China's Zhipu AI raises $137m from state investors as AI race heats up

China's Zhipu AI raises $137m from state investors as AI race heats up

The offices of artificial intelligence tech company Scale AI in Montreal, Quebec, Canada April 7, 2024. REUTERS/Evan Buhler/File Photo

Chinese artificial intelligence startup Zhipu AI has raised more than 1 billion yuan ($137.22 million) in fresh funding, months after securing a 3 billion yuan investment.

The funding round comes amid intensifying competition in China’s AI sector, particularly after rival DeepSeek’s emergence with its large language models that claim to match Western competitors’ capabilities at lower costs.

State-backed Hangzhou City Investment Group Industrial Fund and Shangcheng Capital are among the investors in the round, Zhipu AI said in a WeChat statement on Monday.

The Beijing-based startup said it would use the funds to enhance its GLM large language model and expand its AI ecosystem, with a focus on serving businesses in Zhejiang province and the Yangtze River Delta economic zone.

The deal marks a significant investment from Hangzhou City Investment Group’s year-old industrial fund, highlighting the eastern Chinese city’s push to become a major AI hub.

Hangzhou, which also hosts DeepSeek, has been actively artificial intelligence ventures through state-owned enterprises.

Zhipu AI, founded in 2019 and considered one of China’s “AI tigers”, has completed 16 funding rounds, according to business registration platform Qichacha. The company raised 3 billion yuan in December, with state-backed Zhongguancun Science City among the investors.

The funding push comes as open-source AI models from competitors such as DeepSeek have disrupted the market dynamics for second-tier AI firms. DeepSeek’s models have gained attention for matching the capabilities of leading Western platforms such as OpenAI.

In the WeChat statement on Monday, Zhipu AI announced plans to release a suite of new AI models under an open-source framework, including foundation models, inference models, multimodal models, and AI agents.

Reuters

This is your last free story for the month. Register to continue reading our content