Meta· Llama / infrastructureInfrastructure

MTIA 300: Meta’s First Training Chip with Built-in NICs and Communication-Offloading Engines

Meta announces MTIA 300, its first training chip with built-in NIC chiplets and communication-offloading engines, designed to optimize communication for training recommendation models, achieving 3.9x faster communication than equivalent GPU clusters on a 150-billion-parameter production model.

From the source

MTIA 300 is the first of Meta’s family of in-house training and inference accelerators optimized for training ranking and recommendation models.

engineering.fb.com

Announced

24 Aug 2026 17:45 UTC

Detected

24 Aug 2026 17:46 UTC (+49s after)