Zhipu AI Targets STAR Board IPO to Raise approximately $2.2 Billion as China's AI Leaders Race Back to A-Shares

Zhipu AI, a leading Chinese large language model company, has officially taken a key step toward returning to China's A-share market. On the evening of June 1, Zhipu announced on the Hong Kong Stock Exchange its intention to apply for an A-share listing on the Shanghai Stock Exchange's STAR Board, planning to raise up to 15 billion yuan (approximately $2.2 billion). The move comes less than five months after its Hong Kong debut in January, signaling that the AI giant is accelerating its push to secure a strategic position in China's domestic capital market after achieving global public market pricing.

Just two days earlier on May 29, another leading Chinese AI firm, MiniMax, had submitted its A-share IPO tutoring filing to the Shanghai bureau of the China Securities Regulatory Commission (CSRC). This means the "AI duo" that both listed in Hong Kong in January are poised to reunite on the STAR Board, jointly creating the first batch of pure-play large language model core assets in China's A-share market.

According to Zhipu's announcement, the board of directors unanimously approved the proposed A-share issuance and STAR Board listing at a meeting held on June 1. The filing shows the proposed A-share issuance will represent between 2% and 8% of the company's total post-issuance share capital, amounting to no fewer than 9.0988 million and no more than 38.769 million new A-shares.

The use of proceeds clearly reflects Zhipu's absolute emphasis on foundational technology R&D. Of the 15 billion yuan net proceeds, the artificial intelligence general-purpose foundation model project is allocated 12 billion yuan (approximately $1.8 billion), representing 80% of the total, primarily for R&D iteration of the next-generation foundation model GLM-6 and subsequent series, as well as continued investment in large-scale computing clusters. The MaaS one-stop service platform project is allocated 2 billion yuan (approximately $295.8 million), mainly to enhance enterprise-grade agent service quality and strengthen model training and inference infrastructure. The remaining 1 billion yuan (approximately $147.9 million) will supplement working capital.

Zhipu CEO Zhang Peng explained the company's pricing strategy at an earnings media briefing. As of March this year, Zhipu's API call pricing had risen 83% compared to the end of last year. Zhang stated that price is determined by value, and Zhipu focuses more on the value created per unit Token and the value delivered to customers, with pricing power determined by technological strength and the leadership position brought by long-term trends. Zhipu's continued investment in general-purpose foundation models is closely tied to its need to further consolidate GLM's competitive edge and maintain premium pricing power in the high-end enterprise market.

Financial data shows Zhipu achieved full-year revenue exceeding 724 million yuan (approximately $107.1 million) in 2025, up 131.9% year-on-year. Net loss for 2025 was 4.7 billion yuan (approximately $695.1 million), widening from 2.9 billion yuan (approximately $428.9 million) in 2024, which the financial report attributed primarily to continued increases in R&D spending. On gross profit, 2025 recorded nearly 300 million yuan (approximately $44.4 million), up 68.7% year-on-year, but gross margin declined from 56.3% in 2024 to 41% in 2025. The report attributed this mainly to an increased proportion of cloud deployment business and a phased decline in gross margin for on-premises deployment business. The company's MaaS platform annualized recurring revenue reached approximately 1.7 billion yuan (approximately $251.4 million), a 60-fold increase within 12 months, demonstrating strong commercialization momentum.

Zhipu's path back to A-shares has undergone precise strategic calibration. As early as April 2025, Zhipu had quietly submitted its first A-share tutoring filing, a move that came three months before the CSRC officially released opinions in July 2025 on expanding the scope of the STAR Board's fifth set of listing standards. In February this year, updated IPO tutoring progress on the CSRC's official website showed that Zhipu had withdrawn its initial tutoring filing submitted in April 2025 and immediately filed a new tutoring registration, directly targeting the STAR Board. The IPO tutoring institution was changed from CICC alone to a joint tutoring arrangement between Guotai Haitong Securities and CICC.

This adjustment was not a retreat but a proactive recalibration after assessing capital market dynamics on both sides. In July 2025, the CSRC officially issued opinions that marked a landmark targeted expansion and regulatory clarification of the STAR Board's fifth set of listing standards, opening an institutional pathway for AI companies to list on the STAR Board. Policy tailwinds on one side, Hong Kong pricing on the other, with A-share tutoring following closely behind — forming a precisely orchestrated capital market relay pipeline.

On the industrial investment front, Zhipu's ties with tutoring sponsor Guotai Haitong have deepened further. The latest corporate registry information from Tianyancha shows that Shanghai Zhipu Guotai Haitong AI Industry Venture Capital Partnership (Limited Partnership) was formally registered on May 26 in Shanghai's Zhangjiang area, with registered capital of 540 million yuan (approximately $79.9 million) and a business scope of venture capital investment. The fund was jointly initiated by Zhipu's wholly-owned subsidiary Beijing Zhipu Future Technology Co., Ltd. and Guotai Haitong's wholly-owned subsidiary Guotai Junan Innovation Investment Co., Ltd., while also attracting co-investment from multiple local government guidance funds including the Hubei Integrated Circuit Industry Fund, Shandong New Kinetic Energy Fund, and Guang'an state-owned capital. Guotai Junan Innovation Investment Co., Ltd. also serves as the fund's manager. This deep cooperation at the industrial investment level is viewed by the industry as a model case of capital-industry synergy.

Zhipu has long been active in AI ecosystem investing. As early as August 2024, Zhipu launched its AGI ecosystem fund — Z Fund (Xinglian Capital) — led by Zhipu and jointly established with partners including the Shijingshan Modern Innovation Industry Development Fund and Aofei Data, targeting outstanding AI model startups. By the end of 2025, Z Fund had invested in over 30 companies including Jiliu Technology, Wuwen Xinqiong, Muyan Zhiyu, Caizhi Technology, and Dongyi Technology, covering the entire industry chain from underlying infrastructure and industry solutions to upstream innovative applications. Zhipu's 2025 annual report shows that by end-2025, the total value of Xinglian Capital Fund reached 1.3027 billion yuan (approximately $192.7 million), an increase of 840 million yuan (approximately $124.2 million) from the previous year, representing a surge of 186.26%. Among this, Zhipu's book value amounted to 279 million yuan (approximately $41.3 million).

Founded in June 2019 by a technical team led by Tsinghua University Computer Science Professor Tang Jie, Zhipu is one of China's earliest enterprises focused on general-purpose AI large language model R&D. The company's core product, the GLM series of large models, has become one of the most widely adopted domestically developed models in China, with over 12,000 institutional clients as of September 2025. Zhipu's on-premises deployment has long been its primary revenue pillar, still contributing over 70% of revenue in 2025, with key clients spanning heavily regulated sectors such as finance, government services, and energy.

From a macro capital market perspective, Zhipu and MiniMax's successive decisions to return to A-shares reflect profound shifts in the pricing system for Chinese technology assets. In January this year, Zhipu issued H-shares at HK$116.20 per share, raising net proceeds of approximately HK$4.17 billion (approximately $532.4 million) from its global offering. Since then, Zhipu's secondary market valuation has soared, briefly touching a market capitalization of HK$800 billion (approximately $102.1 billion) during intraday trading on May 29. As of press time, Zhipu's stock price stood at HK$1,433 per share, with a total market capitalization of HK$638.4 billion (approximately $81.4 billion).

As an offshore financial center, Hong Kong has provided Zhipu and MiniMax with public pricing from top-tier global long-term capital. Sovereign wealth funds from the Middle East and Europe, including the Abu Dhabi Investment Authority, Singapore's GIC, and Norway's central bank, have made significant allocations to China's AI sector. The Hang Seng Tech Index announced on May 22 that both companies would be simultaneously included as constituent stocks, and once the relevant rules take effect on June 8, substantial passive index funds will automatically lock in allocations according to index methodology.

What A-shares confer, meanwhile, is the strategic positioning of hard-tech enterprises in this era. The STAR Board's sector positioning means official recognition as a national strategic technology asset — a label that influences government procurement lists, the willingness of central state-owned enterprises to collaborate, and priority access to computing power quotas at critical moments. Long-term capital from social security funds, insurers, national teams, and local government state-owned funds represents genuine patient capital for AI companies that remain loss-making in the long term. From computing power quotas to data compliance, from government procurement to industry standard-setting, the industrial resources accessible to STAR Board-listed companies are irreplaceable.

The Hang Seng AH Premium Index has fallen from its long-standing level above 130 points to 118.52, with some hard-tech leaders even showing rare AH price inversions. After two years of intense competition, global capital is, for the first time, willing to assign higher valuations and pay excess premiums for China's top hard-tech assets in the Hong Kong market outside of A-shares. The two capital markets are no longer shadows or discounted versions of each other — each performs its own function, bearing different strategic foundations.

Against the backdrop of accelerating global AI model capitalization, on the same day Zhipu announced its A-share return, US AI company Anthropic confirmed it had confidentially submitted a draft S-1 registration statement to the US SEC on June 1 in preparation for a proposed initial public offering. The capital race among global large language model companies is unfolding on multiple fronts simultaneously.

Zhipu's proposed A-share issuance of 15 billion yuan far exceeds the approximately 3.6 billion yuan (approximately $532.4 million) it raised during its Hong Kong IPO at the start of the year. From securing global valuations in Hong Kong to obtaining national strategic positioning in A-shares, this "Hong Kong first, then A-share" path is becoming the standard paradigm for China's top AI enterprises. Behind Zhipu and MiniMax, the capital path choices for unicorns such as Moonshot AI, 01.AI, Stepfun, and Baichuan AI now have a clear reference coordinate.