The company’s in-house AI chip push aims to cut dependence on Nvidia and support rapid data centre expansion.
Meta Platforms plans to start manufacturing its in-house artificial intelligence chip, code-named Iris, in September as part of a wider strategy to double its computing capacity to 14 gigawatts next year, according to an internal memo reviewed by Reuters.
The custom-built chip is part of Meta’s four-generation Meta Training and Inference Accelerator (MTIA) programme and is designed to improve the AI systems powering Facebook and Instagram. The company completed testing in just six weeks without major issues, marking significant progress for a project that has faced delays for years.
Meta is working with Broadcom on chip design and Taiwan Semiconductor Manufacturing Co. (TSMC) for production. The move is expected to reduce the company’s dependence on suppliers such as Nvidia and AMD, while lowering the soaring costs of AI computing infrastructure.
This year, Meta expects to deploy seven gigawatts of computing infrastructure before doubling that capacity in 2027. The company also plans to invest up to US$145 billion in AI infrastructure this year and has secured long-term supply agreements with Samsung Electronics, Sandisk and Sumitomo Electric for memory chips, flash storage and fibre-optic equipment.
“You can’t become an AI titan if you are dependent on another company for chips,” said Mike Gualtieri, vice president and principal analyst at Forrester.
Meta plans to introduce a new AI chip roughly every six months through 2027 as demand for AI infrastructure continues to increase.
