Nieke Roos
26 November 2020

With its brain-inspired AI chip, TU Delft spinoff Innatera Nanosystems claims sensor data can be processed 100x faster and with up to 500x less energy than with conventional processors.

Innatera Nanosystems has raised 5 million euros in seed funding to bring its brain-inspired processing technology to sensors and sensor-based devices. The Delft University of Technology spinoff’s neuromorphic processing chip closely mimics the brain’s mechanisms for pattern recognition, enabling sensor data to be processed 100x faster and with up to 500x less energy than with conventional processors. These efficiency and performance gains allow advanced AI to be embedded into the sensor edge, unlocking a wide gamut of applications including intelligent speech processing in human-machine interfaces, vitals monitoring in wearable devices, target recognition in radars and lidars, and fault detection in industrial and automotive equipment.

According to Innatera, its solution is radically different from the traditional AI chips being proposed by its competitors and fundamentally changes how sensor data is processed. The technology relies on a new breed of analog/mixed-signal computing circuits that recreate the behavior of the brain’s fundamental building blocks – spiking neurons and synapses. Neural networks built with spiking neurons possess a precise notion of time, which enables them to be 10-100x more compact than conventional artificial neural networks, especially for applications involving data with high spatial and temporal correlations. As a result of this approach, Innatera’s architecture delivers a combination of ultra-low-power and ultra-short recognition latency, with up to 10,000x higher performance per watt than typical digital processors and conventional AI accelerators.

The edge of the edge

“The most impactful sensor-driven applications today are limited by the efficiency and speed of the processor, and this is more so in small, battery-powered devices than anywhere else,” explains Innatera CEO Sumeet Kumar. “We’re reinventing processing for sensors by combining the energy efficiency of analog/mixed-signal neuromorphic silicon with the performance gains of true spiking neural network algorithms, in a single integrated compute solution.” Innatera says it has been working together with several big international names on applications that it suggests are game-changers. The company expects these developments to surface in the consumer, industrial and automotive markets in the next few years.

Innatera team
Innatera’s management team (clockwise from left): Amir Zjajo (CSO), René van Leuken (Chief Advisor), Sumeet Kumar (CEO), Uma Mahesh (COO). Credit: Innatera

The €5M seed investment round was led by Munich-based deep-tech investors MIG Verwaltungs AG and the Industrial Technologies Fund of BTOV. According to Christian Reitberger, partner at BTOV, “Innatera sets itself apart from the plethora of AI accelerator companies by focussing on the edge of the edge – sensory data processing in the field. Translating truly brain-inspired design principles into state-of-the-art analog/mixed-signal solutions enables a performance envelope not accessible to more conventional solutions”. Sören Hein, partner at MIG adds: “We were particularly impressed by the team, which combines deep academic credentials with practical industry experience at leading semiconductor companies. They realized early on the fundamental importance of algorithms and software to unlock the market potential of SNN chips.”


Device lifecycle management for fleets of IoT devices

Microchip gives insight on device management, what exactly is it, how to implement it and how to roll over the device management during the roll out phase when the products are in the field. Read more. .

Customer commitments

Incorporated in 2018 as a university spinoff, Innatera builds on over a decade of research into computational neuroscience and low-power processing. It currently has about 15 employees at its HQ in Delft and is in the process of setting up a design center in Bangalore, India. The company is developing a suite of proprietary algorithms and a software toolchain to realize the full potential of its neuromorphic silicon. The investment will enable it to scale up its R&D efforts and accelerate product development to deliver on customer commitments through 2021.

One hundred kilometers to the south, Grai Matter Labs is working on a similar brain-inspired chip, with a similar ambition of bringing the fastest AI per watt for sensor analytics and machine learning to every device on the edge. Recently, this fabless semiconductor scale-up with Eindhoven roots got a $14M boost.