Meta revealed a roadmap of four new in-house chips — MTIA 300, 400, 450, and 500 — designed to handle AI inference across its platforms. MTIA 300 is already deployed powering ranking and recommendation systems, while MTIA 400 has completed testing and will enter data centers soon. The final two chips, MTIA 450 and 500, target generative AI tasks like image and video creation and are scheduled for mass deployment in 2027.
The company plans to release new chip generations every six months, far outpacing the typical one-to-two-year industry cycle. Meta achieves this speed through modular, reusable designs. The announcement came just weeks after Meta signed massive deals with Nvidia and AMD, signaling that custom silicon will complement — not replace — third-party hardware.
The move puts Meta alongside Google, Amazon, and Microsoft in the race to build proprietary AI accelerators. By shifting inference workloads to its own chips, Meta aims to reduce costs and latency for billions of daily AI-powered interactions across Facebook, Instagram, and WhatsApp.
