Zhipu AI has raised 2.5 billion yuan ($340 million) in total financing this year, according to an announcement from the Chinese foundation model developer. Its backers include major internet companies and local investment funds, giving the company capital to continue building AI models amid restrictions affecting China’s access to technology and investment.
A broad coalition of Chinese backers
The investment brings together companies that are not usually partners. Alibaba and Tencent, described as archrivals, are both among the investors. Ant Group, Xiaomi, Meituan, Kingsoft, TAL Education Group and Boss Zhipin are also part of the group.
Financial investors include HongShan, Shunwei Capital and Hillhouse Capital. A state fund managed by Legend Capital also participated. The mix of internet businesses, venture capital firms and a state-managed fund gives Zhipu a varied set of domestic sources for its financing.
The funding was raised from yuan-denominated funds. That marks a change from a two-decade pattern in which U.S. dollar funds were the preferred source of financing. The source connects this shift to geopolitical tensions and the widening technology divide between the U.S. and China.
Why the funding matters now
The announcement came during a period of tighter U.S. controls affecting China’s AI ambitions. This week, the Biden administration imposed additional restrictions on exports of Nvidia AI chips to China. The limits make it harder for Chinese developers to obtain the hardware used to train large language models.
Chinese AI companies have been stockpiling semiconductors in anticipation of restrictions from Washington, spending hundreds of millions of dollars on the sought-after chips. That context helps explain why companies working on large AI models need substantial capital: building them requires access to costly computing resources, while that access is becoming more constrained.
Investment has also faced political pressure. In August, President Joe Biden signed an executive order barring U.S. investments in critical Chinese technology sectors, including AI, semiconductors and quantum computing. The stated goal was to curb China’s military build-up, while the order also discouraged American venture capital firms from investing in sensitive areas in China.
Some firms have responded by separating their China operations. Sequoia Capital China was renamed HongShan, and GGV Capital also sought a way to continue operating in the market. The change in financing sources offers one sign of how investment patterns are adapting as the U.S. and China’s technology sectors grow further apart.
From university spinout to model developer
Founded in 2019, Zhipu was spun out of Tsinghua University. The company is led by Tang Jie, a professor in the university’s Department of Computer Science and Technology. Its origins connect the startup to one of China’s prominent academic institutions as it competes in the foundation model field.
Zhipu has released open-source models for conversational AI and broader foundation model work. Its bilingual ChatGLM-6B supports Chinese and English, and is trained on six billion parameters. The company says it can carry out inferences on a single consumer-grade graphics card.
The company also has an open-source foundation model, GLM-130B, trained on 130 billion parameters. Together, the models indicate that Zhipu is working across different scales, from a conversational system designed to run inference on consumer hardware to a much larger foundation model.
Competing in a divided AI landscape
Zhipu is one of several Chinese candidates emerging as the U.S. and China develop separate AI ecosystems. In the U.S., the source points to OpenAI and Anthropic as notable players; Zhipu is presented as one of China’s challengers in the same broad race.
The company’s financing does not remove the constraints around chips, investment or model development. It does, however, show how local capital and partnerships among Chinese technology companies can support an AI developer as international funding channels narrow and access to advanced semiconductors becomes more difficult.