META plans to begin manufacturing its proprietary Iris AI chip in September, aiming to double its compute capacity from 7GW in 2026 to 14GW by 2027. This initiative is part of a broader trend among technology companies, which are developing in-house AI chips to enhance performance and secure supply chains for advanced computing needs. Additionally, META has established long-term supply agreements for memory, flash storage, and fiber optics, anticipating up to $145 billion in investment for AI infrastructure this year.

Meta: Meta Platforms develops and operates leading social media and connectivity services while advancing artificial intelligence technologies across its products. The company is preparing to manufacture its own Iris AI chip in-house starting in September to support internal AI workloads. This move forms part of ongoing efforts to build and control key elements of its computing infrastructure.
Iris AI chip: The Iris AI chip is Meta’s custom-designed semiconductor optimized for accelerating artificial intelligence training and inference tasks. It is scheduled to enter manufacturing in September as part of the company’s strategy for scaling its data center capabilities. The chip will be deployed internally to power Meta’s expanding AI applications.

AI Hardware: Technology companies are developing proprietary AI chips to gain greater control over performance and supply chains for advanced computing needs.
Infrastructure: Firms are pursuing long-term agreements for memory, storage, and optical components to support large-scale AI data center expansion.