--- title: "After a 41-fold increase in large model revenue, what other financial questions does Extreme Perspective still need to answer?" type: "Topics" locale: "en" url: "https://longbridge.com/en/topics/42977915.md" description: "On July 27, Extreme Vision released a positive profit forecast for the first half of 2026. The company expects revenue to range from 153.4 million to 165.2 million yuan, representing a year-on-year increase of 160% to 180%. Adjusted net profit is projected not to exceed 6 million yuan, compared to an adjusted loss of 30 to 35 million yuan in the same period last year. Due to share-based payments and listing expenses, the book net loss remains between 38 million and 44 million yuan. Revenue from large model solutions is expected to exceed 80 million yuan, marking a growth of over 4110% compared to 1.9 million yuan in the same period last year, surpassing the full-year 2025 figure of 66.2 million yuan..." datetime: "2026-07-28T04:07:19.000Z" locales: - [en](https://longbridge.com/en/topics/42977915.md) - [zh-CN](https://longbridge.com/zh-CN/topics/42977915.md) - [zh-HK](https://longbridge.com/zh-HK/topics/42977915.md) author: "[港股研究社](https://longbridge.com/en/profiles/3199113.md)" --- # After a 41-fold increase in large model revenue, what other financial questions does Extreme Perspective still need to answer? On July 27, JiShiJue (Extreme Perspective) released a positive profit forecast for the first half of 2026. The company expects revenue to range from 153.4 million to 165.2 million RMB, representing a year-on-year increase of 160% to 180%. Adjusted net profit is projected not to exceed 6 million RMB, compared to an adjusted loss of 30-35 million RMB in the same period last year. Due to share-based payments and listing expenses, the reported net loss remains between 38 million and 44 million RMB. Revenue from large model solutions is expected to exceed 80 million RMB, a growth of over 4110% compared to 1.9 million RMB in the same period last year, surpassing the full-year 2025 figure of 66.2 million RMB. JiShiJue has completed preliminary validation on the market demand side. The subsequent focus needs to shift from revenue growth speed to customer structure, gross margin levels, delivery reusability, and cash collection. It is not unusual for industrial AI companies to receive projects; it is the continuous accumulation of project experience into products that forms stable operational quality. **Purchasing entities in industrial AI are expanding from single-point algorithms to closed-loop tasks.** Tianyancha data shows that JiShiJue's original business covers standard computer vision, customized computer vision, and software-defined integrated AI solutions. The large model business overlays customer industry knowledge onto general models, combining multi-agent optimization, retrieval-augmented generation, and scenario-specific algorithms to provide customized applications for enterprises. The company's product roadmap aligns with the industrial policy direction of 2026. The "Implementation Opinions on Intelligent Agents" lists perception, memory, decision-making, interaction, and execution as basic capabilities of intelligent agents, and includes smart manufacturing, energy resources, and transportation as key scenarios. Beijing's industrial internet plan further proposes "platform + scenario intelligent agents," emphasizing the combination of standardization and customization, as well as "develop once, adapt to multiple scenarios." This grants visual AI companies a broader product boundary. Traditional computer vision excels at identifying image content, while enterprise intelligent agents must also call upon knowledge bases, business systems, and execution tools to handle tasks after identification. JiShiJue's accumulated visual algorithms, delivery platforms, and industry scenarios enable the company to connect industrial images, cameras, and enterprise processes. Manufacturing, transportation, and energy possess vast amounts of visual data, making it easier to set acceptance conditions around accuracy, response speed, and process improvement. In 2025, JiShiJue's large model solution revenue increased from 62.1 million to 66.2 million RMB, a 6.5% increase, with the number of customers growing from 1 to 12. The gross margin of the large model business rose from 22.8% to 36.6%. While the revenue growth rate was modest, the customer base and delivery economics have already improved. With H1 2026 revenue exceeding 80 million RMB, the expansion of the customer base is beginning to contribute larger absolute increments. Compared to the low-base growth of 41 times, half-year revenue exceeding the previous full year better reflects the speed of business advancement. Enterprise customers are reducing model trials disconnected from business processes, with budgets gradually flowing towards solutions that can be deployed, accepted, and operated continuously. JiShiJue covers vision, multimodal models, and intelligent agents simultaneously, potentially expanding the service boundary of individual projects and creating room for increasing per-customer revenue. The constraints are equally clear. Industry customers require data governance, system adaptation, and on-site implementation. The deeper the product penetrates the production process, the heavier the delivery responsibility. Insufficient standardization will cause revenue to rise synchronously with labor and computing power costs. The company has proven that customers are willing to pay for industry-specific large models; subsequent operational elasticity depends on how many times the same set of capabilities can be reused. **A 41x growth rate needs to be understood within the context of project acceptance.** JiShiJue's large model revenue in H1 2025 was only 1.9 million RMB, reaching 66.2 million for the full year. Calculated based on public data, H2 2025 contributed approximately 64.3 million RMB, accounting for about 97% of the full-year large model revenue. Large model projects still carry a distinct acceptance rhythm: the company fulfills relevant performance obligations after delivering the solution and obtaining customer acceptance. The early surge in H1 2026 revenue is a positive change, making whether orders are more balanced and delivery cycles shortened key points for the next interim report. Cash flow quality is harder to avoid. At the end of 2025, JiShiJue's net trade receivables and notes receivable were 232.6 million RMB, compared to 178.0 million RMB at the end of 2024. The impairment provision for trade receivables increased from 7.461 million to 22.787 million RMB. In 2025, net cash outflow from operating activities was 36.904 million RMB, compared to 17.592 million RMB in 2024. Contract liabilities totaled approximately 16.36 million RMB during the same period, lower than the approximately 23.04 million RMB at the end of 2024. Beyond revenue growth, changes in receivables, impairments, and advance receipts reflect project bargaining power, customer willingness to pay, and working capital pressure. JiShiJue disclosed that large model contracts typically require payment of 10% to 30% of the contract amount within 30 days of signing, with the balance due within 30 days after delivery and customer acceptance. The latest forecast has not yet disclosed the number of customers, single-project proportion, changes in receivables, or contract liabilities for H1 2026. The company is expected to release its mid-year results by the end of August. The most informative content then will be how many customers contributed to the 80 million RMB revenue, whether more advance receipts were formed, and if the collection speed has improved. Front-loaded revenue recognition provides a positive signal. If the number of customers continues to increase and the growth rate of receivables is lower than that of revenue, the order visibility of the large model business will become clearer. If growth is driven mainly by a few large projects, performance volatility will still retain the characteristics of project-based software companies. **Profit repair has emerged, but R&D efficiency is still competing for profits.** JiShiJue's income statement presents two pictures. On an adjusted basis, it improved from a loss of 30-35 million RMB in the same period last year to a net profit not exceeding 6 million RMB, indicating that business expansion is starting to release operating leverage. Under International Financial Reporting Standards, a loss of 38-44 million RMB is still expected, with share-based payments and listing expenses continuing to weigh on book profits. The announcement did not split the two expense items for H1 2026, so the true profit margin of the main business still awaits the interim report. 2025 data explains this tension. JiShiJue's full-year revenue was 292.1 million RMB, a 13.5% year-on-year increase, with overall gross margin rising from 40.2% to 41.5%. R&D expenses increased from 44.8 million to 85.1 million RMB, an 89.8% increase, mainly used for cloud services and data procurement required for large model projects. Listing expenses were 16.947 million RMB, and share-based payments were 18.494 million RMB. JiShiJue had a profit of 8.708 million RMB in 2024, turning into a loss of 45.778 million RMB in 2025. Large model projects increase revenue and gross margins but also add technical service, computing power, and organizational costs. JiShiJue needs to find a balance between R&D intensity and product reusability: too little R&D investment may weaken model iteration, while maintaining high investment long-term will suppress profits. Whether investments can form standard modules, industry knowledge bases, and universal toolchains is more explanatory than the level of expenses in a single year. The 2026 forecast gives directions for improvement. The company stated that the productization and standardization construction of AI solutions have improved delivery efficiency and driven improvements in revenue structure and gross margin levels. The 13.8 percentage point increase in the gross margin of large models in 2025 also indicates that project experience can be translated into lower per-project costs. The industry engineering layer chosen by JiShiJue requires simultaneous understanding of models, visual algorithms, customer data, and business processes. Related capabilities are difficult to replicate quickly simply by expanding computing power. Industry knowledge, data processing experience, and customer trust formed through long-term delivery constitute more practical competitive barriers. Future financial reports will test this process. A more ideal combination would be sustained growth in large model revenue, steadily improving gross margins, slowing growth in receivables, and continued positive adjusted profits. JiShiJue has already advanced the commercialization of large models to a scale where half-year revenue exceeds 80 million RMB; operational quality has become the main issue for the next stage. JiShiJue has proven that industry customers are willing to pay and preliminarily demonstrated that standardization can lift gross margins. Customer dispersion, cash collection, and R&D efficiency will determine whether growth can settle into long-term capabilities. Industrial AI will continue to compete on model capabilities, but the distance between enterprises will come more from whether solutions can be repeatedly delivered, stably collected, and run long-term in customer production systems. JiShiJue has crossed the first leg of commercialization validation. The real test brought by 80 million RMB is whether it can transform a one-time revenue leap into a sustainable operational order. ### Related Stocks - [06636.HK](https://longbridge.com/en/quote/06636.HK.md)