Haiching Zhiyuan completes Hong Kong IPO at 7.20 HK$ per share; stock opens up 314% to 230 billion HK$ market cap.
On June 22, Shenzhen Heaqing Zhiyuan (01392.HK) officially commenced trading on the Hong Kong Stock Exchange. The IPO price was HKD 7.20, with the stock opening up 313.89%, and a market capitalization of HKD 23 billion. Heaqing Zhiyuan is hailed as the "first multi-spectrum AI stock" and has been established in Shenzhen for 13 years.
Notably, Heaqing Zhiyuan's IPO included no cornerstone investors, no stabilizing agents, and no greenshoe options. According to the prospectus, the company has turned profitable on an accounting basis, yet operating cash flow is leaking substantially. Its most profitable segment—"AI large model services"—shows declining gross margins as revenue grows. Heaqing Zhiyuan presents six key investment considerations:
**Revenue and Profit Performance**
From 2023 to 2025, revenues were RMB 117 million, RMB 523 million, and RMB 669 million respectively, representing a three-year compound growth rate of 138.9%. In 2024 alone, revenue surged 346.4% year-over-year, followed by a 27.9% increase in 2025. The revenue base expanded roughly 5.7-fold over three years.
The company achieved "turnaround to profitability": after a loss of RMB 18.413 million in 2023, it swung to a profit of RMB 40.412 million in 2024. However, net profit declined rather than increased in 2025, falling to RMB 29.354 million. Why would profit decline for a company going public with growing revenues?
The company attributed this to "primarily increased general administrative and research and development expenses." Examining adjusted figures clarifies matters: in 2025, adjusted net profit added back RMB 8.465 million in share-based payments and RMB 17.426 million in one-time expenses.
**Cash Flow vs. Accounting Profit**
Looking purely at the profit statement, Heaqing appears to be a growth stock that "turned profitable in 2024 and remained profitable in 2025." But the cash flow statement tells a different story. In 2023, operating activities generated roughly RMB 70 million; in 2024, this turned negative, and by 2025, operating cash flow showed a net outflow of RMB 130 million—while accounting net profit remained positive at RMB 29.354 million.
Where did the money go? The prospectus explains that 2025's outflow was primarily driven by working capital changes—accounts receivable increased by RMB 173 million (including approximately RMB 147 million in increased prepayments to certain suppliers to ensure stable supply of HPC servers amid strong near-term demand) and inventories rose by RMB 55.5 million.
**The AI Large Model Services Story**
The AI large model services segment achieved commercialization and recorded revenue for the first time in 2024, then surged to RMB 355.4 million in 2025, accounting for more than half of total revenue at 53.1%. This expansion forms the core of the company's transformation narrative.
However, gross margin for AI large model services declined from 49.5% in 2024 to 30.4% in 2025. Why would margins compress as this "AI services" business scales?
The prospectus reveals that in 2024, AI large model service projects were primarily hardware-agnostic; by 2025, the vast majority had become integrated hardware-software solutions, with specialized equipment consisting mainly of high-performance computing (HPC) servers. Direct materials and supplies as a percentage of cost of sales rose steadily from 69.9% in 2023 to 91.9% in 2025. In other words, more than 90% of costs are spent on procuring raw materials and equipment.
Thus, the 2025 "AI large model services" revenue of 53% comprises more than 70% as integrated projects bundled with server sales, with hardware priced at near-market rates. While this business includes proprietary "Heaqing Origin Large Models" and algorithms (genuine competitive advantages that have passed the Cyberspace Administration's algorithm registration and been selected as Guangdong Province's first batch of industry-specific large models), the main driver of 2025's doubled revenue stems substantially from "server resale and custom integration" rather than pure high-margin software. This also explains why, despite growing profits, the company needed to prepay RMB 147 million to stockpile HPC servers.
Prepayments to suppliers rose from RMB 2.73 million in 2023 and RMB 25.09 million in 2024, to RMB 156 million by year-end 2025. On December 31, 2025, the company made one-time prepayments of approximately RMB 146.9 million to four suppliers "to ensure stable and timely supply of equipment (primarily HPC servers in high demand in recent markets) required for multi-spectrum AI large model services and perception terminals." Supplier K and Supplier D alone accounted for RMB 66.9 million (42.8%) and RMB 21.6 million (13.8%) respectively. For a company with net profit of RMB 29.354 million, prepayments on supplier accounts equal five times its annual profit.
Additionally, trade receivables turnover days were 68, 60, and 95 days for 2023, 2024, and 2025 respectively.
**Customer Concentration**
Heaqing's three-year revenue expansion of more than 5-fold is evident from top-five customer concentration: 38.3% in 2023, 59.0% in 2024, and 46.8% in 2025.
The single largest customer contributed RMB 23 million (19.6%) in 2023, surged to RMB 186 million (35.5%) in 2024, and declined to RMB 94 million (14.1%) in 2025.
In 2024, the largest customer alone drove 35.5% of revenue, serving as the engine for that year's three-fold revenue explosion. In 2025, this engine slowed, with AI large model services taking the lead instead.
However, AI large model services operate on a project basis with revenue recognized upon customer project acceptance. Among 41 AI large model projects in 2025, 26 came from unlisted private enterprises, contributing RMB 294 million (82.7%); only 4 projects from listed private customers contributed RMB 12.4 million (3.5%) combined. Project-based models offer strong revenue bursts but uneven streams and questionable sustainability—explaining the wild swings in accounts receivable and inventories.
Supply chain concentration mirrors customer concentration: top five suppliers account for 54% to 66% of purchases, with the largest supplier alone accounting for 39.9% in 2024. The company depends on external suppliers for core components (microprocessors represent 15-30% of costs, CMOS sensors 10-20%), limiting negotiating power.
**Valuation Acceleration**
Valuation accelerated dramatically in 2025: the C+ round (December 2024) valued the company at only RMB 1.01 billion; by the D round (July 2025), valuation jumped to RMB 3.55 billion, a 3.5x increase in six months—precisely coinciding with AI large model services revenue explosion. The RMB 3.55 billion D-round valuation occurred just one year before the IPO. The IPO offering price corresponded to a market cap of approximately HKD 5.57 billion, only about 36% higher than the D-round post-investment valuation; however, at the grey market price, market cap surged to approximately HKD 20.2 billion, roughly 4.9 times the D-round valuation.
A-round per-share cost was approximately RMB 0.57—representing a 90.8% discount to the offering price, with book gains of approximately 10-fold; at the grey market price of RMB 22.5, gains reached approximately 38-fold.
D-round at RMB 5.15 per share—only a 17% discount to the offering price. The investor "Zhishijiu No.9" entered in July 2025, getting the timing right but at already-elevated costs.
**Early Winners**
The early winners in A and B rounds were the Zhejiang Merchants Venture Capital system (Taolyue New Energy, Taolyue Xinwang, Jinlan Sunshine, collectively holding 14.86%) and the Shenzhen Hi-Tech Investment Group system (Shenzhen state-owned capital under Shenzhen Investment Holding, collectively 4.54%). Taolyue New Energy, led by Zhejiang Merchants Venture Capital, entered the A-round at approximately RMB 0.57 per share, making it one of the biggest winners in discount and returns.
**Management and Ownership**
Heaqing's founder, chairman, and general manager is Zhou Bo, who has devoted approximately 26 years to computer vision and thermal imaging. He graduated in 1999 from Xi'an University of Science and Technology (formerly Xi'an Mining College) with a degree in automatic control, subsequently serving as a video research engineer at Chengdu Coli Electronics Research Institute, then as product manager at Tu Min Intelligent Video (833318), and general manager and director at Shenzhen Yellow River Digital Technology Co., Ltd. from 2005 to 2012, before founding Heaqing in 2013.
His core team consists of longtime colleagues from "Xi'an University of Science and Technology and Yellow River Digital": Executive Director and Vice General Manager Miao Rui (age 40, computer science degree from Xi'an University of Science and Technology, former Yellow River Digital employee), Executive Director and Supply Chain Director Zou Xiaogang (age 43), and Executive Director and Technology Expert Chen Yonggang (age 54, with 28 years of technical experience)—most joined when the company was founded in 2013.
Through three employee stock ownership platforms—Zhongcheng Tianyi, Zhongzheng Tianyi, and Zhongzhi Tianyi—where he serves as sole executive partner, plus direct holdings, Zhou Bo controls 48.87% of voting rights pre-IPO.
**Market Position and Outlook**
Heaqing Zhiyuan's story is that of a technology company that has become China's industry leader in multi-spectrum AI, securing position in a high-growth sector (China's multi-spectrum AI market valued at approximately RMB 20 billion in 2025, projected to reach RMB 79.4 billion by 2030, with 31.8% CAGR). Its proprietary large models, edge-side lightweight capabilities, and "National Specialized, Refined, Distinctive, Innovative Small Giant" designation represent genuine competitive credentials.
However, "first place" requires proper contextualization—according to Frost & Sullivan, Heaqing ranked first among Chinese multi-spectrum AI enterprises by revenue in 2025 with only a 3.3% market share; the top five collectively hold 8%, representing "relative first" in a highly fragmented market.
A technical team from Xi'an University of Science and Technology spent thirteen years turning "equipping devices with super-smart vision" into Hong Kong's hottest physics AI IPO. What answer the market will provide remains to be seen.