MMOOZZGG.com

AI investment

AI investment in China: deals, IPOs and startups of 2026

5 min read3 sources

AI investment in China was $12.41 billion in 2025, according to Stanford HAI, and in 2026 a single DeepSeek round brought in $7.4 billion. Below: the large rounds, the stock market listings of chip and robot makers, and what of this matters for a Russian company.

On this page · 5 sections

China in numbers

Stanford HAI counted $581.69 billion of corporate AI investment worldwide in 2025 in its AI Index 2026. China took $12.41 billion of it, the US $285.88 billion and Europe $20.92 billion.

The Stanford HAI figure counts money that went through visible deals: private investment, mergers and acquisitions. Worldwide, private AI investment in 2025 reached $344.66 billion, and mergers and acquisitions $214.44 billion.

In 2026 the picture for China is clearly different. By our count, just two events, the DeepSeek round of $7.4 billion and the $8.6 billion listing of ChangXin Memory Technologies, gave more than Stanford HAI counted for China in all of 2025.

Compare these sums directly with care: a round and a listing are different kinds of money. But the direction is clear: large Chinese AI companies raise sums of global scale in 2026.

The $12.41 billion figure is a useful baseline. It shows how fast Chinese deals grew in a year, and in a year from now it will show whether the growth held.

Large rounds

The largest Chinese round of the year in our sources: DeepSeek, $7.4 billion in Q2 2026. KPMG put it among the three largest deals of the quarter worldwide, after Anthropic with $65 billion and Project Prometheus with $12 billion.

DeepSeek is known for open models that companies can run on their own servers. The $7.4 billion round gives it money for the next versions of such models.

In Q3 2026, Crunchbase noted two Chinese rounds of more than $3 billion each:

1. Moonshot AI, a developer of language models.

2. Kling AI, a service that creates video with AI.

The same Crunchbase list for Q3 has Databricks and Safe Superintelligence with $5 billion each, as well as Crusoe, Mistral AI and Nscale. Chinese companies stand in the same row as US and European ones by round size.

All rounds with amounts, dates and sources are in the deal tracker of our guide.

IPOs of chip and robot makers

In Q3 2026, three Chinese companies went through an IPO, a first sale of shares on a stock exchange. All three make hardware for AI, not software.

1. ChangXin Memory Technologies, memory for computing: $8.6 billion at a valuation of $85.5 billion.

2. Enflame, AI chips: $912 million at a valuation of $25.5 billion.

3. Unitree Robotics, robots: $905 million at a valuation of $9 billion.

For China, this is a way to build its own supply chain: memory, computing chips and robots are now funded through the stock exchange, not only through state programs and funds.

AI company listings in 2026 are covered on a separate page, and investment in robots on our page on robotics.

Comparison with the US

By the results of 2025, the gap is large. Stanford HAI counted $285.88 billion of corporate AI investment in the US and $12.41 billion in China.

In single deals of 2026, the gap is smaller. The DeepSeek round of $7.4 billion is larger than the $5 billion rounds of Databricks and Safe Superintelligence, and the $85.5 billion valuation of ChangXin is comparable to large private US AI companies.

The structure of the money differs too. The largest US rounds of the year went to model developers: Anthropic got $65 billion. Chinese money is more visible in hardware: memory, chips and robots.

One figure for a year is not enough to judge the Chinese market. Look at deals by quarter and at the segments they go to, or the picture will be out of date by the next report.

What matters for Russia

Chinese chip and robot makers raised hundreds of millions of dollars on the stock exchange: Enflame $912 million and Unitree Robotics $905 million. They have money for development, and for a Russian buyer of hardware this means these suppliers will offer new product versions.

For a Russian company, this leads to three practical steps:

1. When you choose an AI hardware supplier, look at how it is funded. A listed company publishes its financial reports, so its stability is easier to check.

2. If a solution is built on an open Chinese model, plan for a way to replace it. The model market changes within quarters, and $7.4 billion rounds come to new players too.

3. Compare a Chinese solution with others by the cost of ownership over several years, not by the price of the first delivery.

China also shows a general lesson: the country invests in its own memory, chips and robots to depend less on other countries’ supplies. The same principle helps a Russian company on a small scale: an AI solution that runs on your servers with a model you can replace depends less on other people’s decisions.

Sources and fact-checking

Checked by the editors on 14 October 2026.

  1. Stanford HAI, AI Index 2026, chapter “Economy”, 13 April 2026 (hai.stanford.edu)
  2. Crunchbase News, 5 October 2026 (news.crunchbase.com)
  3. Kommersant, citing KPMG, 13 August 2026 (kommersant.ru)

Found a mistake? Tell us

Publications

Mike Mozg

Agentic AI use cases at AT&T: nearly 500 processes now run on agents

AT&T no longer uses AI only to help staff work faster: it is rebuilding whole business processes around agents. SiliconANGLE reported on 14 November 2026 that nearly 500 agentic AI use cases at AT&T are already in production. Here is how the company rebuilds them, where a human still checks the agent, and what other large companies can take from it.

News4 min
Contact us