Imagine a chip that works like a human brain, only waking up when it notices something important. Instead of sending every bit of data to a distant cloud server, it processes information right where the sensor is. This means devices can stay 'awake' and smart without draining their batteries in a few hours.
Battery-powered devices struggle to process continuous sensor data without draining power or relying on expensive, slow cloud uploads. Current hardware cannot balance the contradictory needs of high AI performance, small size, and low cost.
A silicon prototype of a Spiking Neural Processor featuring an analog-mixed signal computing fabric for near-sensor AI processing.
If you are a wearable device manufacturer dealing with short battery life in smartwatches — this project developed a spiking neural processor that enables always-on pattern recognition within an ultra-narrow power envelope.
If you are a smart home security provider dealing with high data costs and latency in video doorbells — this project developed a silicon prototype that eliminates the need to send over 95% of irrelevant captured data to the cloud.
If you are a wireless earbud developer dealing with extreme size and power constraints — this project developed a compact analog-mixed signal computing fabric that improves integration density and power dissipation.
Based on available project data, specific pricing is not disclosed, but the project focuses on reducing cost as one of the three primary constraints for small consumer devices.
The project focused on optimizing the computational fabric for manufacturability and delivered a silicon prototype, suggesting a path toward industrial production.
The project involved the development of new IP to enable novel functional capabilities and energy efficiency gains within the processor architecture.
It is designed as a near-sensor processing architecture that allows sensors to intelligently make sense of data locally before transmission.
The project was executed between May 2023 and April 2025, concluding with the validation of a silicon prototype.
The project is led by a single partner, Innatera Nanosystems BV, a Dutch SME and semiconductor spin-off from TU Delft. With a 100% industry ratio and no university or research partners in the consortium, the project is streamlined for commercial execution and rapid prototyping of the silicon hardware.
Contact Innatera Nanosystems BV regarding their Spiking Neural Processor silicon prototype.
Contact SciTransfer for a detailed technical deep-dive into neuromorphic processor integration.
