Popular Edge & IoT AI tools

14 category leaders in Edge & IoT, selected from the full directory.

Edge Impulse (Qualcomm)

  • End-to-end edge ML: data, training, optimization, deployment.
  • MCU to gateway targets; Qualcomm-owned since 2023.
  • Partners include Nordic, ST, Arduino, AWS, NVIDIA.
  • Standard toolchain for industrial and consumer IoT AI.

NVIDIA Jetson

  • Default edge AI modules from Orin Nano to Thor.
  • JetPack stacks CUDA, TensorRT, and robotics tooling.
  • Powers robots, drones, and on-device generative AI.
  • Largest developer ecosystem in embedded AI compute.

Coral

  • Google Edge TPU hardware and tooling for on-device ML.
  • About 4 TOPS at roughly 2W for USB, M.2, and SoM form factors.
  • TensorFlow Lite / LiteRT deployment path for makers and OEMs.
  • Architecture also appears in Synaptics Astra IoT processors.

Qualcomm AI Hub

  • On-device model hub for Snapdragon and IoT silicon.
  • Optimized models, conversion, and profiling for edge apps.
  • Covers vision, audio, and generative inference on-device.
  • Default path for Qualcomm-powered phones and gateways.

Ambarella

  • CV SoCs combining ISP and neural acceleration for cameras.
  • Targets security, automotive, and robotics edge vision.
  • Low-power multi-camera perception without a discrete GPU.
  • Longstanding OEM footprint in AI camera platforms.

LiteRT

  • Google on-device ML runtime succeeding TensorFlow Lite.
  • Runs optimized models on phones, MCUs, and edge silicon.
  • Conversion and delegation tooling for production deploy.
  • Default mobile/edge path across Android and Coral stacks.

ExecuTorch

  • PyTorch-native runtime for on-device inference.
  • Deploys models to mobile, MCU, and edge silicon.
  • Official edge path for the PyTorch ecosystem.
  • Pairs with LiteRT as a primary on-device stack.

Arm Ethos

  • Arm NPU IP for MCUs and edge SoCs.
  • Ethos-U for microcontrollers; Ethos-N for richer chips.
  • Licensed into mainstream IoT silicon vendors.
  • Foundational IP for milliwatt on-device inference.

Hailo

  • Edge AI accelerators for high-throughput vision inference.
  • Hailo-8 and modules for cameras, gateways, and PCs.
  • Strong TOPS-per-watt for always-on edge CV.
  • Widely designed into industrial and smart-city systems.

Axelera

  • Metis edge AI accelerators for high-throughput vision.
  • In-memory compute aiming at multi-camera inference nodes.
  • Strong performance-per-watt vs discrete edge GPUs.
  • Leading independent edge AIPU vendor alongside Hailo.

SiMa

  • MLSoC Modalix for physical AI under ~10W.
  • On-device CNNs, transformers, and multimodal models.
  • SoMs and LLiMa stack for industrial and robotics edge.
  • Top-tier edge SoC challenger to Jetson-class platforms.

NVIDIA DeepStream

  • Multi-camera edge vision pipelines on Jetson and dGPU.
  • Decode, infer, track, and overlay in one SDK stack.
  • Default NVIDIA path for smart-city and retail analytics.
  • Pairs with Jetson and IGX for production edge vision.

Luxonis

  • OAK AI depth cameras with on-device neural inference.
  • Spatial AI pipelines for robotics and industrial vision.
  • Wide adopter base from hobbyists to factory deployments.
  • Go-to edge vision camera brand for embedded AI teams.

Ambiq

  • Ultra-low-power SoCs for battery-powered edge AI.
  • Apollo and Atomiq chips for wearables and IoT sensing.
  • SPOT tech enables always-on ML on tiny energy budgets.
  • Hundreds of millions of devices shipped for edge AI.