Which companies has the 60 billion yuan national AI fund publicly invested in?

The national team is still ramping up investment in artificial intelligence. On October 19, Wei Wei, director of the Department of Science and Technology of the Ministry of Industry and Information Technology, said at a press conference of the State Council Information Office that the 60 billion yuan National Artificial Intelligence Industry Investment Fund has already invested in a batch of high-quality projects. In 2025, the scale of China's core AI industry exceeded 1.2 trillion yuan, and the penetration rate of AI technology applications among manufacturing enterprises above designated size exceeded 30%. As of the end of June this year, the number of AI enterprises in China exceeded 6,600.

The official list of exactly which "high-quality projects" this refers to has not been disclosed. According to Caixin's tally on October 20, there are currently three publicly disclosed direct investment projects, all of which went to leading large-model companies. Combining listed company announcements, business registration information and financing news, the fund has made at least eight publicly disclosed investments, covering five links: models, applications, embodied intelligence, computing power chips and AI security.

01. Official list of investees not disclosed

The National Artificial Intelligence Industry Investment Fund was established on January 17, 2025, led by the Ministry of Industry and Information Technology and the Ministry of Finance, with a capital contribution of 60.06 billion yuan. The third phase of the National Integrated Circuit Industry Investment Fund contributed 60 billion yuan and is the fund's sole limited partner; the manager is Guozhitou (Shanghai) Private Equity Fund Management Co., Ltd., whose actual controller is the Shanghai Municipal State-owned Assets Supervision and Administration Commission. The fund is based in Xuhui, Shanghai, and has a term of 13 years.

Twenty-one months after its establishment, the official side has not disclosed the name or amount of any invested project. Wei Wei's introduction on October 19 stopped at "a batch of high-quality projects."

The most systematic official introduction of the fund came from an embodied intelligence symposium held by the Shenzhen Stock Exchange on April 10, 2025. Zhang Jianhua, deputy director of the Planning Department of the Ministry of Industry and Information Technology and deputy head of the preparatory group for the National AI Fund, said at the time that the fund would carry out investment layout around the entire AI industry chain, covering links such as computing power, algorithms, data and enabling applications, follow the laws of industry development, "appropriately invest early, small and in the frontier," and have both industrial investment and venture capital attributes. The introduction did not involve specific projects.

Since then, official statements have focused on positioning. In January this year, the Ministry of Industry and Information Technology said it would strengthen the fund's support for the humanoid robot field; on August 26, Vice Minister Xin Guobin proposed at a State Council Information Office press conference that the AI fund be promoted to pool financial "living water" and increase investment.

The disclosed capital contributions can be sorted into eight deals.

According to Caixin, in DeepSeek's first round of financing, the fund contributed about 980 million yuan. Caixin said the post-investment valuation of that round was 350 billion yuan, while Securities Times and Caijing reported a figure of nearly 400 billion yuan. The fund also participated in Moonshot AI (Kimi)'s Series F financing, which had a post-investment valuation of $35 billion.

A listed company announcement disclosed another deal. On August 31, Kuaishou announced on the Hong Kong Stock Exchange that the fund invested 1.4 billion yuan in the domestic entity of its video generation large model Kling AI, for a stake of about 1.14%. Based on this ratio, the domestic entity of Kling AI is valued at more than 120 billion yuan. Because it fulfilled mandatory disclosure obligations, this deal has the most complete information.

Unmanned delivery vehicle company Neolix is the fund's first investment in that track. On October 20, Neolix announced the completion of nearly $400 million in financing, with the official wording being "led jointly by Boyu Investment and a certain national-level fund"; Xinhua Finance and other outlets reported that the fund is the National AI Fund. To date, Neolix's cumulative financing has exceeded 10 billion yuan.

Four other contributions appeared in earlier reports. In March, Galbot completed 2.5 billion yuan in financing, with investors including the fund, at a post-investment valuation of more than 20 billion yuan; at the end of January, the fund appeared on the shareholder list of smart-driving chip company Xinxin Hangtu; on January 27, AI security company Chaitin Tech completed 500 million yuan in financing, with investors including the fund; and according to Xinhua Finance, the fund also invested in AI computing power chip company MetaX.

Among the eight contributions, the publicly verifiable amounts total about 5.4 billion yuan, less than one-tenth of the fund's 60.06 billion yuan scale.

02. Head companies concentrate, capital flows to computing power

The eight contributions are distributed by position in the industry chain: two in the model layer, DeepSeek and Moonshot AI, both adopting the open-source route—Kimi K3 was fully open-sourced on July 27 this year, with 2.8 trillion parameters, making it the largest known open-source model in the world; two in the application layer, Kling AI and Neolix; one in embodied intelligence, Galbot; two in computing power chips, Xinxin Hangtu and MetaX; and one in AI security, Chaitin Tech. The eight invested companies are spread across five links, and most have already formed products and customers in their niche tracks.

The flow of funds goes beyond the invested companies. The main use of financing for model companies is to purchase computing power. DeepSeek completed 51 billion yuan in first-round external financing in June this year, at a post-investment valuation of nearly 400 billion yuan; according to Caijing and other reports, the company subsequently advanced the construction of a large data center in Ulanqab, Inner Mongolia, planning to deploy at least 160,000 Huawei Ascend accelerator chips, with chip procurement of about 17 billion yuan. The fund also directly invested in MetaX, giving it a presence on the computing power supply side.

This flow is already reflected in macro data. National Bureau of Statistics data show that in the first three quarters, the value-added of digital product manufacturing above designated size contributed more than one-third of the growth of industry above designated size, and integrated circuit output grew 23.3%.

03. The national team's money follows AI that can actually be used

Why these companies, and why this timing?

A research report published by CITIC Securities on October 21 offers one explanation. The report estimates that the combined ARR (annual recurring revenue) of domestic models is about $8 billion, equivalent only to the level of overseas models 18 months ago; valuations are about one-tenth of those of overseas models. The report argues that once the capability gap narrows, the low base, fast growth and low valuations of domestic models could drive a systemic re-rating.

The capital pace of invested companies is accelerating. Moonshot AI completed a Pre-IPO round after its F round at a valuation of about $50 billion, confidentially filed with the Hong Kong Stock Exchange in September, and plans to list in the first quarter of 2027; DeepSeek's second funding round was upsized to about 100 billion yuan, with a pre-money valuation of about 500 billion yuan, led by CATL and Tencent, and has engaged CITIC Securities to plan a STAR Market listing, targeting a 2027 listing.

In addition, AI's impact on the employment structure has begun to show. According to an audit industry source close to the Smart Economy Bureau, accounting firms have introduced AI-assisted work this year, hiring has fallen compared with the previous year, and intern quotas have also been reduced. National Bureau of Statistics spokesperson Wang Guanhua said on October 19 that AI is rapidly moving from a "technology option" to a "production standard."

Moreover, AI development is a "money-burning" battle. A Goldman Sachs research report forecasts that global AI investment will reach about $1.02 trillion in 2026, of which about $581 billion will be in the United States; the combined 2026 capital expenditure guidance of Microsoft, Amazon, Alphabet and Meta is about $725 billion, up 77% from about $410 billion in 2025, with the vast majority going to data centers and AI chips.

Investment intensity determines the competitive threshold. Each step up in industrial scale requires hundreds of billions of dollars in sustained funding; the payback period for such investment is measured in five or ten years, beyond the duration limit of most market-based capital. Long-cycle capital thus becomes necessary, which is the direct background for fiscal and state-owned capital to set up dedicated funds.

Finally, 60.06 billion yuan, equivalent to 5% of the scale of the AI core industry in 2025, with publicly verifiable contributions amounting to less than one-tenth. The money is not much, but the direction has been drawn.

What comes next is delivery. Moonshot AI has confidentially filed with the Hong Kong Stock Exchange, DeepSeek's second funding round has doubled to 100 billion yuan, CATL and Tencent are on the lead investor list, and companies the national team has invested in are moving one by one toward the secondary market. By the time they list, "what the national AI fund invested in" will be written into the shareholder list of every prospectus, and into every candlestick chart. Whether the money was well spent—the market is already voting with its feet.

This article comes from the WeChat public account "Caineng Quan" and is published by 36Kr with authorization.