AWS Inferentia2
Chip
AI Accelerator
Shipping
Amazon's inference chip — cheap tokens for production serving.
You're looking at a chip record: what this processor is, its key specs, who designs, manufactures and supplies its parts, and where it gets used.
Inferentia2 is the serving half of Annapurna Labs' silicon line: 32 GB of HBM per chip and a compiler stack (Neuron) that targets cost per inference rather than peak FLOPS. Inf2 instances gang up to 12 chips for large-model serving.
Specifications & key facts
Each figure is labeled by basis: vendor spec, vendor claim, reported, estimate, or official. "Not publicly disclosed" means exactly that — we never invent values.
Overview
Designer
AWS (Annapurna Labs)
GA
April 2023
Memory
Memory
32 GB HBM per chipVendor spec
Compute
FP16/BF16
190 TFLOPSVendor spec
Physical
Scale
up to 12 chips per Inf2 instance, NeuronLink interconnectVendor spec
How this connects to the rest of the stack
This record's links at a glance — click a group to jump to it.
Designed by
Available from
Related
Sources
Record reviewed as of 2026-01-15 — see Methodology.