AI-defined vehicles hit compute, memory and validation limits

A Semiconductor Engineering industry overview finds automakers split between centralized "server-on-wheels" compute and distributed, right-sized architectures. Raw camera and radar data is straining memory, interfaces, in-vehicle networks and power efficiency, while AI decisions and updates must be validated across edge cases over a vehicle's long service life. Physical AI is seen as a $1 trillion-plus market over the next decade.

• OEMs split between central compute and distributed architectures
• Raw camera and radar data strains memory, networks and power
• AI decisions and updates need validation across edge cases
• Physical AI market seen at over $1 trillion in a decade