Unitree Robotics, China's leading humanoid robot maker, priced its initial public offering on Shanghai's STAR Market at 150.80 yuan per share on August 6, 2026, implying a valuation of 61 billion yuan (approximately US$9 billion); DeepSeek, the Chinese AI lab that upended global AI markets in early 2026, took a strategic placement stake; Unitree is seeking to raise 6.1 billion yuan (around US$904 million); subscriptions open August 10
China-based DeepSeek disrupted global AI markets in early 2025 with open-weight models competitive with US frontier labs but trained at a fraction of the cost.
full dossier ›China's Alibaba unveiled Qwen3.8-Max, its largest AI model at 2.4 trillion parameters, on August 3, sending its shares higher; simultaneously DeepSeek released V4-Flash, priced at more than 100 times below comparable Anthropic models according to analyst research; the two launches on the same day represent opposite strategies among Chinese AI firms: scale on capability versus slashing inference costs
Google DeepMind pushed Gemini 3.5 Pro from June to July 17, 2026, after discarding the existing 2.5 Pro architecture entirely; Alphabet lost US$225 billion in market value in a single session; four senior DeepMind researchers departed for rivals; DeepSeek is forcing an API migration for all users by July 24
China's DeepSeek is developing its own inference AI chip, reportedly to be manufactured at SMIC, according to anonymously sourced reporting from Reuters; the move would mark a major strategic shift for the company beyond model development into hardware
China's Alibaba Group runs the Qwen LLM family, the most-downloaded Chinese open-weight model series, shaping the global open-vs-closed AI frontier debate.
Eight AI labs, split across the US, Europe and China, are writing the capability ceiling for the technology reshaping global industry and statecraft.
China's Zhejiang capital is the country's second-largest technology cluster after Beijing, anchored by Alibaba and now producing a new generation of global AI and robotics firms.
A 1M-token sparse-attention model lands above GPT-5.5 on its own coding benchmark, below Claude Opus 4.8, with weights still withheld
The 744B-parameter sparse MoE model with a real 1M-token context tops SWE-bench Pro among open-weight models and undercuts frontier closed-weight pricing sharply; MIT licence removes every regional restriction
DeepSeek and Qwen keep open models near-frontier and cheap; Meta and Alibaba keep their best tiers proprietary
V4-Pro at 1.6T params (49B active) and V4-Flash at 284B, both 1M-context, released open-weight on Hugging Face