Pulsar Spiking Neural Processor for Ambient Intelligence
The first mass-market, neuromorphic microcontroller for the sensor edge
Innatera Pulsar is the first neuromorphic microcontroller built for real-time intelligence at the sensor edge. Delivering brain-like efficiency in a micro-watt power envelope, it enables always-on, responsive devices across wearables, IoT, and industrial systems. Combined with the Talamo SDK, Pulsar brings scalable, event-driven AI to developers and sensor makers – powering a new generation of smarter, self-contained products.
Benefit 01
Always-on intelligence without draining battery
Pulsar brings real-time, event-driven intelligence directly to sensor-powered devices, enabling sub-millisecond responsiveness at microwatt power levels. This means application processors stay asleep until truly needed. Enabling a much greater battery life for features like human detection, gesture recognition, and smart environmental sensing without cloud dependence.
Benefit 02
Smarter sensors for the edge
Pulsar empowers sensor makers to deliver smarter, self-contained modules. By embedding neuromorphic intelligence directly at the edge, Pulsar enables faster development cycles, lower component costs, and smarter, event-driven sensors that are ready for integration across applications.
Benefit 03
Development made easy with Talamo SDK
Building brain-inspired AI doesn’t mean starting from scratch. Pulsar’s Talamo SDK allows developers to create spiking neural network models or port existing TensorFlow and PyTorch models easily. This is a complete neuromorphic toolchain – from model design and training to hardware deployment. No need for a PHD or new training.
Build your model journey
with Talamo
Pulsar brings real-time, event-driven intelligence directly to your devices,
enabling sub-millisecond responsiveness at microwatt power levels.
Specifications
At the heart of Pulsar lies a brain-inspired Spiking Neural Network (SNN) engine, designed to run models quickly and efficiently. This accelerator executes neural operations in a massively parallel fashion using an array of analog mixed-signal processing elements, each one mimicking the behavior of biological neurons. Every time these elements fire, they do so with minimal latency and extremely low energy, enabling real-time inference directly on sensor data.
This neuromorphic core is part of a heterogeneous compute architecture that brings together three complementary fabrics. The SNN accelerator performs event-driven computation for temporal data streams. A CNN accelerator provides a traditional path for spatial pattern recognition tasks, giving developers flexibility to deploy the best approach for each problem. Finally, an integrated RISC-V CPU subsystem manages all system control and housekeeping around the neural engines, creating a unified platform for end-to-end sensor data processing within a single, compact device.
Compute
Heterogeneous neuromorphic architecture with SNN, RISC-V, CNN, and FFT accelerators
01
Memory
Embedded low-latency SRAM optimized for real-time sensor processing
02
Package
2.8 × 2.6 mm WLCSP, industrial range –40 °C to 125 °C
03
Real-World use cases
Consumer electronics
- Speech/audio recognition
- Audio processing
- Human presence detection & recognition
- Gesture recognition
Sensor inputs: camera, microphone, radar
IoT & Smart Home
- Presence and motion sensing
- Ambient audio and anomaly detection
- Ultra-low-power, always-on operation
Sensor inputs: camera, radar, infrared
Industrial IoT
- Predictive maintenance
- On-board perception for autonomous operation
Sensor inputs: IMU, microphone
Wearables
- ECG pattern analysis
- IMU-based motion analysis
- Fall and anomaly detection
Sensor inputs: ECG, PPG, EMG
Comparison & Related
Products
State-of-the-art
Pulsar
Audio scene classification
Over 100x lower
Energy per inference
Over 33x lower
Model size
Sound recognition (e.g. KWS)
33x lower
Energy per inference
1.4x shorter
Inference latency
4x smaller
Model size
Radar gesture recognition
42x lower
Energy per inference
177x shorter
Inference latency
30x smaller
Model size
Our executive team.
Innatera’s Pulsar chip has the potential to redefine what’s possible at the edge. By using brain-inspired Spiking Neural Networks, it brings real-time processing to ultra-low-power devices without leaning on the cloud.
David Harold
senior analyst, Jon Peddie Research
Innatera is the first to bring neuromorphic microcontrollers to the mass market. We combine deep scientific heritage with proven industry collaborations, from radar sensing to smart safety systems.
Matthias Neumann
Senior Marketing Manager
Innatera is the first to bring neuromorphic microcontrollers to the mass market. We combine deep scientific heritage with proven industry collaborations, from radar sensing to smart safety systems.
John smith
safety systems.
Innatera is the first to bring neuromorphic microcontrollers to the mass market. We combine deep scientific heritage with proven industry collaborations, from radar sensing to smart safety systems.
John smith
safety systems.
Innatera is the first to bring neuromorphic microcontrollers to the mass market. We combine deep scientific heritage with proven industry collaborations, from radar sensing to smart safety systems.
John smith
safety systems.
Awards & Recognitions:
Embedded World Startup Award
AI Hardware Summit
SSCC 2025
Computex
Integrate your Pulsar chip today
Discover how Pulsar’s neuromorphic architecture can transform your next-generation devices with real-time intelligence and ultra-efficient performance at the edge.
