Meta Expands AI Infrastructure With AWS Graviton Deal for Agentic AI

TLDR:

• Meta will deploy tens of millions of AWS Graviton cores for expanding AI infrastructure demands
• The deal makes Meta one of the world’s largest AWS Graviton customers for compute capacity
• Agentic AI workloads require stronger CPU power for reasoning, search, and task execution
• Meta says diversified infrastructure helps match the right compute to the right AI workload

Meta Platforms will add tens of millions of AWS Graviton cores to its compute portfolio as it expands infrastructure for agentic AI systems. The deal makes Meta one of Amazon Web Services’ largest Graviton customers and deepens a long-running cloud partnership between both companies.

The first deployment starts with tens of millions of cores, with room for future expansion as Meta’s AI systems grow. The move reflects rising demand for CPU power as AI shifts toward reasoning, planning, and multi-step task execution.

Meta and AWS Expand AI Compute With Graviton Cores

Meta confirmed the agreement through its AI at Meta account and a detailed company blog post. The company said the deployment supports systems behind Meta AI and broader agentic AI products used by billions of people.

AWS Graviton5 processors will handle CPU-heavy workloads tied to these services. These include reasoning tasks, search functions, and systems that coordinate multi-step AI actions.

Meta said processing cores remain critical because they handle workloads that GPUs alone cannot manage efficiently. As AI products become more autonomous, CPU demand continues rising across production systems.

According to Meta, no single chip architecture can support every workload at scale. The company said this drives its broader strategy of combining internal infrastructure with external cloud partnerships.

Santosh Janardhan, Meta’s head of infrastructure, said AWS has remained a trusted cloud partner for years. He noted that Graviton helps Meta manage CPU-intensive workloads with better efficiency and performance.

Agentic AI Push Drives Meta’s Diversified Infrastructure Strategy

Meta said agentic AI requires different compute needs than earlier chatbot-style systems. These models must reason continuously, make decisions, and complete tasks across multiple steps without constant user input.

That shift increases demand for faster data handling and stronger bandwidth across CPU systems. Meta said AWS Graviton5 cores were designed for exactly those requirements.

Amazon also framed the deal as part of a broader AI infrastructure push. AWS Vice President Nafea Bshara said infrastructure must combine compute, data, and inference services to support global-scale AI systems.

He said purpose-built silicon works best when paired with the full AWS AI stack. Meta’s larger Graviton deployment, he added, shows how large-scale AI companies are changing infrastructure priorities.

Meta described the agreement as part of its portfolio approach to compute. The company continues investing in its own data centers while also using partners that offer specific hardware advantages.

Adding Graviton cores gives Meta more flexibility when assigning workloads across its infrastructure. The company said that balance helps it build faster while keeping performance and efficiency aligned with demand.

The first phase starts immediately, while future deployments will depend on how Meta AI and agentic products expand. For AWS, the agreement also strengthens Graviton’s position in large-scale enterprise AI infrastructure.

The post Meta Expands AI Infrastructure With AWS Graviton Deal for Agentic AI appeared first on Blockonomi.