Description The Coral M.2 Accelerator A+E Key adds a Google Edge TPU coprocessor to a compatible host through a compact M.2 card slot. It delivers up to 4 TOPS of int8 machine-learning performance at 2 TOPS per watt, enabling low-latency TensorFlow Lite inference without relying on continuous cloud connectivity. The 22 x 30 mm module is designed for embedded computers, mini PCs and industrial systems that expose a compatible PCIe lane through an A- or E-keyed M.2 slot. On-device processing can reduce response time, limit data transfers and keep sensitive input close to the application. Features Google Edge TPU coprocessor for local machine-learning inference 4 TOPS peak performance for int8 workloads Power efficiency of 2 TOPS per watt M.2 A+E Key form factor with PCIe Gen2 x1 interface Integrated power management Supports TensorFlow Lite models compiled for the Edge TPU Compatible with supported Debian-based Linux and Windows systems Specifications ML accelerator Google Edge TPU coprocessor Peak performance 4 TOPS (int8) Power efficiency 2 TOPS per watt Hardware interface M.2 A+E Key, M.2-2230-A-E-S3 Serial interface PCIe Gen2 x1 Supply voltage 3.3 V ±10% Dimensions 22 x 30 x 2.35 mm Weight 3.1 g Operating temperature -20 to +85 °C Storage temperature -40 to +85 °C Relative humidity 0 to 90%, non-condensing Applications Local image classification and object detection Embedded vision and edge AI systems Industrial gateways and monitoring equipment Low-latency inference in mini PCs and compact computers Privacy-focused processing without continuous cloud access Product List 1 x Coral M.2 Accelerator A+E Key Other Content Host compatibility: An A- or E-keyed connector does not by itself guarantee compatibility. The host slot must provide a PCIe Gen2 x1 connection, and the system must support MSI-X. Check the host documentation before installation. Setup: Shut down the host before inserting the module. The Coral PCIe driver, Edge TPU runtime and a supported TensorFlow Lite or PyCoral environment are required to run inference. Thermal and handling guidance: Use suitable system-level thermal management to keep the Edge TPU below its rated operating temperature. The PCIe driver supports temperature monitoring, dynamic frequency scaling and thermal shutdown controls. Handle the module in a static-safe environment. Resources Coral M.2 Accelerator Datasheet PDF Coral M.2 and Mini PCIe Accelerator Setup Guide Coral PCIe Module Temperature Management Guide Edge TPU Inferencing Overview TensorFlow Lite Model Compatibility Guide FAQ Will this accelerator work in every M.2 A- or E-keyed slot? No. The host slot must expose a compatible PCIe Gen2 x1 connection, and the system must support MSI-X. Confirm these requirements in the host documentation. Is this module an M.2 storage device? No. It is a PCIe machine-learning accelerator and does not provide storage. Which models can run on the Edge TPU? It runs compatible TensorFlow Lite models that have been compiled for the Edge TPU. Does the module require cooling? The thermal solution depends on the host enclosure and workload. The system should be designed to keep the Edge TPU within its rated temperature range, with driver-based monitoring and throttling available where needed.
