Meta has confirmed plans to begin manufacturing its first internally designed AI chip, known as "Iris," in September 2026. The chip is part of the company's Meta Training and Inference Accelerator (MTIA) program and is designed specifically to power Meta's artificial intelligence systems more efficiently. The project represents one of the company's largest investments in custom semiconductor technology.
According to internal plans, Meta intends to dramatically expand its computing infrastructure, aiming to reach 14 gigawatts of AI computing capacity by 2027. The company has already deployed about 1 gigawatt of infrastructure during the first half of 2026 and expects to continue expanding rapidly to support increasingly advanced AI models and services.
Meta is developing the chip in partnership with Broadcom, while manufacturing will be handled by TSMC. The company believes that designing its own processors will lower long-term infrastructure costs, improve performance for its AI workloads, and reduce reliance on third-party GPU suppliers. Despite the move, Meta says it will continue using NVIDIA and AMD hardware alongside its own chips.
Technology analysts view the announcement as another major milestone in the global AI race. With companies such as Microsoft, Google, Amazon, Apple, and Meta investing hundreds of billions of dollars in AI infrastructure, custom chip development is becoming a key competitive advantage that will shape the future of artificial intelligence and cloud computing.
