China’s DeepSeek and Xiaomi Accelerate AI Model Releases

what happened

Chinese developers, including DeepSeek and Xiaomi, released 16 AI models in September, intensifying competition with OpenAI, Anthropic and Google. Nikkei Asia said the figure covers launches by leading companies rather than every model introduced in China during the month.

The average gap between upgraded-model releases at major Chinese and U.S. laboratories fell from 125 days between January 2023 and March 2026 to 44 days from April through September 2026. DeepSeek is emphasizing lower-cost open-weight systems that developers and businesses can adapt for private, local or specialized use.

Bloomberg reported that DeepSeek was nearing a funding round of at least 80 billion yuan, equivalent to about $12 billion, backed by investors including Tencent and Contemporary Amperex Technology. On October 12, Bloomberg also reported that DeepSeek had agreed to lead a seed round worth tens of millions of dollars for robotics startup Winverse AI. The proposed deal would value Winverse at about $400 million after funding, although it had not been finalized.

Anthropic CEO Dario Amodei has called for a slower pace of advanced-AI development and stronger independent safety checks. Bloomberg reported that Chinese developer Moonshot AI reached an approximate $50 billion valuation after its latest fundraising and was preparing for a possible Hong Kong IPO in early 2027.

why it matters

The faster release cycle and larger funding rounds intensify competition for computing capacity, researchers and customers. They also sharpen the commercial question of whether revenue and practical adoption can support the industry’s substantial infrastructure and training costs.

instruments