Topic
China's AI sector, including labs like DeepSeek, Moonshot and Alibaba, their funding, state-linked capital and global competitive standing.
DeepSeek is reportedly doubling its second round to 100 billion yuan at a valuation near 500 billion yuan, four months after the first. Government-backed and corporate funds are setting that price, on revenue figures attributed to The Information.
Reality
The seven-month numbers behind DeepSeek's pre-IPO round pair the best gross margin anyone reports with one of the smallest revenue lines. A Shanghai filing would put both on the record.
Reality
Anthropic says six campaigns since February 2026 pulled reasoning transcripts out of Claude through proxy relays built on fictitious identities, stolen credit cards, and API keys harvested from legitimate companies and individuals.
Reality
A developer found a 313MB archive of his commercial project queued for upload and could not open it, because the private key sits on Z.ai's back end. Z.ai says the data is destroyed once the page is built.
Reality
Rhodium Group's tally of Chinese AI model revenue comes to about $10.5 billion against $105 billion for OpenAI and Anthropic. Two of the Chinese startups now heading for listings carry the highest revenue multiples in the set.
Reality
Zhipu closed about $5 billion two months after raising $4 billion in a July share sale. The largest sum in the same tally is the $29.6 billion that ByteDance borrowed from roughly 30 banks at home and abroad.
Reality
The forecast, carried by state broadcaster CCTV, describes compute demand that is billed every time somebody uses a model. Europe is building for the same demand with seven gigafactories on a EUR 30bn budget, of which about EUR 1bn is committed.
Reality
The Beijing lab behind Kimi K3 has filed confidentially for a Hong Kong listing and is now weighing Shanghai's STAR Market too. Nothing in the reporting puts a number on what training its models costs.
Reality
Bloomberg says the model would stream worlds at about 20 frames a second with 50 milliseconds of latency. That is one frame of headroom, and every minute a Pico user spends inside one is compute somebody has to pay for.
Reality