Live
AI hardware

Meta MTIA 300

FP16 throughput
6×10¹⁴ FLOP/s
Memory
201 GB
TDP
800 W

Meta MTIA 300 is an AI accelerator designed by Meta (Other), released in March 2026.

It delivers 6×10¹⁴ FLOP/s of tensor FP16/BF16 throughput, 1.2×10¹⁵ FLOP/s at FP8. It carries 201 GB of memory at 5.5 TB/s of bandwidth. Its thermal design power is 800 W.

Full record
Manufacturer
Meta
Type
Other
Release date
Mar 11, 2026
Tensor FP16/BF16
6×10¹⁴ FLOP/s
FP8
1.2×10¹⁵ FLOP/s
Memory
201 GB
Memory bandwidth
5.5 TB
TDP
800 W
Foundry
TSMC
More from Meta
SourceEpoch AI, 'Machine Learning Hardware'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/data/machine-learning-hardware. Licensed under CC BY 4.0.
← All ai hardware