Xiaopeng Claims Monthly 300 Million Yuan Investment in Autonomous Driving AI Learning, Now Parity with Tesla's FSD v13

Xiaopeng revealed it is allocating approximately 300 million yuan monthly to autonomous driving AI learning, achieving performance comparable to Tesla's FSD v13. The company's VLA 2.0 architecture eliminates an intermediate language processing step, directly converting visual data to driving actions while retaining language for human instruction interpretation. Xiaopeng emphasizes that language processing introduces unnecessary computational overhead, opting for a camera-centric sensing strategy with redundant safety layers for emergency braking and steering. The firm also detailed its 5 billion USD annual investment in AI training, highlighting 4360% efficiency gains and 90% GPU utilization rates, while comparing its data-centric approach with Tesla's technological philosophy.