Chinese artificial intelligence startup Zhipu AI reported first-half revenue of 953.9 million yuan ($141.96 million), marking a 400% increase from the same period a year earlier, according to financial figures published by Reuters. The Beijing-based large language model developer, which recently completed a major Hong Kong initial public offering, drove sales growth by scaling demand for its models while rolling out lower-cost products and expanding its use of domestic semiconductor chips.
Zhipu AI Hong Kong IPO and Market Debut
Zhipu became the first of China’s prominent “AI tigers” to go public following a $558 million initial public offering in Hong Kong, according to reporting from CNBC. Shares of the firm climbed as much as 16% above their offer price of 116.20 Hong Kong Dollars ($15) on their debut, with approximately 37.4 million shares on offer. The Beijing-based startup ultimately closed its first trading session at $131.50, up 13.1%. Founded in 2019 by researchers from a top Chinese university, the company was valued at around HK$4.3 billion by the flotation. Zhipu’s prospectus indicates that the firm plans to direct 70% of the IPO proceeds toward the research and development of general-purpose large AI models.
Financial Performance and R&D Spending
For the six months ended June 30, Zhipu posted a net loss of 2 billion yuan, narrowing from a loss of 2.4 billion yuan a year earlier, as reported by Reuters. During the same timeframe, the company increased its research and development spending by 36.6% to 2.1 billion yuan. Market competition remains intense across China’s AI sector, with domestic rivals including Alibaba, ByteDance, and startup Moonshot cutting prices and accelerating new model launches. A competing Chinese AI startup, MiniMax, reported a 283% jump in first-half revenue to $116.6 million, while its adjusted net loss more than doubled, according to Reuters data.
US Export Controls and Domestic Chip Adoption
Zhipu’s operations have faced regulatory pressure from the United States, including its placement on the U.S. Commerce Department’s Entity List in January of the previous year over alleged ties to China’s military, according to CNBC. U.S. restrictions on advanced semiconductor technology and expertise have constrained the firm’s ability to train its AI models. In response, Zhipu has accelerated a pivot away from imports by expanding the use of domestic chips. The company released its lower-cost GLM-5.3-Flash model this month, noting that it was tested entirely using Chinese-made chips, according to Reuters.
Global Expansion and Competitive Positioning
Despite generating revenues that remain a fraction of those reported by U.S. laboratories—such as Anthropic, which topped a $65 billion annual revenue run rate by the end of July, and OpenAI, which exceeded $25 billion earlier in the year—Zhipu’s open-source models have gained international traction, according to Reuters. The firm has established international offices in the United Kingdom, Singapore, Malaysia, and the Middle East, alongside joint innovation center projects in Southeast Asian nations including Indonesia and Vietnam, as reported by CNBC. To differentiate its products, Zhipu has focused heavily on coding and cybersecurity models. The company stated that its flagship GLM-5.3 model matched Anthropic’s Mythos 5 on specific white-box code-review and vulnerability-discovery tests, though it lagged behind on more complex vulnerability-exploitation tasks.
