MX3 is a production-proven AI accelerator built on a pipelined dataflow architecture, designed to run inference for camera and video workloads with low batch-one latency and small on-chip weight storage.
It integrates through PCIe and USB interfaces and ships both as silicon and inside the vendor's M.2 accelerator modules, positioning it as an inference companion to an existing host processor.

MX3 is a production AI accelerator chip that runs inference for edge vision and video systems.
Ara-1 is the first-generation Ara discrete neural processing unit for edge AI vision inference.

Ara240 is the second-generation Ara discrete neural processing unit for generative AI inference at the edge.


Kneron's Edge AI Platform delivers on-device NPUs and full-stack software for edge inference.

Ara-1 is the first-generation Ara discrete neural processing unit for edge AI vision inference.
Ara240 is the second-generation Ara discrete neural processing unit for generative AI inference at the edge.
Kneron's Edge AI Platform delivers on-device NPUs and full-stack software for edge inference.

Hailo-10H is an edge AI accelerator for generative AI inference on devices rather than in the cloud.



Analog compute-in-memory processor delivering 25 TOPS for edge AI inference.



Axera's Smart Vehicle Chips deliver AI SoCs for advanced driver-assistance and intelligent cockpit systems.


Axera's Smart Vision Chips power AI-enabled cameras and visual perception devices.
Analog compute-in-memory processor delivering 25 TOPS for edge AI inference.

Low power machine learning system on chip for physical AI devices.

Axera's Smart Vehicle Chips deliver AI SoCs for advanced driver-assistance and intelligent cockpit systems.

Axera's Smart Vision Chips power AI-enabled cameras and visual perception devices.
