Pulsar - Innatera

Innatera's Pulsar is the first commercially available, brain-inspired microcontroller for sensing at the edge, offering real-time intelligence at microwatt power levels.

At A Glance

Release Year: 2025
Status: Released
Chip Type: Mixed-signal
Software: Talamo SDK
Applications: Wearables, Hearables, Smart Home, IoT, Industrial Monitoring
Neurons: N/A
Synapses: N/A
Power: < 1mW (microwatt levels)

Pulsar is a brain-inspired neuromorphic microcontroller designed for ultra-low-power, real-time intelligence in edge devices. It combines a Spiking Neural Network (SNN) engine with a RISC-V MCU and CNN acceleration to enable smart sensing applications without cloud dependency.

Overview

Pulsar is Innatera’s revolutionary brain-inspired microcontroller, engineered to deliver real-time, event-driven intelligence for a new generation of smart devices. As the world’s first commercially available neuromorphic processor for the mass market, Pulsar is designed to process sensor data with ultra-low power consumption, enabling “always-on” capabilities in wearables, IoT devices, and industrial systems without relying on the cloud.

The core of Pulsar’s innovation lies in its Spiking Neural Network (SNN) engine. Unlike traditional AI that continuously processes dense data, SNNs operate on event-driven spikes, activating compute power only when new information is available. This drastically reduces energy consumption to microwatt levels while achieving sub-millisecond inference speeds.

Architecture

Pulsar features a hybrid architecture that provides both efficiency and flexibility:

  • Spiking Neural Network (SNN) Engine: The primary engine for processing sensor data like motion, audio, and vibration with maximum power efficiency.
  • RISC-V MCU: A conventional microcontroller unit for general-purpose tasks and control.
  • CNN Acceleration: Hardware support for Convolutional Neural Networks, offering versatility across different types of AI workloads.

This combination allows Pulsar to function as the sole microcontroller a sensor needs, integrating neuromorphic intelligence directly at the edge.

Software and Development

Development for Pulsar is streamlined through the Talamo SDK. This software development kit allows developers to:

  • Create Spiking Neural Network models.
  • Easily port existing models from popular frameworks like TensorFlow and PyTorch.
  • Develop applications without requiring deep expertise in neuromorphic engineering.

This approach lowers the barrier to entry for building brain-inspired AI solutions.

Applications

Pulsar’s unique capabilities make it ideal for a wide range of power-constrained and latency-critical applications:

  • Wearables & Hearables: Enabling always-on health monitoring, gesture recognition, and fall detection.
  • Smart Home & IoT: Providing real-time event detection and environmental awareness without cloud reliance, enhancing privacy and responsiveness.
  • Industrial Applications: Facilitating predictive maintenance and anomaly detection through ultra-efficient, continuous monitoring of machinery and systems.

Help Us Improve this Guide

Our hardware guide is community-maintained. If you know of a chip we should add, see an error, or have updated information, please let us know by opening an issue on our GitHub repository.

Open Neuromorphic Evolves: Announcing Our Charter and First Executive Committee Election

Open Neuromorphic Evolves: Announcing Our Charter and First Executive Committee Election

Open Neuromorphic is introducing an organizational charter and holding its first Executive Committee elections to foster growth and collaboration. Learn more and get involved!

NorthPole, IBM's latest Neuromorphic AI Hardware

NorthPole, IBM's latest Neuromorphic AI Hardware

A deep dive into IBM's NorthPole, a brain-inspired AI accelerator. Understand its architecture, 10 core axioms, and how it achieves groundbreaking energy efficiency for neural inference.

Efficient Compression for Event-Based Data in Neuromorphic Applications

Efficient Compression for Event-Based Data in Neuromorphic Applications

Discover methods to efficiently encode and store event-based data from high-resolution event cameras, striking a balance between file size and fast retrieval for spiking neural network training.