Meta to begin AI chip production in September as it ramps up computing capacity

Meta Platforms is preparing to begin manufacturing its latest in-house artificial intelligence chip in September, as per a Reuters report..According to an internal company memo reviewed by Reuters, the chip—code-named Iris—forms part of Meta's multi-generation Meta Training and Inference Accelerator (MTIA) programme, which is focused on developing custom silicon to power AI models used across Facebook, Instagram and other services.The memo indicates that testing of the chip was completed in just six weeks without major technical issues, suggesting the project has gathered momentum after years of development. Meta has been working to build its own AI processors for more than five years as demand for computing power continues to surge.Meta is collaborating with Broadcom on the chip's design, while manufacturing will be handled by Taiwan Semiconductor Manufacturing Co. (TSMC). By developing custom processors, the company aims to optimise AI workloads for its own platforms, improve efficiency and lower the long-term costs associated with purchasing large volumes of graphics processing units (GPUs) from suppliers such as Nvidia and AMD.Also read: Apple signs $30 bn Broadcom deal to expand US chip productionThe new chip is intended to complement, rather than replace, the GPUs Meta already deploys for AI training and inference. The internal memo notes that integrating the latest third-party GPUs across Meta's large-scale infrastructure has proven both time-consuming and operationally complex.Meta first introduced Iris under its technical MTIA branding earlier this year alongside three additional AI processors. The company now plans to release new AI chips approximately every six months through 2027, accelerating a product development cycle that typically spans a year or longer in the semiconductor industry.Alongside its chip roadmap, Meta is also expanding its AI infrastructure at an aggressive pace. The company expects to deploy seven gigawatts of computing capacity this year and plans to increase that figure to 14 gigawatts in 2027, according to the memo.To support the expansion, Meta has secured multi-year supply agreements with several hardware partners, including Samsung Electronics for memory chips, SanDisk for flash storage and Sumitomo Electric for fibre-optic equipment. The long-term contracts are aimed at ensuring stable component supplies amid growing demand for AI infrastructure and ongoing pressure on semiconductor supply chains.The infrastructure build-out is expected to contribute to Meta's broader AI spending plans, with the company projected to invest up to $145 billion in AI infrastructure this year.