极视角 (06636.HK) 2026智通财经夏季路演大会
文章语言：
简
繁
EN
Share
Minutes
原文
会议摘要
The dialogue discussed the impact of big model and computer vision (CV) business on earnings, noting that big model gross margins are about 36%, lower than traditional Internet, but industry gross margins are on the rise. The commercialization of the AI industry needs to go through different stages, and the CV sector has been profitable. AI token economic logic is established, the underlying application optimization to promote the development of the industry. Set Perspective Technology, a young team focusing on large models and AI infrastructure facilities, has upgraded from a computer vision algorithm platform to an enterprise-level AI operating system provider, with hundreds of patents and more than 100 AI research and development personnel. It is committed to the deep integration of AI technology and enterprise business processes and the realization of large-scale application through intelligent body construction and operation and maintenance services. The company's business model covers large B- side and small B- side. It cooperates with large customers to verify the value of AI transformation. In the future, it plans to deepen the integration of business processes with customers and become a unified entrance for enterprise AI operation base and token consumption.
会议速览
Set Perspective Technology: The Road to Transformation from Algorithmic Platforms to Enterprise-Class Big Model SolutionsIntroduced the company's young management team, since its establishment in 2015, from the computer vision algorithm mall to the AI engineering platform, to the enterprise-level large model solution upgrade process, emphasizing the cooperation with industrial capital, to provide enterprises with AI landing full chain services.
Enterprise AI Operating System Construction and Large Model Application PracticeThis paper expounds the three-layer technology base for the company to build enterprise-level AI operating system infrastructure, including visual language big model, AI development platform and agent application platform, emphasizes the transformation from AI tool chain to big model era, as well as the in-depth cooperation with industrial customers, and discusses the challenges and opportunities faced by enterprise-level big model application.
Enterprise AI Landing: The Road to Transformation from Model Deployment to Business ValueIt is discussed that the key to the landing of enterprise-level AI is to solve complex problems such as historical systems, internal processes, and data silos, rather than simply pursuing model capabilities. Through the trust relationship established with customers, the value of visual algorithms and tool chains is transformed into large model solutions, operation and maintenance authorization and intelligent body usage revenue, emphasizing the importance of the last mile service that enterprises AI land.
AI Grid Platform: A Commercial Engine for Enterprise-Level Agent Ecology and Token ConsumptionThrough the analogy between power grid and power industry, this paper expounds the importance of AI power grid platform in enterprise application, emphasizes the ability of platform standardization, product and service, and the value of forming intelligent body ecology and continuous consumption of Token after in-depth business process, which indicates the change of core index of large model industry from efficiency tool to commercial engine.
The Historical Law of Enterprise Technology Landing and the Trend of Service IntegrationThis paper discusses the law of the successful integration of enterprise-level technology into the core process, points out that the final form of technology is often the combination of software and service, and emphasizes the importance of going deep into the customer's business process and the stable income model brought by it.
Big Model Enterprise Application Burst: Data Accumulation, Cost Drop and Engineering MaturityThe key to the outbreak of large-model enterprise applications lies in the original accumulation of data assets, the reduction of system deployment costs and the gradual maturity of engineering, which will be embedded in the enterprise operating system in the future, promote the upgrading of knowledge management, intelligent body interaction, etc., realize human-machine collaboration, and change the enterprise operation mode.
Building an enterprise-class AI base: from project delivery to long-term value creationThis paper discusses the whole process AI solution for large B- end customers, including consulting, knowledge base construction, intelligent body development and operation and maintenance, and emphasizes the realization of sustainable and expandable subscription income through deep coverage of enterprise processes, precipitation of intelligent bodies and active users, and the transformation from short-term project delivery to long-term value creation.
Building an Enterprise AI Operating System: Transforming from Project Services to Platform ServicesThis paper discusses the cost-effective token service of Skill Mall for small B- end users, as well as the ecological model of mall created by large B- end user project precipitation, developer intelligence, large factory product access and user UGC. This paper introduces how the company transforms project-based services into platform-based services, builds enterprise-level AI operating systems, realizes intelligent body construction, use and management, meets the security requirements of government and enterprise customers, and aims to create the front entrance of enterprise-level AI operating systems.
Big Model Enabling Enterprise Transformation: Practical Cases of Base Station Middle Station and Intelligent Body PlatformAs a background system for enterprise large model capacity production tuning and evaluation, the base station supports multi-modal data access and efficient processing, integrates reasoning engines and fine-tuning methods, preforms a variety of mainstream models, and helps enterprises tune low-threshold and high-efficiency models. Through the case of rail transit group, it shows how the big model reconstructs the enterprise-level business process, realizes the AI transformation, improves the fault repair efficiency, shortens the process time, and reflects the deep interaction and continuous use value of enterprise-level users.
Extreme Perspective Competitive Advantage: Full-stack technology and scale delivery of industry data under the triple moat.The extreme perspective builds a triple moat with industry data full stack technology, scale delivery capabilities, and customer stickiness. Its advantages include rich experience in government and enterprise projects, professional AI consulting team, one-stop closed-loop delivery capability, and re-purchase demand based on platform precipitation, forming a long-term value base.
Large model business growth strategy: horizontal expansion and vertical deepening of the two-wheel drive.The growth of the large model business relies on AI brand influence, regional cooperation, rapid demand assessment mechanism and rich project experience, reserve customers covering multiple fields, revenue elasticity comes from the increase in the number of internal intelligence of customers, forming a compound growth curve of horizontal expansion and vertical deepening.
Extreme Perspective Big Model Transformation: From CV Business to Enterprise AI Energy Management EntryThrough years of AI business accumulation, the extreme perspective has verified the AI's ability to commercialize, including sustained revenue growth, efficiency improvement and profitability model maturity. The company is transforming from CV business to large model technology provider, committed to becoming a unified portal for enterprises to use large models and tokens, building multi-modal perception capabilities and industry data assets, and promoting the deep application and energy management of AI in enterprises.
The relationship between business income model and large model economy is discussed.Business revenue models were discussed, including traditional offline charges, platform service fees, and token metering charges. Mention the economic links with large model suppliers and 24 years of profit sources, including large model and CV.
The trend of gross margin of large model and the profit prospect of AI industry are discussed.The dialogue revolves around the upward trend of large model gross margin and the earnings outlook of the AI industry. It is pointed out that the gross profit margin of the large model is gradually rising and may be close to the traditional Internet level in the future, but the specific value is difficult to predict. Emphasize that AI companies need to go through different stages of profitability, the CV field has achieved profitability, and the AI token economic logic is feasible, the key lies in the continuous optimization of the underlying application model and scenario expansion.
要点回答
Q：Dear investors, could you please introduce the management team of Ji Perspective Technology?
A：Of course, our management team is very young, the chairman and two co-founders are all post-90s. Among them, the chairman is a young man from Macau, who started a business in the Mainland and succeeded in listing in Hong Kong stocks. He is also a talent in the National Ten Thousand Talents Program. The team is composed of two co-founders with different backgrounds. One is responsible for the business direction, and the other is focused on the direction of artificial intelligence technology. It is a composite team that integrates technology product delivery and capital market.
Q：What is the history of the development of the set perspective technology shares?
A：The company has a history of 11 years since its inception, from 15 years to the core of computer vision, and in 17 years or so to create a computer vision algorithm mall, to solve the problem of algorithm supply and industry demand matching. By 19 to 22 years, we had built a tool chain and platform for algorithm development, management training, deployment reasoning, and AI engineering capabilities. 24 years later, the company launched the multi-modal large model intelligent body platform and the push-push integrated platform, from the algorithm supply platform to the AI engineering platform to the enterprise-class large model infrastructure and token aggregation platform upgrade.
Q：How is the technical base of the set perspective technology shares formed?
A：Our technology base is divided into three layers: the bottom layer is the foundation of the visual language large model, the middle layer is a platform that includes AI development training, reasoning evaluation and computing power scheduling, and the upper layer is the ARG model warehouse of the intelligent body-built enterprise, multi-user isolation knowledge base and other application platforms. This shows that the set perspective is not only developing a single application or model, but also building the infrastructure required for an enterprise-level AI operating system.
Q：What are the characteristics of the cooperation and industrial customers of the company?
A：The company has hundreds of patents, software and qualification certifications, and has more than 100 R & D teams with AI as the core. In recent years, R & D has grown rapidly. In the past, there were more than 3,000 government and enterprise customers, covering energy manufacturing, automobiles, construction, real estate and other industries. These customers will become the priority conversion targets of AI operating systems and token services for large-model intelligent enterprises in the future.
Q：How does the company convert customer resources into a source of revenue?
A：The company plans to convert its customer base into three types of revenue: the first is revenue from large model solutions, the second is ongoing operations and licensing revenue, and the third is token consumption revenue from the deepening use of intelligences.
Q：What is the current status of the AI industry and the market positioning of the set perspective technology shares?
A：At present, even powerful model companies have begun to attach importance to the landing services of enterprises. Many medium-sized enterprises have purchased AI packages such as ChatGPT enterprise, but a small number of companies are actually embedding AI into core business processes. This is because enterprises are faced with complex systems, processes, data islands and other issues, which cannot be solved only by providing APIs. Therefore, the set perspective technology shares are committed to doing enterprise AI landing token aggregation intelligent system and operation and maintenance, to make up for the gap of enterprise-level landing services, become a unified entrance for enterprise AI use, with the increase in the number of enterprise intelligence and business process intervention deepened, platform stickiness and value will continue to improve.
Q：What stages is the big model industry currently going through?
A：The big model industry is going through three phases. The first stage is the general application outbreak stage, AI can be used for simple generation tasks such as question answering, writing and retrieval. The second stage is the professional post efficiency stage, AI in-depth design, programming, financial analysis, office collaboration and other fields. The third stage is to AI deeply embedded in enterprise business processes and become part of organizational collaboration, which is also the biggest opportunity. In this phase, the consumption of tokens will undergo a qualitative change, from low-frequency dispersion to the basic consumption of enterprise business operations.
Q：What are the changes in the core indicators of the big model industry? What is the positioning of the polar perspective?
A：As large models enter the core processes of the enterprise, the core metrics will shift from model parameters to landing value, from efficiency tools to business engines. A truly valuable platform is a platform that can enter business processes, precipitate scene data, form an intelligent body ecosystem, and continuously consume tokens. The polar perspective positioning revolves around providing a platform, emphasizing full-process services and companionship, helping customers embed their intelligences deeper into their core business processes, resulting in more stable, higher platform fees and token consumption.
Q：What are the historical patterns for the landing of enterprise-level technology?
A：History shows that the technology that successfully enters the core processes of large enterprises is often not pure tools, but a combination of software plus services, and process reengineering. Companies such as SAP, Salesforce and Pantie have achieved deep integration with customers and long-term value creation in this way.
Q：What are the key changes in the current large model landing with explosive conditions?
A：Key changes include: the completion of the original accumulation of data assets, the reduction of system deployment costs, the reduction of illusion problems, the gradual maturity of basic engineering, the emergence of multi-modal capabilities and the formation of human-machine collaboration mechanisms, so that large models can be landed on a large scale and continue to operate.
Q：What is the role of the future big model in the enterprise?
A：The future big model will become the internal operating system of the enterprise, connecting all kinds of business systems, knowledge base, approval management authority system and front-end intelligence, through each business flow, system call and human-machine collaboration to bring token consumption, and with the context accumulation consumption continues to accumulate.
Q：What are the main business models and service contents for big B- end customers from an extreme perspective?
A：The polar perspective mainly provides the whole process of large model consulting products and operation and maintenance services for large B- end customers, including four-step consulting services such as status quo diagnosis, scenario design and feasibility assessment. In addition, it is responsible for knowledge base and data construction, intelligent body development, and continuous operation and maintenance to generate tens of millions to hundreds of millions of dollars in solution revenue, as well as tens of millions to tens of millions of dollars in operation and maintenance and licensing fees, and build a long-term and stable AI budget entry and subscription revenue model.
Q：Once an enterprise embeds AI into the whole process, why does the replacement cost gradually increase?
A：When the AI platform within the enterprise becomes dominant, it co-ordinates model access, smart body calls, permission control, billing, and cost management, and as the depth of use increases, so does the replacement cost.
Q：What is the reason for the large consumption?
A：When enterprises deeply embed AI into business processes, the average daily token consumption can reach hundreds of millions of levels, and the annual consumption may even reach hundreds of billions of levels. Therefore, customers' professional needs for token procurement scheduling, cost optimization, and performance management are increasing.
Q：How will the company measure key indicators of project success in the future?
A：In the future, the company will no longer only focus on the number of contracted projects, but more on how many enterprise processes the platform covers, how many smart bodies it deposits, how many active users it generates and how much it consumes, which is the key bridge for the company to move from making money to long-term value.
Q：What is the company's plan for small B- end users?
A：The company plans to provide a cost-effective skill mall that includes semi-standardized capabilities and abstracts customized content into tools, such as creative generation and AI programming assistants, to serve small B- end users.
Q：What are the ways the company is commercialized?
A：Commercialization methods include payment by token consumption, monthly or annual resource packages, large model service fees, customized service fees, etc. The company is committed to transforming project-based services into platform-based services, providing a deep benchmark for the big B end, providing scale and ecology for SMB, connecting both ends to form the company's ultimate products.
Q：What is the strategic positioning of the "extreme" platform?
A：The "Extreme" platform is the core platform for enterprise intelligence applications, not only a chat robot platform, but also an enterprise-level platform for building, managing and operating intelligence within the enterprise, supporting local deployment in the cloud, with enterprise-level security features such as full-link operation audit and architecture isolation.
Q：What is the role of the "base" integrated middle stage?
A：The "base" is an integrated middle stage, which serves as a background system for enterprise large model capacity production tuning and evaluation, supports multi-modal data access, completes data pre-processing, enhancement and verification, and provides low-threshold and high-efficiency model tuning capabilities through integrated reasoning engines and advanced fine-tuning methods.
Q：What kind of value does a rail transit group case show?
A：This case reconstructs the group's enterprise-level business process by combing through 535 intelligent body requirements, providing a full set of AI transformation capabilities including consulting, platform, knowledge base, model training and operation and maintenance. The project amount is large, the process is deeply embedded, the subsequent operation and maintenance and investment consumption will continue, and the intelligent body application significantly improves the efficiency, such as the power supply system fault repair process consumption reduced by about 80%.
Q：Does token consumption become a tight metric after AI take over the business process, and why customers are willing to continue to use Talking?
A：Token consumption is not a simple compact indicator now, it represents the actual usage of the business process after it has been taken over by the AI. Customers are willing to use Talking continuously because they can feel the shortening of the process and the improvement of efficiency.
Q：What are the main aspects of the competitive advantage of the extreme perspective?
A：The competitive advantage of the polar perspective can be summarized as a triple moat: industry data full stack technology and scale delivery. Specifically, it includes three aspects: first, with rich management consulting experience and a professional team of AI consulting engineers, it can transform complex requirements into a landing solution; Second, it has one-stop closed-loop capability from scheme delivery, development, deployment, operation and maintenance to compliance, and is especially good at first-line customized delivery. Third, it has the advantage of high customer stickiness. Through the precipitated customer platform and in-depth business application, the customer is prompted to generate new demand and re-purchase, so that customers need to replace the knowledge base, process permission model configuration, etc. when changing suppliers, forming a unique long-term value base.
Q：What is the future trend forecast for large model business customers?
A：In the future, the development trend of large model business customers is to serve large B- end customers in the medium term, as a replication case. At the same time, the growth of token consumption will be used to enlarge revenue and support the company's growth.
Q：What are the factors supporting the company's growth?
A：The factors supporting the company's growth mainly include: high-quality clues brought by brand influence in AI fields; Regional partners and key regional resources help to cover major customers; The formed mechanism for quickly sorting out and evaluating the needs of large models of enterprises and providing consulting solutions; As well as rich bidding experience, demo verification and project promotion can be completed quickly.
Q：What is the current reserve of customer resources?
A：At present, the company has a reserve of large customers such as rail transit, urban investment operation group and port, and the growth of the number of customers is the first layer, while the income elasticity comes from the superimposed effect of the number of internal intelligences of each customer, the increase in demand and the expansion of process coverage, that is, horizontal expansion and vertical deepening together form the company's growth curve.
Q：What proven business foundations are built on the transformation of the extreme perspective model?
A：The transformation of the extreme perspective model is based on three verified businesses: AI products can be continuously sold to enterprises and their income continues to grow, and the increase in the number of customers, delivery volume and re-purchase rate proves the commercialization ability of enterprise-level AI; The AI can continuously optimize itself to improve efficiency, gross profit margin continues to rise, and sales and distribution expenses decrease. AI real operating results have been formed, and the company has achieved adjusted profits in 24 years, the proven business model is validated, and the next phase will be to become a unified operating system provider and portal for enterprises to use large models and tokens.
Q：The role of CV business for the company and its relationship to the transformation of the big model?
A：As an important basic market of the company, CV business provides stable cash flow and customer base, maintains stable growth and maintains a good gross profit level, reflecting the ability of platform delivery and product reuse. More importantly, the industry customer scenario data, image understanding capabilities and project delivery experience accumulated by the CV business over the past decade have become an important technical capability base for large model application scenario migration, helping the company to succeed in large model transformation.
Q：In the future, can the unity of standardized applications be achieved through the collaboration of visual large models and natural language cues?
A：Yes, in the future, more standardized applications can be achieved through the synergy of visual large model size models and cues for natural language. CV will transform from an algorithmic mall to an enterprise large model base that provides multi-modal perception capabilities, industry data assets and public delivery capabilities.
Q：What is the position of the visual language big model in the company's big model technology system?
A：The visual language big model is an important part of the company's big model technology system. Its positioning is to provide accurate and stable multi-modal perception and understanding capabilities for enterprise management agents. Instead of competing with other general model big factories for parameters, it uses 1 billion-level real business data sets, more than 50 million industry data sets and a large amount of context instruction data accumulated in the past ten years to generate different versions through fine visual perception and understanding capabilities, to meet the needs of end-side general-purpose landing and cloud-based complex applications.
Q：What is the overall revenue model of the company, and in particular how the token is charged?
A：At present, the company's business revenue mainly adopts the offline-based way, and after signing up with customers, it charges according to the single platform service, such as the overall cost of AI and the cost of different intelligences. At the same time, the company gradually began to measure the token and develop billing plans.
Q：Does the company have an economic link to the company that provides the big model?
A：The company will choose different large models to adapt to specific application scenarios according to customer needs, and sign contracts with relevant large companies.
Q：Is the adjusted profitability improved by the original business or the large model business?
A：Earnings improvement in 2024 came from both the large model and CV components.
Q：Will the gross profit margin of the big model be closer to that of the traditional Internet industry? What level of gross profit margin do you think the big model business can eventually achieve?
A：The gross margin trend of the big model is somewhat similar to that of the Internet industry, but it is not yet possible to fully determine whether it will reach the level of the traditional Internet industry. However, the company's internal large model business gross margin did show an upward trend in the past two years. The gross profit margin of large model business is on the rise, but it is impossible to give a clear figure for the time being, because the major manufacturers have invested heavily in infrastructure, and it is uncertain whether they can see the critical point of business transformation.
Q：Under what circumstances can we see that the transformation of AI companies is clearer?
A：For the transformation of AI companies, there are different levels at different stages, and some companies are already profitable in specific areas such as CV. In the AI token economy, with the development of the underlying application, the use and transformation of the model is a tenable logic, and different nodes will be formed in the future.
