China’s open-weight AI surge rattles Google and US dominance

This piece captures a strategically important shift: China isn’t just racing on raw model capability, it’s weaponising price and openness. Moonshot’s Kimi K3 and Alibaba’s Qwen 3.8 Max are framed as surprisingly strong, open‑weight models that are cheap or free to use, while US incumbents struggle with delayed launches and fragile business models tied to high-margin APIs. That’s a classic disruption pattern: good‑enough products undercutting premium offerings on cost and flexibility.

For the race to AGI, this dynamic matters because it broadens who can experiment at the frontier. If capable models are downloadable and runnable on commodity hardware, thousands more researchers, startups and even state actors can probe safety limits, invent new agent architectures and push capabilities in directions the big US labs didn’t prioritise. At the same time, Washington’s response—talk of restricting access to Chinese models—shows how quickly AI competition is becoming entangled with trade controls and national security.

The other thread is capacity. Beijing’s subsidised compute and power give Chinese labs room to run aggressive open‑weight strategies that are uneconomical in the US. If that persists, we may see a world where state-backed ecosystems outside the US iterate faster on open AGI precursors, while American firms are boxed in by monetisation and export-control constraints.