Meta MTIA 400
Meta chip that handles both AI training and ads-ranking inference
N/A
Datacenter
288GB
HBM3E
1150
GB/s bandwidth
1200W
TDP
3nm
process
Meta detailed the MTIA 400 at Hot Chips 2026, built on a 3nm process with 288GB of HBM3E and a 1200W TDP. Unlike earlier MTIA chips split into separate training and inference lines, the 400 is built to run both jobs on the same silicon, reaching about 12 petaflops of MXFP4 compute and 6 petaflops at FP8. Meta says it plans to ship a new MTIA generation roughly every six months going forward.
// similar hardware in Datacenter
NVIDIA B200
NVIDIA Blackwell flagship datacenter GPU for frontier model training
NVIDIA · Datacenter · N/A
NVIDIA H100 SXM
Hopper-architecture datacenter GPU for large-scale AI training
NVIDIA · Datacenter · ~$30000
NVIDIA H200
High-memory Hopper refresh for large-model inference
NVIDIA · Datacenter · ~$32000
AMD Instinct MI300X
AMD flagship datacenter accelerator for large-model AI training
AMD · Datacenter · ~$15000
AMD Instinct MI325X
High-memory CDNA 3 refresh for large-model inference
AMD · Datacenter · ~$16000
NVIDIA A100 80GB
Ampere-generation datacenter GPU, still widely deployed
NVIDIA · Datacenter · ~$15000