Li Auto (02015.HK) is reportedly exploring an in-house cloud inference chip; the
project is in an early stage and may adopt a dataflow architecture similar to
its vehicle-side autonomous driving chip. Li Auto said: "We have not heard
anything about this so far." Industry insiders say cloud inference demand will
rise as autonomous-driving large models, in-car agents and multimodal
interaction evolve. Future competition among automakers may hinge not only on
in-vehicle compute but on the ability to build cloud AI infrastructure that
supports continuous model iteration, low-cost invocation and scalable delivery.
