Meta is deploying its custom MTIA chips to handle inference tasks for recommendation systems, ad optimization, and generative AI. The company reports that these ASICs provide a 44% reduction in total cost of ownership and 40% better power efficiency compared to general-purpose GPUs.
The roadmap includes four generations of chips: MTIA 300, 400, 450, and 500. The MTIA 300 is currently in production, while the MTIA 400, codenamed Iris, is scheduled for production in September 2026. Meta aims to release a new generation approximately every six months using a modular chiplet design.
The development program involves collaboration with Broadcom and manufacturing by TSMC. While Meta continues to partner with Nvidia and AMD for model training, the custom silicon is intended to optimize high-volume inference workloads.
Meta projects capital expenditures for 2026 to be between $115 billion and $135 billion. Additionally, the company plans to scale its computing capacity to 14 gigawatts by 2027.
