FrontierThe story, in brief
Meituan's LongCat-2.0 shows China can train massive AI models without Nvidia
1.6 trillion parameters. Meituan just trained a massive LLM entirely on Chinese chips—no Nvidia required.
The KeyNews take
Why it matters
China's ability to train frontier-scale models without US chip dependence reshapes the geopolitical AI landscape and signals a critical shift in the global compute hierarchy.
The key facts
4 to knowMeituan LongCat-2.0: 1.6 trillion parameters
Trained entirely on Chinese chips (no Nvidia GPUs)
Demonstrates China's indigenous chip capability for large-scale model training
Published June 30, 2026
Go to the source
The Decoderthe-decoder.com
Publisher excerpt: Meituan trains a 1.6 trillion parameter AI model entirely on Chinese chips, no Nvidia required.