I’ve said that autonomous vehicles (AVs) will be on public roadways sooner than we all think, but not everywhere we would want. Sensors, of course, are critical to this autonomous vehicle future. But, with those sensors come huge processing and power demands. To address these challenges, I’m excited to announce our investment in Recogni — a startup that’s building low-power, high-performance AI processing systems.
Depth cameras, LiDAR and radar sensors generate terabytes of data per hour. All this data pushes the power budget of AVs into the kilowatts. Recogni’s inference module processes sensor perception data much more efficiently than similar systems (details mum, for now). This decrease in power consumption helps tomorrow’s AVs more efficiently reach their goal of identifying and reacting to other vehicles, pedestrians, and obstacles on the road in real-time.
What impressed us most about Recogni is how significantly its hardware increases the total computing capability of AV systems. Recogni’s chip architecture can provide 1000 Tera Operations Per Second (TOPS) of performance for every five watts of power — far beyond comparable AI accelerator solutions. By targeting both power consumption and performance simultaneously, Recogni is poised to provide cloud-class processing power to AVs at the edge.
Recogni was founded by RK Anand, Ashwini Choudhary, Valerie Chan, Giles Backhus, and Eugene Feinberg — all serial entrepreneurs with backgrounds in networking and mobile hardware. They’ve assembled a strong team of AI, computer vision, and distributed systems professionals to create their proprietary chip technology for AV processing.
We’re thrilled to welcome Recogni to the Toyota AI Ventures portfolio and participate in their recent $25M Series A funding round alongside GreatPoint Ventures, BMW i Ventures, Faurecia, and DNS Capital. For more information about how Recogni is designing a vision-oriented AI processing system, visit their website.