ZhongAn Xinke's Second Resubmission: Enterprise AI Begins to Write Technical Capabilities into the Income Statement

Last week, Zhong'an Xinke submitted its IPO application to the Hong Kong Stock Exchange again. The new prospectus updates disclosed data through the first five months of 2026, showing revenue of 199.3 million yuan, a year-on-year increase of 62.7%; net profit of 4.16 million yuan, compared to a loss of 8.99 million yuan in the same period last year; and a comprehensive gross margin rising from 32.7% to 36.5%. From 2023 to 2025, the company's net profits were 10.08 million yuan, 33.23 million yuan, and 39.07 million yuan respectively, marking three consecutive full years of profitability.

Based on the above data, the outline of enterprise-level AI commercialization is gradually becoming clear: while large models have rapidly popularized technical capabilities, they have also raised acceptance standards for enterprise clients. While demo performance, model parameters, and the number of agents will naturally attract attention, long-term contracts still rely on business results. Specifically, we need to see if customer acquisition costs have decreased? If marketing conversion has improved, risk identification faster? And whether old systems can be integrated smoothly?

In the field of enterprise AI, technology suppliers must first help customers create value before having the opportunity to turn one-off projects into continuous procurement. Zhong'an Xinke's second filing resembles more of an implementation progress report for enterprise AI. Currently, scale growth for Zhong'an Xinke has emerged, but product reuse, customer retention, and cash collection still require further testing.

Enterprise AI is shifting from model procurement to process transformation, with delivery capability becoming a scarce asset

The prospectus cites Frost & Sullivan data indicating that China's enterprise-level AI solutions market grew from 19.1 billion yuan in 2021 to 69 billion yuan in 2025, and is expected to reach 474 billion yuan by 2030; the corresponding financial sector market is projected to reach 237 billion yuan. In 2025, Zhong'an Xinke ranked fourth among Chinese enterprise-level AI solution providers with vertical large model capabilities based on revenue.

Overall, the industry space is vast, but with nearly a thousand participants coexisting, competitive returns will certainly not be evenly distributed.

Facing data isolation, permission management, regulatory compliance, and legacy system transformation, enterprise clients typically procure consulting, model integration, knowledge engineering, process orchestration, on-site implementation, and subsequent operations and maintenance. Such projects have longer cycles than consumer-grade AI and stricter acceptance phases. The advantage is that once solutions enter marketing, customer service, risk control, or operational management processes, suppliers accumulate business understanding, process data, and migration costs, deepening client relationships.

Zhong'an Xinke treats intelligent marketing and intelligent operations management as two main lines. Solutions can integrate with clients' existing IT foundations, reducing the need for large-scale infrastructure transformation. Data shows that by the end of May 2026, the company had cumulatively served 409 clients, up from 88 at the end of 2023; the customer retention rate in 2025 was 53%, higher than 42.4% in 2024. Client numbers validate demand breadth, while retention rates are closer to commercial quality.

The company disclosed the formation of over 1,200 AI agents and more than 50 AI super assistants. Zhong'an Xinke already possesses a significant competitive advantage because while sheer numbers are easily 追赶 ed (caught up with), the industry knowledge, process data, and implementation experience behind these agents are harder to replicate.

Competition in enterprise AI will focus on two questions: How many clients can the same set of capabilities cover? How many scenarios can a single client expand into? If a marketing project can extend to client segmentation, rights operation, and product matching, then enter customer service, compliance, and internal management, revenue continuity will improve.

Intelligent marketing releases profits first, giving growth quality a more concrete pivot point

In the first five months of 2026, Zhong'an Xinke's intelligent marketing revenue was 113.8 million yuan, accounting for 57.1% of total revenue, compared to 58.37 million yuan in the same period last year; intelligent operations management revenue was 85.55 million yuan, accounting for 42.9%. The gross margin for intelligent marketing rose from 40.6% to 45.1%, while that for intelligent operations management was 25.1%. The prospectus attributes the improvement in intelligent marketing gross margin to more clients adopting self-operated channels, reducing reliance on traditional telecom outreach services.

Early enterprise AI was largely driven by customized projects, so revenue growth often accompanied simultaneous increases in outsourcing, on-site personnel, and delivery costs. Currently, Zhong'an Xinke's cost structure has changed: in the first five months, the proportion of external procurement in cost of sales dropped from 56.1% in the same period last year to 42.8%, while personnel costs rose to 56%.

Reduced dependence on external procurement helps preserve gross margins, while increased personnel investment indicates that scale expansion still requires heavy delivery resources.

The product structure changes are as follows: in the first five months of intelligent operations management revenue, system modules contributed 46.91 million yuan, exceeding service revenue of 38.64 million yuan; intelligent marketing remains primarily service-driven, with system module revenue at 12.24 million yuan. Intelligent operations management has a higher degree of modularization, while intelligent marketing shows stronger gross margin performance.

The two business lines have different focuses and can complement each other: marketing business is responsible for acquiring clients and contributing gross margins, while operations management modules are responsible for entering more operational links, extending the client lifecycle.

Data shows that R&D expenses for the first five months of 2026 were 43.8 million yuan, of which R&D personnel costs were 36.4 million yuan, mainly invested in the XK-QianAI and QianNexus agent matrices; meanwhile, sales expenses were 12.16 million yuan and administrative expenses were 15.29 million yuan. Profit was only 4.16 million yuan, related to the company still being in a phase of product investment and team expansion. Zhong'an Xinke's R&D investment has not contracted despite turning profitable. Subsequent operational efficiency depends on whether deployment cycles can be shortened, whether the proportion of module revenue can increase, and whether the dependency of new revenue on new personnel can decrease.

Zhong'an Xinke has a positive signal that is easily overlooked. The company's R&D expenses reached 92.59 million yuan in 2025, nearly 3.5 times that of 2024, yet it still achieved a net profit of 39.07 million yuan for the year. High R&D intensity has not yet crippled profitability; gross margin growth covered a considerable portion of investments. Enterprise AI companies generally face triple pressures of rapid product iteration, high model invocation costs, and delivery team expansion. Currently, Zhong'an Xinke has proven that revenue scale can absorb R&D investment; the next step is to prove this absorption capacity can be sustained.

Profitability has appeared; cash flow and client structure determine how far the business model can go

Data shows that operating cash flows for 2024 and 2025 were net outflows of 28.1 million yuan and 39.4 million yuan respectively, with a net outflow of 14.8 million yuan in the first five months of 2026. The prospectus attributes cash flow pressure mainly to increases in accounts receivable and contract fulfillment costs. Trade receivables stood at 199.2 million yuan by the end of May 2026, with an accounts receivable turnover of 149 days; as of the latest practicable date, 29.2% of ending receivables had completed subsequent collection.

Data from a single period is easily affected by project acceptance and collection rhythms. Enterprise client payment cycles are long, and sustained profitability ultimately needs to convert into cash. Zhong'an Xinke's improvements are already written into the income statement, but the cash flow statement still leaves problems to be solved.

Next, key observation should focus on whether the growth rate of accounts receivable is lower than revenue growth, whether modular products improve collection cycles, and whether operating cash flow can cover R&D and sales investments. Only when cash collection improves will book profits possess more solid quality.

Additionally, client concentration needs continued observation. Data shows: the revenue share of the top five clients dropped from 74.7% in 2023 to 43% in 2025, rebounding to 54.7% in the first five months of 2026; the largest client accounted for 25.1%. The annual decline indicates the client base is diversifying, while the short-term rebound shows major client projects still impact revenue rhythm. A more ideal structure would retain cooperation depth with top clients while increasing independent third-party clients, multi-industry clients, and repeat purchase revenue.

Major clients play a dual role for enterprise AI companies: they not only provide complex scenarios, data feedback, and industry endorsement, but also bring customization requirements, longer acceptance cycles, and stronger bargaining power. Zhong'an Xinke's future profit elasticity largely depends on whether the methodologies 沉淀 ed (accumulated) by top clients can be transformed into standard modules and sold to larger client groups with lower delivery costs.

Zhong'an Xinke completed Series B financing in 2025 with a post-money valuation of 2.215 billion yuan. Based on 2025 revenue of 477.1 million yuan, this corresponds to approximately 4.6 times price-to-sales ratio; based on 39.07 million yuan net profit, it corresponds to approximately 56.7 times price-to-earnings ratio.

The Series B financing price has already factored in AI growth attributes, but future performance still needs to digest the pricing. Because while revenue scale provides development space, gross margin, customer retention, module revenue, and cash collection truly determine enterprise quality.

In the future enterprise AI industry, suppliers lacking industry knowledge tend to stay at model integration, and companies highly dependent on human labor are constrained by delivery radius. Enterprises that can 沉淀 modules, embed them into workflows, and form continuous procurement are more likely to achieve stable per-capita output and cash flow.

Zhong'an Xinke's second filing begins to pull the evaluation standards for enterprise AI back to operations itself: models can only open possibilities, products and delivery turn possibilities into revenue, and ultimately, cash flow must prove this business can go further.