Trade restrictions on high-end chips remain the primary friction point between Washington and Beijing. These chokepoints limit hardware access for Chinese labs. The resulting divide forces a divergence in model training strategies. Practitioners must now navigate fragmented supply chains and diverging regulatory frameworks for AI development.