Kneron nabs $40 mln

San Diego-based Kneron Inc, an on-device edge artificial intelligence company, has secured $40 million in funding.

San Diego-based Kneron Inc, an on-device edge artificial intelligence company, has secured $40 million in funding. The lead investor was Horizons Ventures.


San Diego, CA, January 23, 2020 – Kneron, Inc., a leading on-device edge artificial intelligence (AI) company based in San Diego, California, announces it has raised an additional $40M dollars in funding. Horizons Ventures, Hong Kong businessman Li Ka-Shing’s venture capital firm, is the lead investor of this round and has invested in previous rounds.

Kneron previously raised $33 million in funding from Horizons Ventures, Alibaba Entrepreneurs Fund, CDIB, Himax Technologies, Inc, Qualcomm, Thundersoft, and Sequoia Capital.

Kneron provides complete end-to-end integrated hardware and software solutions that enable on-device edge AI inferencing in mobile devices, personal computers, and IoT use cases including smart home devices, surveillance, payments, and smart cars. Their solutions augment cloud-based AI to accelerate AI inferencing on any device.

Albert Liu, Kneron’s founder and CEO, said “We are excited to continue our journey with partners like Horizons Ventures who share our passion and dedication towards our mission to enable AI on any device, democratize AI, and build the Edge AI Net.”

While many announcements in the AI space center on concepts and technologies that aren’t yet market-ready, Kneron’s on-device edge AI solutions are already being distributed to the market. “We are excited to see Kneron’s AI software and hardware solutions integrating into partner products that are shipping worldwide,” said Jonathan Tam from Horizons Ventures. “We expect to see a surge in demand for on-device AI compute going forward, and Kneron is perfectly positioned to equip these devices with highly efficient AI capabilities without sacrificing power.”

As the entire on-device Edge AI industry is still emerging, Kneron’s early investment and commercialization of its technology have positioned it in a leadership position to enable AI adoption in mass-market devices.
Kneron product offerings include:

· SoC AI chipsets – Kneron’s KL520 AI chip accelerates neural network models whether from Kneron or 3rd parties on mass-market devices enabling 2D/3D visual recognition and audio recognition applications in everyday devices
· Edge AI algorithms – Machine learning algorithms, which have among the smallest memory footprints in the industry according to recent NIST test results, include face detection, facial recognition, body detection, and gesture recognition
· Neural Processing Units (NPU) – Kneron’s NPU provides commercially proven, high-efficiency solutions designed for devices with low power, low thermal profiles and yet complex neural network computational requirements

Reconfigurable on-device edge AI in real-time
Their edge AI solutions are reconfigurable allowing for real-time switching between audio recognition and 2D/3D visual recognition depending on application needs on the device. Furthermore, this reconfigurability is compatible in real-time with major AI frameworks such as Tensor Flow, ONNX, Keras, Caffe, and PyTorch as well as major CNN models such as VGG16, ResNet, GoogleNet, YOLO, Tiny YOLO, LeNet, MobileNet, Densenet, and more.

Edge AI Net
Because of Kneron’s reconfigurable technology, they are uniquely positioned to realize their vision of the Edge AI Net, or AIoT 3.0. In short, the Edge AI Net will democratize AI and create more Wall-Es and EVAs with fewer Skynets. The Edge AI Net will allow Edge AI devices to communicate with each other to create collective actions that can be independent of centralized cloud-based AI services.

Balance of size, power, performance, and cost
Kneron also separates itself in the on-device edge AI space through solutions that balance power usage, memory footprint, and cost while still performing above their model “size” class as evidenced by their results at the NIST Facial Recognition Vendor test of 2019. This balance is critical when storage space, size, and power are limited in many use cases like security cameras, smart doorbells, smart door locks, smartphones, etc. In addition, its solutions are compatible with all major AI platforms and can reconfigure in near real-time to adapt to different application needs.