From voice-first interaction and intelligent assistants to AI-optimised hardware, the next generation of automotive cockpits will be built around agentic AI.
Yu Yang, Principal Lead Analyst – Automotives & Robotics at Yole Group and an Advisor on the InCabin China Advisory Board explores how this shift is changing vehicle architecture, semiconductor requirements and the future of human-vehicle interaction.Â
AI Agent is re-shaping cockpits. There are increasing cases of agentic Ais integrated to cockpits. Probably the most known one is Grok from Tesla. Other cases cover ChatGPT from OpenAI, Alexa+ from Amazon, Doubao from ByteDance, Qwen from Alibaba, DeepSeek from DeepSeek, alone with multiple proprietary agents from OEMs’ own development. The latest one is Gemini from Google to be introduced to Volvo’s new models.Â
There has been a successfully implemented function of interactive navigation. And some OEMs are also using agents as ChatBots, or news feeding. However, the real potential lies in the capability of operating various control functions, understand and implement complicated commands, and deeply customized services. This large language model (LLM) enabled voice recognition will be the center of human-vehicle interaction, rather than the current paradigm of touch screen based one with multiple layer of commands. Of course, functional safety must be carefully scrutinized in vehicle controls. We believe these AI Agents would be the key of future cockpit, and we are just at the beginning of new era.Â
This new trend clearly requires an update of SoC and memories capable of running AI models with more parameters, further pushing the needs for SoCs with more NPUs, which has been largely needed in ADAS. There are some suppliers are working on AI Box, which is an independent hardware embedded with and optimized for AI Agents for cockpits. A more economic way is AI accelerator chips, with large mount of NPU blocks. Some OEMs start using the same SoCs for both ADAS and cockpit applications (of course, separate chips, but the same type), as both requiring large AI models on the edge.Â
Another type of semiconductor falls within our scope is ICs for audio functions. This includes specific audio DSP, interface ICs for audio bus, SoCs or MCUs with embedded audio functions, etc. Traditionally audio chips are mainly for advanced audio experiences in premium vehicles. However, if speech recognition is the base of future interaction onboard, voice collection (via microphones), noise cancellation, signal amplification, processing and transmission, will be critical.Â
Connect with us on WeChat via the QR code below ⬇
Interested in exterior vehicle sensing technology?
With a pass to InCabin China, you’ll also get full access to our co-located sister event, AutoSens. Find out more here >>
Watch more interviews and technical discussions from across the InCabin community on our YouTube channel ⬇
