Benchmarking Palantir and MiniMax, how does Haizhi Technology(02706) justify supporting a trillion-dollar market cap expectation?

The Hong Kong stock market at the start of 2026 continues to see AI new stocks as the core focus of capital chasing. Following the explosive popularity of stocks like Biran Technology, MiniMax, and Zhipu AI, Haizhi Technology (02706), dubbed the “First Stock of Image-Model Fusion” and the “Chinese version of Palantir,” surged 242.2% on its first trading day. The next day and third day saw continued gains of 29.6% and 28.4%, with increased trading volume, and secondary market momentum far exceeding market expectations.

Market discussions mainly focus on its “AI Deception Removal” niche label, but few delve into why industry capital and top institutions like Zhipu Capital are willing to pile in as cornerstone investors.

This is not merely following the AI trend, nor is it baseless hype.

Haizhi Technology’s ability to stand out in the red ocean of general large models—within an industry saturated with competition—relies on its early, precise focus on the core contradictions of large model industry deployment. With a differentiated strategic vision and unique technical logic, it has built a competitive moat far beyond its peers in the long and snow-covered 2B industrial AI track. Its value proposition and development path are highly aligned with global enterprise AI leader Palantir, validating the company’s long-term growth potential. Driven by benchmarks like MiniMax, Haizhi Technology has already opened up a trillion-yuan market cap imagination space.

Strategic First-Mover: Securing an Irreplaceable Vertical Entry Point to Bridge the Core Gap in Large Model 2B Deployment

Recently, Stanfordberg, founder of open-source AI intelligence system Open Claw, made a statement that caused a stir in the global AI industry—he publicly predicted that 80% of apps will disappear in the future, and personal AI assistants will become the core force replacing these applications. He also clarified that applications in specialized vertical fields such as healthcare, finance, and law will be among the few core categories unlikely to be fully replaced by general AI.

This statement hits the core contradiction in the current AI industry: the internal competition and wave of replacement in C-end general scenarios are surging, while the B-end professional vertical fields are the true core entry points that AI will not easily overturn and that hold long-term commercial value.

To date, the industry consensus is that the technology race for general large models has entered a homogenized, internal competition stage. C-end traffic dividends have peaked, and price wars are intensifying. The real determinant of long-term enterprise value lies in the ability to deploy vertical scenarios in the 2B sector. However, a persistent pain point is the insurmountable gap between general large models and industry scenarios—while these models have powerful semantic understanding and reasoning capabilities, they lack internal business logic and data processing logic, cannot automate personalized needs in vertical scenarios, and cannot solve the “zero tolerance” problems in industries like finance, government affairs, and energy.

Haizhi Technology’s core strategic foresight lies in its early 2023 September proposal of the “Image-Model Fusion” technical path, preemptively recognizing the development boundaries of general large models and the entry value of vertical scenarios. Instead of following the crowd into the costly internal competition of general large models, it chose to build the “roads and bridges” between all large models and industry vertical scenarios, precisely occupying this core track unlikely to be replaced by general AI.

According to the prospectus, Haizhi’s Atlas intelligence system is compatible with over 100 large language models, including open-source and commercial models like DeepSeek and Tongyi Qianwen, as well as industry-specific models such as State Grid’s Guangming large model.

Haizhi’s orders are highly “defensive,” as its systems are often embedded directly into clients’ core business flows, creating high customer stickiness and switching barriers. In 2024, the company’s government clients’ renewal rate reached 92%. It added three state-owned banks, eight provincial government clients, and four leading internet giants, with finance, government, and internet clients accounting for over 80%. For core clients like banks and government agencies, switching vendors involves a 1-2 year migration cycle, costs exceeding tens of millions, and significant business risks. This strong lock-in property has helped Haizhi build a solid foundation.

This irreplaceable vertical scenario entry point is also why Zhipu AI chose to become Haizhi’s cornerstone investor. For general large model providers, Haizhi’s 360+ enterprise clients and numerous deployment scenarios are essential core entry points.

Deep cooperation with Zhipu AI allows Haizhi to quickly capture every update in large model technology, achieve bidirectional adaptation of industry scenario needs and large model capabilities, and form a business closed loop of “scenario data iteration model capability, model upgrades feeding scenario optimization,” creating a competitive ecological moat in the industrial AI track.

Technical Barrier: Benchmarking Against Global Leaders to Build a Local Decision-Making Operating System for Chinese Scenarios

If strategic foresight has enabled Haizhi to seize the AI industry’s long-term growth wave, its over a decade of expertise in knowledge graphs and graph computing has built an uncopyable technological moat. The long-term value of this system has been fully validated overseas—its underlying logic and commercial deployment capabilities are comparable to global enterprise AI leader Palantir; its graph-based enterprise decision context construction aligns closely with the approach of Glean, a prominent overseas company.

Palantir, a three-year, 35-fold super stock, claims to have “built a modern enterprise decision-making operating system,” with a current market cap of $322.3 billion. The core similarities between Haizhi and Palantir are reflected in two dimensions:

First, in technical approach: both focus on graphs, with large models as auxiliary, constructing enterprise ontologies of business and data assets through knowledge graphs, forming a comprehensive enterprise intelligence management panorama. This allows AI to deeply participate in core business functions (assist or autonomous analysis, decision-making, and execution) and meet the full capabilities of enterprise-level AI operating systems.

Second, in business model: both adopt a “2B + 2G” dual-driven approach, focusing on deep vertical scenario applications (not standardized products), achieving high growth and profitability inflection points.

Haizhi’s second growth curve aligns precisely with the latest overseas AI industry trend—using context graphs to realize full-chain translation of enterprise decision data. Glean’s core strength is building enterprise-level contextual relationship systems via knowledge graphs, enabling AI to understand all internal data assets and implicit knowledge, supporting automated analysis, decision-making, and execution, laying the foundation for enterprise automation. Similarly, Haizhi’s Atlas graph system structures decision data across enterprise operations, production, and management, restoring the full context of decision logic in graph form, allowing large models not only to understand data and business logic but also to apply and execute—autonomous analysis, decision-making, supervision, and rapid, precise response.

Despite differences in primary scenarios (Palantir rooted in Western military and geopolitical contexts), Haizhi has established its own clear geopolitical and scenario moat: it focuses on domestic social governance, public services, finance, energy, and other refined management scenarios, as well as the digital standardization and intelligent automation transition of central and state-owned enterprises. Its product evolution revolves around data compliance, business abstraction, risk control, lean management, and decision scenarios tailored to China’s market needs, making it a truly localized decision-making operating system.

According to Frost & Sullivan, based on 2024 revenue, Haizhi ranks first among Chinese AI intelligent system providers centered on graphs, with about 50% market share, forming an absolute monopoly advantage.

Valuation Reassessment: Benchmarking MiniMax, a Hundred-Billion Market Cap Target with Trillion Growth Potential

Ultimately, market enthusiasm depends on fundamentals and valuation support. Currently, the valuation system for Hong Kong AI stocks shows that Haizhi is still severely undervalued.

JPMorgan Chase’s latest AI industry report explicitly sets the long-term value for China’s AI industry: the sector is shifting from the “hundred-model battle” to a new stage focused on commercialization and deployment capabilities. JPMorgan’s coverage of Zhipu and MiniMax assigns an “Overweight” rating, with target prices raised to HKD 400 and HKD 700 respectively by December 2026. Using bottom-up quantitative analysis, it predicts that by 2030, the global AI market will reach $1.4 trillion, with the B2B application market accounting for about $1.1 trillion—long-term growth engine of AI industry (see chart). This assessment significantly raises Haizhi’s valuation ceiling.

(Chart source: JPMorgan latest report, data as of February 9, 2026)

From core fundamentals: Haizhi’s revenue far exceeds that of listed top large model companies. In 2024, MiniMax achieved about RMB 220 million in revenue, Zhipu AI RMB 312 million, while Haizhi’s total revenue reached RMB 503 million, 2.28 times MiniMax and 1.61 times Zhipu.

In terms of growth: from 2022 to 2024, compound annual growth rate (CAGR) was 26.8%. In the first half of 2025, revenue growth accelerated to 38.4%, with Atlas intelligent system revenue soaring 872.2% YoY in 2024.

More importantly, Haizhi’s valuation logic perfectly aligns with JPMorgan’s core AI valuation framework, with strong certainty support.

First, precise track positioning: JPMorgan emphasizes that B2B AI applications are the core value pool of the global AI market, three times the size of C-end markets. Haizhi’s focus on government, enterprise, finance, energy, and transportation verticals are the most promising high-value areas for AI commercialization, avoiding the price wars of C-end general models.

Second, solid commercialization barriers: unlike many AI companies still in early tech monetization stages, Haizhi has achieved large-scale deployment with over 360 enterprise clients, forming a healthy business cycle of “high unit price + high renewal rate,” fully aligning with JPMorgan’s view that “long-term value depends on commercialization capability.”

Third, ecosystem synergy unlocking growth potential: deep strategic cooperation with Zhipu AI allows sharing of large model iteration dividends and rapid scenario expansion via Zhipu’s developer ecosystem, highly consistent with JPMorgan’s favored “API monetization + global ecosystem expansion” growth path.

Fourth, incremental capital certainty: the company fully meets the quarterly inclusion standards for Hong Kong Stock Connect, and being listed will attract continuous inflows of mainland capital, creating a positive valuation and stock price cycle.

Fifth, clear profit inflection point: Haizhi’s adjusted net profit turned positive in 2024. As Atlas business scales, economies of scale will continue to release, profitability will improve, and it is expected to achieve full profitability earlier than Zhipu and MiniMax.

Initially, Haizhi’s static PS ratio was only 20x; even after recent surges, the static PS remains just over 100x, leaving ample room for valuation catch-up and aiming for a trillion-yuan market cap.

Recently, UBS, more optimistic than JPMorgan, set a target price of HKD 1,000 for MiniMax. UBS’s valuation is based on 125x P/ARR (Price/Annual Recurring Revenue), applied to the December 3, 2026, ARR of $318 million (see table), cross-validated with an SOTP approach.

(Chart source: UBS research report)

If we conservatively apply UBS’s 125x P/ARR to maintain 38.4% YoY growth in 2025 and 2026, Haizhi’s reasonable market cap at the end of 2026 should be around RMB 120 billion.

In the long-term growth wave of the global AI industry as defined by JPMorgan, the revaluation of B-end applications has just begun, and Haizhi’s market cap approaching that of Zhipu and MiniMax—trillions of yuan—is highly probable.