How can Ruixin (02687) make up for the huge gap in AI investment per capita?

The Zhitong Finance App learned that on average how many students does a graduate mentor in a Chinese university take? The figures given in the Ministry of Education's “2025 National Annual Report on the Quality of Postgraduate Education” are that the ratio of undergraduate students in ordinary colleges and universities is about 1 to 24, 1 to 18.5 for graduate students, and more than 1 to 25 for some engineering majors. In the same period, the ratio of teachers to students at graduate level at American research universities was about 1 to 6.8, MIT was about 3 to 1, Princeton University was about 5 to 1, Stanford University was about 6 to 1, and Harvard University was about 7 to 1.

In terms of conversion, the number of students per Chinese graduate mentor is about 3 to 4 times that of similar institutions in the US.

The gap at the undergraduate level is also stark. The average ratio of teachers to students in four-year colleges in the US is about 1 to 17, which is about 30% lower than in China. However, the annual enrollment scale for graduate students in China has exceeded 1.3 million. The supply of teachers is far from keeping up, and the student-teacher ratio continues to rise. Chinese university teachers are training a larger student population at several times the load of their American peers.

This is an arithmetic problem with almost no solution in the short term. The large-scale recruitment of teachers is constrained by the teacher training cycle and preparation, and the drastic increase in average student funding is constrained by financial affordability. The physical world is slow to catch up, yet another set of numbers reveals a more pressing gap.

In 2025, the total investment in informatization and AI in Chinese universities was about 28 billion yuan, an increase of about 33% over the previous year. This figure looks pretty small, but when it comes to 47.63 million students, the per capita investment in AI is about 588 yuan, which is less than 81 US dollars. Moreover, this 28 billion yuan is mainly concentrated in the hardware procurement process, and the proportion of software, services, and continuous operation is still very low; it can only be considered the starting line for the entire college AI market.

American colleges and universities do not have a unified official statistical standard, but according to the comprehensive per-seat subscription model and college IT budget structure estimates, the per capita AI investment is in the range of about 200 to 500 US dollars (EDUCAUSE 2025 IT Budget Benchmark), which is about 2.5 to 6 times that of China.

Other than the monetary gap, the structural issues are even more glaring. A significant proportion of AI spending in Chinese universities goes to hardware infrastructure such as servers, GPU clusters, and network transformation. The US side invests more in software, services, and ecological construction, and the marginal benefits of the latter type of investment are usually higher.

The white paper of the China Education Equipment Industry Association revealed an intriguing figure. The average construction cost of an AI laboratory in Chinese universities reached 12.8 million yuan, yet the equipment idle rate was close to 47.8%. Nearly half of the equipment is not converted into actual teaching output. The money has been spent, and the classroom has not changed.

The difference between teachers and students is 3 to 4 times, and the difference in AI investment per capita is 2.5 to 6 times. When the two sets of numbers are superimposed, this is the reality that higher education in China must face. Resources are limited, and the two burdens of scale expansion and quality catch-up are on the shoulders at the same time.

The traditional path doesn't work; catching up can only find a breakthrough from the perspective of new technology.

AI's changes in higher education go far beyond adding a tool to the existing teaching process. What it changes is education productivity itself.

On the teaching side, the AI-assisted system can generate personalized exercises and feedback according to each student's learning progress. It is technically feasible for one teacher to provide differentiated guidance to 200 students at the same time. The same workload requires 4 to 5 teaching assistants to work together under the traditional model. In postgraduate training, AI research assistants undertake repetitive tasks such as document screening, data analysis, and initial screening of experimental plans, freeing mentors from inefficient labor and concentrating on providing high-value academic guidance. The field of experimental training is also changing. Physical AI technology brings high-risk, high-cost, and high-consumable experimental scenarios into a digital twin environment. Students can operate repeatedly without being limited by scarce equipment.

How far are these scenes from college? A timeline has been given for policy-level deployment.

Since 2025, documents such as “Opinions of the State Council on Further Implementing the “Artificial Intelligence +” Action Plan, the “Artificial Intelligence + Education” Action Plan by five departments including the Ministry of Education, and the “Outline Plan for Building a Strong Education Country (2024 to 2035)” have been intensively issued.” The “15th Five-Year Plan Outline” incorporates “Further Implementation of the Education Digitalization Strategy” into the Higher Education Section. The 2026 National Education Work Conference further proposed that the application of artificial intelligence in the field of education should be promoted in practice, and that artificial intelligence general education should also be promoted at an accelerated pace throughout the school sector.

From the overall deployment of the State Council to the rigid indicators of the ministries and departments, the policy signals are highly consistent. As AI enters higher education, it has changed from an incentive item to a must-answer question.

The demand-side data also illustrates the problem. In 2025, the AI solution market in Chinese colleges and universities grew by more than 30% year on year, maintaining rapid expansion for three consecutive years, becoming the fastest growing segment of education informatization investment. Driven by hard policy constraints and actual resource gaps, this market is still accelerating.

Looking deeper, the current AI investment in colleges and universities is mainly hardware procurement, and software and services account for a low proportion. The 47.8% equipment idle rate shows that what colleges really lack is AI ability to sink into the teaching process; computer rooms and servers are only prerequisites. Once AI evolves from an auxiliary tool to teaching infrastructure and is deeply embedded in the entire process of curriculum design, academic evaluation, experimental training, and scientific research assistance, the market space will far exceed the scope that can be framed by current informatization budgets.

Simply calculate an account. China has 47.63 million students, and the per capita investment in AI is currently less than 81 US dollars. Even if it only catches up to the low-end level of the US at $200 in the next few years, the corresponding market size for college AI alone will be close to 70 billion yuan. If it reached the US median of $350, that figure would exceed 120 billion yuan. This is still just a section of higher education. It doesn't count software and services as additional hardware alternatives, and it doesn't include the K12, vocational education, and lifelong learning sectors.

Open up your perspective a little more. The digitization of education is a systematic project covering all school segments and scenarios. Higher education is the sector with the strongest payment capacity, the most rigid demand, and the strongest policy support. It is also the most mature testing ground for AI technology. Moving from colleges and universities, extending to vocational education and basic education, and moving from single-point applications to full-process teaching infrastructure, the long-term ceiling of this racetrack is far above 100 billion dollars.

It is at this node that Zhuoyue Ruixin (02687) enters the field of vision.

The direction this company is aiming for is very clear. There is a structural shortage of higher education resources in China, and to what extent AI can fill this gap. The company is based on self-developed full-modal large-scale models to drive AI capabilities to advance in depth from single-point teaching assistance to full-process teaching infrastructure, covering the complete chain of curriculum design, personalized learning, academic evaluation, scientific research assistance, and reality training. This set of abilities has been refined and matured from the college scene, and is naturally expandable to be replicated in neighboring markets such as vocational education and corporate training.

The full-modal large model is the core technical asset of Excellent and Innovative Technology. According to public information, the company has connected to the World Labs world model to form four major technical matrices covering text, speech, vision, and 3D generation, equipped with visual frameworks such as VOM and VLM. From the bottom model to the upper level teaching scenario, the entire chain of capabilities is in your own hands.

Technology needs to be implemented in the classroom, depending on the service network. Excellent Ruixin has built a service system covering the whole country, connecting the path between technology research and development and classroom implementation. This is particularly important in the Chinese university market. Colleges and universities in different regions and at different levels have huge differences in infrastructure, teacher level, and degree of digitalization, and it is impossible for a single plan to cover the world. The nationwide service network enables the company to localize deployment and continuous operation according to the actual situation of different colleges and universities, and transform cutting-edge capabilities into teaching productivity that can be replicated on a large scale.

The actual results can already be seen. With Zhuoyue's new AI system, it is possible for one professor to simultaneously provide differentiated guidance to hundreds of students. Expensive experiments that were originally limited by equipment and space have been transformed into digital twin experiences that can be iterated over and over again. AI is embedded in everyday teaching processes; it is no longer a device that falls dust in a laboratory.

In May 2026, Zhuoyue Ruixin was selected as one of the first constituent stocks of the “Hong Kong Commercial Daily · Hong Kong Stock Technology Index”. The capital market's recognition of its technological positioning and growth logic can be seen from this.

Back to the original math problem. The gap in the student-teacher ratio in Chinese universities will not be bridged overnight, nor will financial investment expand indefinitely. But digital efficiency provides a viable path to bridging the quality gap with limited resources. One professor plus an AI system can handle the workload of the past five teaching assistants. A digital twin lab allows students to repeatedly operate expensive devices they would have dared to see and touch.

28 billion dollars is only a subset of the current information technology budget. The 33% annual growth rate is still climbing, 2.5 to 6 times the catch up space per capita has not been realized, and the pain point of nearly half of the idle equipment is waiting to be solved. Taken together, these numbers outline a market that has just started from the starting line and has a long-term ceiling of over 100 billion dollars. Policies are being promoted, demand is rising, and technology is maturing. College AI has reached an inflection point from buying hardware to using software and services.

What Zhuoyue Rui has done in a new way is to stand on this lever. The technical breadth of the full-modal model is uneven in the depth of demand for higher education, and the flexibility of platform-based deployment is uneven in the distribution of resources in Chinese colleges and universities. Resource gaps exist objectively; only those that can leverage it are efficient.