Honda and Mythic have announced a joint development agreement in which Honda R&D will license Mythic’s Analog Processing Unit (APU) technology and the companies will co-develop an automotive-grade AI SOC for deployment in Honda’s next-generation software-defined vehicles (SDVs) by the late 2020s/early 2030s.
In line with Honda’s safety approach, Mythic’s ultra-efficient analog compute-in-memory architecture will enable advanced driver-assist and autonomous features. The goal is to boost on-board AI capabilities while minimizing power draw – a critical factor as Honda pursues its global goal of zero traffic collision fatalities involving Honda motorcycles and automobiles by 2050.
Mythic said its analog compute-in-memory technology delivers about 100x better energy efficiency than conventional digital AI chips by combining memory and computation into one layer. This efficiency advantage translates into roughly 100x more AI processing within the same power budget. The joint development envisions future vehicles with more than 100,000 TOPS (trillions of operations per second) of AI compute.
“Cars are quickly becoming petascale supercomputers on wheels – in the near future, the most powerful computer in your home will be parked in your garage,” said Mythic CTO Dave Fick in a statement. “Vehicles will soon require computing performance on par with data centers, but with far tighter energy budgets. That’s exactly what Mythic’s analog technology delivers. We’re thrilled to partner with Honda to usher in a new era of energy-efficient automotive computing, where every vehicle can have the AI brainpower of a data center, without the power draw.”
The new AI system-on-chip (SoC) will be capable of running advanced machine learning models to make Honda’s vehicles smarter and safer. These range from vision transformers for perceptive awareness, to physics-informed neural networks for vehicle dynamics and control, to cloud-free large language models for in-car assistants.
Initial prototype chips from the Honda-Mythic collaboration are expected to be tested in vehicles by the late 2020s/early 2030s, with the jointly developed analog AI SoC slated to enter production shortly after completing successful trials.
