Honda Motor Co., Ltd. and Mythic announced a joint development agreement to co-develop an automotive-grade analog AI system-on-chip (SoC) for Honda’s next-generation software-defined vehicles. Honda R&D will license Mythic’s analog compute-in-memory technology, with the partners targeting vehicle testing in the late 2020s and production in the early 2030s.
The collaboration focuses on sharply improving on-board AI performance under tight automotive power budgets. Mythic’s architecture integrates memory and computation, aiming to reduce data movement and deliver large gains in energy efficiency versus conventional digital AI accelerators. The companies said the roadmap targets more than 100,000 TOPS of AI compute for advanced driver-assist and autonomous functions while aligning with Honda’s long-term safety and carbon-neutral goals.
The jointly developed SoC is designed to support a range of in-vehicle workloads, including vision models for perception, control-oriented neural networks, and local language models for assistants without relying on the cloud. Prototype chips are expected to enter vehicle trials after initial development and qualification, followed by automotive-grade production once testing completes.
- Joint development of an automotive-grade analog AI SoC for Honda SDVs
- Licensing of Mythic’s analog compute-in-memory APU technology by Honda R&D
- Targeting 100x energy-efficiency gains and 100,000+ TOPS within vehicle power envelopes
- Initial vehicle testing planned for late 2020s; production targeted early 2030s
- Focus on ADAS, autonomous driving, and on-board AI workloads without cloud dependence
“Mythic’s cutting-edge analog compute technology makes them a strategic partner for Honda as we develop the next generation of intelligent, safe vehicles,” said Atsushi Ogawa, Chief Operating Officer at Honda R&D Co., Ltd.
🌐 Analysis
Automotive OEMs are pushing more AI compute on-board as software-defined vehicles scale, but power and thermal limits remain a hard constraint, driving interest in alternative architectures beyond GPUs and NPUs. Honda’s move follows broader industry experimentation with neuromorphic and in-memory compute approaches as competitors and suppliers seek differentiated paths to higher AI performance-per-watt for advanced driver-assist and autonomy.
MythiC is a Palo Alto–based semiconductor startup developing ultra-low-power AI inference processors built around an analog compute-in-memory architecture. Founded in 2012, Mythic’s mission is to dramatically reduce the energy and cost of running neural networks by performing matrix-multiply operations directly in analog flash memory, minimizing data movement—a core bottleneck in conventional digital AI accelerators. The company’s core intellectual property centers on this analog compute fabric and a tightly coupled software stack that maps trained neural networks onto its hardware. Mythic has been led by CEO Mike Henry, who previously held senior roles at Intel, and was co-founded by a team with backgrounds in semiconductor design and embedded systems. Its flagship products include the M-series AI processors and development platforms targeting edge and embedded inference use cases such as industrial vision, automotive, and IoT. Mythic has raised over $170 million from investors including SoftBank Ventures Asia, DCVC, and AME Cloud Ventures, and has reported partnerships and evaluations with system integrators and OEMs exploring power-constrained AI deployments. Key milestones include the commercial launch of its analog AI silicon, public demonstrations of sub-watt high-TOPS inference performance, and ongoing work to adapt its technology for automotive-grade and safety-critical systems.
🌐 We’re tracking the latest developments in networking silicon. Follow our ongoing coverage at: https://convergedigest.com/category/semiconductors/
