ByteBrief
We're a portrait publication through and through. Turn your phone back and your briefing picks up right where you left it.
(We tried widescreen once. It wasn't us.)
Meta presented its MTIA custom AI silicon at Hot Chips 2026, detailing the MTIA 400 inference chip. The 400 offers 12 PFLOPs of FP4 performance, 9.4TB/second of HBM3e bandwidth, and scales to 72 ASICs per domain. Four generations are planned: MTIA 300, 400, 450, and 500.
Tap to vote and see what everyone thinks.
Summary by ByteBrief