China AI Lines

The long-running stories in Chinese AI: open weights, the price war, robotics, chips and regulation. Each line states a working thesis, its evidence strength and what would change it.

5 active linesSnapshot

Working judgments, revised when the evidence changes.

Each line separates the current thesis from its evidence state and revision history.

Open-weights competition

Chinese labs compete by releasing open weights, not only through closed APIs

Current thesis

Leading Chinese labs release model weights under permissive licenses and use downloads, fine-tunes and developer adoption outside China as a distribution channel alongside their own APIs.

Direction
Unknown
Evidence
Insufficient · 0 events
Freshness
Current
Revision
1

Model price war

Chinese model APIs compete on price as much as on capability

Current thesis

Chinese model providers use API price cuts, cache discounts and free tiers to win developers, which puts pressure on model margins at home and abroad.

Direction
Unknown
Evidence
Insufficient · 0 events
Freshness
Current
Revision
1

Robotics & embodied AI

Chinese robotics makers are turning AI models into shipped hardware

Current thesis

Chinese humanoid and robotics companies are combining AI models with low-cost hardware supply chains and moving from demonstrations toward products sold to customers.

Direction
Unknown
Evidence
Insufficient · 0 events
Freshness
Current
Revision
1

Chips & compute self-reliance

Chinese AI is adapting to limited access to leading-edge chips

Current thesis

Export controls limit Chinese access to leading-edge AI chips, so labs and chipmakers respond with domestic accelerators, efficiency-focused training and inference, and shared compute.

Direction
Unknown
Evidence
Insufficient · 0 events
Freshness
Current
Revision
1

AI regulation

China governs AI through filing, labeling and service rules

Current thesis

China regulates AI mainly through rules on services offered to the public, algorithm filing, content labeling and technical standards rather than one broad AI law.

Direction
Unknown
Evidence
Insufficient · 0 events
Freshness
Current
Revision
1