XPeng Inc. — AI supply-chain exposure

XPeng is a Chinese volume EV maker that sits on the application edge of the AI chain, but it behaves more like a software-and-silicon company than a traditional automaker. It writes its own advanced driver-assistance stack in-house, runs its own Turing AI chip program, and trains driving foundation models on large GPU clusters. In the model's terms it is an integrator: it consumes accelerators and turns them into a product — autonomous-driving software shipped inside cars — rather than making the underlying compute.

The structural hook is that XPeng is one of the most AI-forward carmakers building at scale, which makes it a genuine downstream demand sink for training compute and, over time, for in-car inference. Its exposure runs through GPUs, both for the fleet-scale training that improves the driving model and for its ambition to design silicon itself. That dual role — heavy compute buyer today, aspiring chip owner tomorrow — is why the model treats it as a real, if modestly weighted, node in the chain rather than a pure end-consumer.