State-backed Chinese investors eye top AI startups

State-backed Chinese investors are taking stakes in top AI startups including Beijing-based Moonshot AI, the maker of Kimi, as the country pushes for greater self-sufficiency in large language models.

Moonshot AI has added investors including the Beijing AI Industry Investment Fund, a fund under Shanghai Guosheng Group, and China Mobile, while Meituan’s venture arm is leading its latest financing round.

Separate media reports said DeepSeek is seeking more than 50 billion yuan (US$7.35 billion) in funding at a valuation above 350 billion yuan (US$51.4 billion), with China’s state-backed semiconductor fund in talks to lead the round.

Researcher Wang Peng said state capital can also bring computing power, data and industrial support.

🔗 Source: China Daily

🧠 Food for thought

Implications, context, and why it matters.

State backing goes beyond direct funding

These investments sit inside a broader state support system that helps Chinese AI startups deal with bottlenecks such as access to computing power 1.
Provincial governments also offer “computing vouchers” that repay part of a startup’s computing bill, often 20% to 50% depending on the location 2.
Favored companies such as DeepSeek can receive “national high-tech enterprise” status, a government label that opens the door to lower taxes and grants 1.
Startups may also win public-private partnership work on AI-powered “City Brains,” software platforms local governments use to manage traffic, public services, and other urban operations. Those projects can bring in revenue plus access to large datasets 1.

State money brings obligations for AI firms

Taking state funding often means matching national goals, including a more self-reliant AI technology stack during a broader push for “self-reliance and self-strengthening” across hardware and software 2.
Chinese authorities urged DeepSeek to train a new model on Huawei hardware, according to reports. Huawei is the Chinese telecoms and technology company that also makes AI chips. Training later moved back to Nvidia hardware, while Huawei Ascend graphics processing units (GPUs) handled inference because Huawei chips still posed problems for training 3.
The pressure grew after U.S. export controls. DeepSeek’s founder said the company’s real obstacle was bans on advanced chips rather than money 4.
That setup can give state-backed AI firms better access to scarce domestic resources, including limited advanced manufacturing capacity at Semiconductor Manufacturing International Corporation (SMIC), China’s leading contract chipmaker, where Huawei is said to hold outsized leverage and a dominant position 2.