🧠 Food for thought

Implications, context, and why it matters.

Hyundai builds robot AI with an in-house edge brain and DeepX

• The generative AI work builds on a three-year partnership between DeepX and Hyundai Motor Group’s Robotics Lab on low-power on-device AI for robots 1.
• Early prototypes ran on Nvidia’s Orin AGX platform, a computing module for AI systems, though Hyundai Motor Group’s Robotics Lab found it too expensive and power-hungry for wide deployment 2.
• DeepX says the lab then moved to a setup that pairs Intel central processing units (CPUs) with DeepX neural processing units (NPUs), chips that speed up AI workloads, to cut cost and power use 2.
• That work produced an on-device “edge brain” AI chip that uses under 5 watts. Hyundai Motor Group’s Robotics Lab has put it into technologies such as DAL-e Delivery, a delivery robot project now in demonstration and validation 1.

Low-power AI chips could lessen the role of GPUs in edge robotics

• The partnership suggests physical AI systems may rely less on general-purpose graphics processing units (GPUs), especially where power use and heat matter 3.
• Hyundai Motor Group says the co-developed chip prepares optimized solutions for future mass-produced robots while giving the supply chain more flexibility and stability 1.
• DeepX has outlined a partner model where it provides chips, while robot hardware makers, along with software developers, create platforms and applications 3.
• If more industrial companies choose specialized low-power chips from designers like DeepX, the edge AI market could become more split instead of centering on one hardware supplier 3.