Meta’s custom-built Iris AI chip is expected to strengthen its AI infrastructure while reducing reliance on Nvidia and AMD processors.
Meta Platforms plans to begin manufacturing its own artificial intelligence (AI) chip in September as part of a broader strategy to expand its AI infrastructure and reduce reliance on third-party chip suppliers, according to an internal memo reviewed by Reuters.
The custom-designed data center chip, code-named Iris, is part of Meta’s Meta Training and Inference Accelerator (MTIA) program, a multi-generation initiative aimed at powering AI models across the company’s platforms, including Facebook and Instagram.
According to the memo, the chip successfully completed testing in just six weeks, with engineers reporting no major issues. The rapid testing process marks a significant milestone for Meta’s in-house semiconductor project, which has faced delays and technical challenges since it was launched more than five years ago.
Meta is developing the AI chip in collaboration with Broadcom, which is assisting with the chip’s design, while Taiwan Semiconductor Manufacturing Company will manufacture the processor.
The company believes custom-built silicon will improve the performance and efficiency of its AI systems while helping reduce the substantial costs associated with purchasing high-end processors from suppliers such as Nvidia and Advanced Micro Devices.
Rather than replacing graphics processing units (GPUs), the new Iris chip is expected to complement the large number of AI processors Meta already acquires from Nvidia and AMD, strengthening the company’s growing AI infrastructure.
The reported production timeline and successful completion of bug testing have not previously been disclosed. Meta declined to comment on the internal memo.
The move underscores Meta’s long-term investment in AI hardware as competition intensifies among major technology companies racing to develop faster, more efficient AI computing platforms.
