centers later in 2026.