The Hong Kong stock market in early 2026 continues to see AI-focused new listings as a primary driver of capital interest. Following fervent investor enthusiasm for companies like Biren Technology, MiniMax, and Zhipu AI, HAIZHI TECH GP (02706), hailed as the "pioneer in graph-model integration" and the "Chinese version of Palantir," surged 242.2% on its debut. The stock extended gains by 29.6% and 28.4% on the second and third trading days respectively, with significant trading volume indicating secondary market demand far exceeding expectations.
While retail investors are largely attracted by its positioning in the niche "AI hallucination elimination" sector, the strategic rationale behind investments from industrial backers like Zhipu AI and Wuji Capital as cornerstone investors warrants deeper analysis. This support is not merely speculative trend-following but reflects HAIZHI TECH GP's unique competitive positioning. Amid intense competition in general-purpose large language models, the company has carved out distinct capital appeal and a robust growth trajectory by identifying and addressing the core challenge of industrial AI implementation years ahead of the broader industry. Its differentiated strategy and proprietary technological approach have established formidable barriers in the enterprise AI sector, a field with long-term potential.
The company's value proposition and development path closely align with Palantir Technologies Inc., the global leader in government and enterprise AI, underscoring its long-term growth prospects. From a valuation perspective, spurred by peers like MiniMax, HAIZHI TECH GP has ignited expectations for a potential trillion-dollar market capitalization.
**Strategic Foresight: Securing an Irreplaceable Vertical Niche to Bridge the AI Implementation Gap**
A recent assertion by Open Claw's founder, Steinberg, sent ripples through the global AI industry, predicting that 80% of current applications could disappear, replaced by personal AI agents. However, he highlighted specialized verticals like healthcare, finance, and law as domains likely to remain essential and resistant to full displacement by general AI. This insight underscores a central industry dilemma: while consumer-facing AI faces saturation and substitution, B2B vertical applications represent the sustainable, high-value frontier.
Market consensus holds that competition in general-purpose LLMs has become homogenized, with consumer market growth plateauing and price wars intensifying. The critical differentiator for long-term enterprise value is now the capability to deploy AI in specific industrial contexts. A significant challenge persists: a substantial gap exists between general LLMs' semantic understanding capabilities and the specific operational logic, data structures, and stringent zero-error tolerance requirements of sectors like finance and government. General models lack inherent understanding of internal business processes and cannot autonomously handle specialized, high-stakes tasks.
HAIZHI TECH GP demonstrated prescience by introducing its "graph-model fusion" technical pathway as early as September 2023, anticipating the limitations of general LLMs and the strategic importance of vertical market access. Instead of joining the capital-intensive race for general model development, the company positioned itself as the essential "bridge" connecting various LLMs to industrial applications, securing a vital and defensible niche.
According to its prospectus, HAIZHI TECH GP's Atlas agent is compatible with over 100 LLMs, including open-source and commercial models like DeepSeek and Tongyi Qianwen, as well as industry-specific models such as the State Grid's Bright model. The company's contracts exhibit strong defensive characteristics, often embedding directly into clients' core business workflows, creating high customer stickiness and significant switching costs. In 2024, its government client renewal rate reached 92%, with new clients including three major state-owned banks, eight provincial government entities, and four leading internet firms. Clients from finance, government, and internet sectors collectively constitute over 80% of its business.
For core clients like banks and government agencies, replacing HAIZHI TECH GP's systems could entail migration periods of 1-2 years, costs exceeding tens of millions, and substantial operational risk. This strong lock-in effect provides a stable business foundation. This irreplaceable vertical market access is precisely why Zhipu AI chose to become a cornerstone investor. For LLM developers, HAIZHI TECH GP's portfolio of 360+ enterprise and government clients and over 100 deployment scenarios represents a crucial channel for real-world implementation.
For HAIZHI TECH GP, deep collaboration with partners like Zhipu AI ensures it remains at the forefront of LLM advancements, enabling a virtuous cycle where real-world scenario data iteratively improves model capabilities, and model upgrades, in turn, enhance application performance. This creates a powerful business moat in the industrial AI landscape.
**Technical Moat: Building a Native Decision-Making OS for China, Aligning with Global Leaders**
If strategic foresight positioned HAIZHI TECH GP for the industrial AI wave, its decade-long expertise in knowledge graphs and graph computing constitutes a formidable and difficult-to-replicate technical moat. The long-term value of this technological architecture is validated by global counterparts. Its underlying logic and commercial deployment capabilities are comparable to Palantir Technologies Inc., while its approach of using graphs to construct enterprise decision-making context aligns closely with the "context graph" methodology of the promising overseas firm Glean.
Palantir Technologies Inc., a stock that surged 35-fold over three years and is renowned for building a "modern enterprise decision-making OS," currently boasts a market cap of $322.3 billion. HAIZHI TECH GP shares key similarities on two fronts: Technically, both center on graphs augmented by LLMs, using graphs to model enterprise data and business ontologies, creating a comprehensive map for intelligent management. This allows AI to deeply integrate into core operations for analysis, decision-making, and execution, fulfilling the role of an enterprise AI operating system. Commercially, both pursue a dual B2B and B2G strategy, focusing on deep vertical integration rather than standardized products, achieving high growth alongside profitability inflection points.
HAIZHI TECH GP's secondary growth curve also aligns with the emerging trend of using context graphs for end-to-end translation of enterprise decision data. Glean's core competence lies in building enterprise-wide contextual associations via knowledge graphs, enabling AI to comprehensively understand internal data assets and implicit decision-making knowledge, thereby facilitating accurate and efficient automated analysis and execution. Similarly, HAIZHI TECH GP's Atlas system can structurally decompose decision-making data across business, production, and management processes, reconstructing the complete contextual logic of enterprise decisions in graph form. This allows LLMs not just to understand data and business logic but to apply it—autonomously analyzing, deciding, and overseeing execution with precision and speed.
While the primary service domains differ from Palantir Technologies Inc. and Glean, HAIZHI TECH GP has established a clear geo-specific and scenario-specific moat. It deeply serves China's social governance, public services, financially sensitive commercial sectors, and the digital-to-intelligent transformation of state-owned enterprises. Its product evolution is tailored to the data compliance, business abstraction, risk control, and decision-making needs of Chinese enterprises, making it a genuinely native decision-making operating system for the domestic market.
According to Frost & Sullivan data, based on 2024 revenue, HAIZHI TECH GP ranks first among graph-centric AI agent providers in China, holding approximately a 50% market share, indicating a dominant position.
**Valuation Re-rating: Benchmarking Against MiniMax, Mapping a Path from Billions to Trillions**
Ultimately, market enthusiasm must be underpinned by fundamentals and valuation. Within the current Hong Kong AI stock valuation framework, HAIZHI TECH GP appears significantly undervalued. J.P. Morgan's recent AI industry report emphasized the long-term value of China's AI sector, transitioning from a "hundred-model battle" to a phase prioritizing commercial deployment capability. The report initiated coverage on Zhipu AI and MiniMax with "Overweight" ratings, setting target prices of HKD 400 and HKD 700 respectively by December 2026.
The bank's bottom-up analysis projects the global AI market to reach $1.4 trillion by 2030, with B2B applications accounting for approximately $1.1 trillion, serving as the core growth engine. This outlook significantly raises the valuation ceiling for companies like HAIZHI TECH GP.
Examining core fundamental metrics, HAIZHI TECH GP's revenue substantially exceeds that of listed LLM leaders. In 2024, MiniMax reported revenue of approximately RMB 220 million, while Zhipu AI reported RMB 312 million. In contrast, HAIZHI TECH GP's total revenue reached RMB 503 million, 2.28 times that of MiniMax and 1.61 times that of Zhipu AI. Growth is also robust, with a revenue CAGR of 26.8% from 2022 to 2024. Revenue growth accelerated to 38.4% year-on-year in H1 2025, driven largely by the core Atlas agent business, which saw revenue surge 872.2% YoY in 2024.
Crucially, HAIZHI TECH GP's valuation thesis aligns strongly with J.P. Morgan's AI pricing framework. First, its strategic niche is precise; J.P. Morgan identifies B2B AI applications as the core value pool, a market three times larger than the consumer segment. HAIZHI TECH GP's focus on government, finance, and energy aligns perfectly with these high-value, commercially viable verticals, avoiding the price wars of the consumer LLM space. Second, its commercial barriers are solid. Unlike many AI firms in early monetization stages, HAIZHI TECH GP has achieved scaled deployment across 360+ clients, creating a virtuous cycle of high contract values and high renewal rates, consistent with the "commercialization capability dictates long-term value" thesis. Third, ecosystem synergies expand the growth ceiling. The deep partnership with Zhipu AI provides access to LLM advancements and leverages Zhipu's developer ecosystem for rapid scenario expansion, mirroring the "API monetization + global ecosystem growth" path favored by the bank. Fourth, incremental fund inflows are highly probable. The company meets the criteria for fast-track inclusion in the Southbound Stock Connect, which would channel sustained mainland capital into the stock. Fifth, the path to profitability is clear. The company achieved positive adjusted net profit in 2024. As the Atlas business scales, economies of scale should further enhance profitability, potentially allowing it to reach sustained profitability ahead of Zhipu AI and MiniMax.
At its IPO, HAIZHI TECH GP traded at a static P/S ratio of just 20x. Even after recent significant gains, its static P/S remains slightly above 100x, suggesting substantial room for valuation expansion towards a trillion-dollar market cap. Recently, UBS expressed even greater optimism than J.P. Morgan, assigning MiniMax a target price of HKD 1,000. This was based on a 125x P/ARR multiple applied to an estimated Annual Recurring Revenue of $318 million by December 2026, validated by a sum-of-the-parts analysis.
Applying a similar conservative 125x P/ARR multiple to HAIZHI TECH GP, and assuming it maintains a 38.4% growth rate through 2026, implies a reasonable market capitalization of approximately RMB 1.2 trillion by the end of 2026. As J.P. Morgan outlines a long-term growth cycle for global AI, the re-rating of B2B application value is just beginning. It appears highly probable that HAIZHI TECH GP's market capitalization will converge towards the trillion-yuan benchmarks set by peers like Zhipu AI and MiniMax.
Comments