The most powerful industrial brain composed of "large models + intelligent agents" has truly arrived this time.
Since the release of GPT-o1, the technological capabilities of AI have witnessed an unprecedented leap. At the beginning of this year, the release of DeepSeek R1 significantly reduced the implementation cost of this useful deep - thinking model, truly achieving mass popularization. AI is now comprehensively and deeply penetrating into people's lives. Recently, the State Council issued the "Opinions on Deeply Implementing the 'Artificial Intelligence +' Initiative", further accelerating the process of AI's penetration into all industries.
However, looking globally, many technology practitioners are still troubled by the fact that AI cannot access a wider range of application scenarios. An AI practitioner once told 36Kr that currently, the most thoroughly transformed scenarios are still those with a high degree of onlineization, such as programming and office work. "It is estimated that AI will greatly improve the efficiency of these scenarios in the future."
From the Internet to the mobile Internet, and now to the AI era, the industrial field has always been one of the toughest nuts to crack for AI technology implementation. Not to mention the strict requirements of industrial scenarios for safety and accuracy, the deep integration with the complex real - world physical environment has also given rise to significant data demands. Under these difficulties, who will take on the challenge of cracking this tough nut has always been a core issue that concerns the industrial community.
In recent years, with their in - depth accumulation in the industrial field, many leading enterprises have also shown great interest, jointly heating up the industrial AI market. It is reported that the scale of China's industrial AI market is expected to exceed 120 billion yuan in 2025, with a compound annual growth rate of up to 28.6%.
But to what extent is the current industrial AI usable? Will an industrial AI Agent with autonomous capabilities be born? These questions have found answers after a recent major release.
Exploring a 3.5 - trillion - yuan Market: The Formal Debut of Industrial AI Agents
In fact, over the past year or so, the consensus in the industrial community has quietly shifted. The debate about "whether to integrate AI" has disappeared, and instead, there has been an increase in AI investment and a full embrace of large - scale models.
According to an incomplete statistics by 36Kr, in the industrial field, since 2023, iFlytek has released the Lingyang Industrial Large - scale Model, Simo has launched the industryGPT Industrial Multi - modal Large - scale Model, China Unicom has created the Yuanjing Industrial Large - scale Model, China Mobile has released the Jiutian Industrial Large - scale Model, NetEase has developed the Fuxi Industrial Large - scale Model, and SUPCON Technology has released the Time - series Pre - trained Transformer (TPT) large - scale model.
On one hand, major technology companies are accelerating the R & D of industrial large - scale models. On the other hand, industrial enterprises are quickly adopting them. Even, with the support of industrial AI, the first batch of results have been seen. Taking the TPT large - scale model released by SUPCON Technology as an example, in the waste liquid treatment process of a global top 50 chemical company, TPT shortened the pH adjustment cycle from the original 6 - 8 hours to less than 1 hour, with an efficiency improvement of over 80%. In the Datang Duolun Coal Chemical Project, TPT constructed an integrated energy management and control system of "perception - prediction - regulation", which is expected to replace coal - fired power generation with green electricity by 496.902 million kilowatt - hours annually, reducing carbon dioxide emissions by 419,400 tons.
Behind these results lies a more significant trend: Industrial AI Agents are becoming the best form for AI implementation in the industrial field. The latest industry report shows that the global industrial intelligence market scale will exceed 3.5 trillion yuan this year, with the Chinese market accounting for over 40%. Industrial AI Agents are no longer simple software that executes instructions but are integrated hardware and software platforms with autonomous perception, cognition, decision - making, and learning capabilities. More importantly, they are trained based on the core database of the industrial industry, with further improved execution efficiency and certain dynamic adaptive capabilities, which will completely change the logic of human - machine interaction and production management.
However, since industrial AI Agents are still in the early stage of development, the current core focus of the industry is still on who has a more valuable and accurate core database and can train large - scale model capabilities based on these core "data fuels" to achieve better results in the implementation of industrial AI Agents.
Currently, there are typical representative enterprises in the industrial AI track, such as SUPCON Technology mentioned above. Its developed time - series large - scale model TPT has achieved a paradigm shift from "mechanical execution" to "autonomous decision - making". Focusing on the industrial AI field, SUPCON Technology has a strong brand identity. It has entered various core industrial scenarios and has accumulated more than 30 years of basic skills in the industrial control field. It is reported that SUPCON Technology has served more than 37,000 customers globally, covering more than 50 sub - industries in the process industry, and has deployed more than 100,000 sets of industrial control systems and approximately 100 million I/O points. These elements constitute the core fuel for training industrial AI - the industrial database.
In the past, based on a deep understanding of industrial scenarios, SUPCON Technology made many explorations in mining the operation rules and potential features of equipment hidden in time - series data, and finally developed a truly implementable time - series large - scale model.
On August 28, 2025, SUPCON Technology's time - series large - scale model was officially upgraded to TPT 2 (Time - series Pre - trained Transformer 2) and released globally. This release is not only a technological advancement but also a paradigm leap.
The core upgrade points include the adoption of the MoE architecture (Mixture of Experts model), multi - technology integration, and the realization of Agent construction. It is like an all - around expert in the industrial field, capable of deeply understanding and solving many complex problems in industrial scenarios, especially in helping users ensure production safety, improve product quality, reduce energy and material consumption, enhance equipment maintenance efficiency, and achieve autonomous operation.
It can be said that the release of TPT 2 is not only a huge upgrade of SUPCON Technology's own products but also a pioneer in industrial scenarios. By creating a "smart brain" for equipment operation, it has truly promoted the transformation of industrial applications from traditional tool - type assistants to Agents with the capabilities of active anomaly recognition, intelligent risk assessment, and autonomous decision - making and execution. It reshapes the industrial production paradigm with scenario - based intelligent solutions and realizes another round of innovation in the industrial software technology system and application mode.
Cui Shan, the Chairman and President of SUPCON Technology, also told 36Kr in an interview after this release, "We hope it is a more open technology, enabling both small, medium - sized and large enterprises in the industrial industry to use AI. At the same time, SUPCON Technology will enter the platform - based era and truly empower ecological partners."
An Industrial AI Agent with Autonomous Capabilities Has Arrived
In the past, industrial R & D often relied on engineers' experience and used relatively inefficient trial - and - error methods. Industrial AI Agents are like extremely experienced masters, greatly improving the efficiency of R & D. In the production and manufacturing process, they can autonomously optimize and improve the efficiency of production process control and operation, achieve more refined and automated management of the production process, and efficiently maintain equipment.
More importantly, industrial AI Agents are not simply a superposition of technologies but a complex and comprehensive proposition.
SUPCON Technology's TPT 2 takes the large - scale model as its core brain and combines more industrial knowledge, turning this large - scale model into a powerful brain in the industrial field. Finally, it has taken the lead in realizing the application of Agents in industrial scenarios.
The release site of TPT 2
To understand the value of TPT 2 more specifically, we need to break it down into three dimensions:
Firstly, it is intensive. One MoE model can handle a variety of problems and generate different Agents, directly replacing N industrial software in the processes of simulation, control, optimization, prediction, evaluation, and statistics.
Secondly, it provides dialogue - based diagnosis and helps with decision - making. The dialogue - based solution is like having an industrial expert face - to - face. Even if the operator lacks certain professional knowledge, they can handle things smoothly and make accurate decisions quickly.
Thirdly, it is a real industrial AI Agent, realizing a full - link closed - loop. From perception, recognition, decision - making to execution, the Agent launched by SUPCON Technology this time is a highly complete one, truly helping enterprises achieve less - manned or unmanned management.
The release of TPT 2 will help AI further penetrate into industrial scenarios, and more and more factories with a sense of the future will emerge. Imagine walking into a modern factory and seeing batches of robots busily working, unmanned vehicles shuttling between production lines, and an AI with autonomous decision - making awareness scheduling thousands of devices in real - time...
Implementation in Hundreds of Devices: Efficiency Improvement Subverts Tradition
According to the introduction at the release site, TPT has achieved breakthrough applications in more than a hundred sets of devices of leading industry customers such as Wanhua Chemical, Sinopec Zhenhai Refining & Chemical, and Datang Duolun Coal Chemical.
Taking the optimization of the ethylene plant of PetroChina Lanzhou Petrochemical Yulin Chemical Co., Ltd. as an example, the design quality of equipment often determines whether a project can be put into production normally and whether it can operate in the long - term after production. These equipment have also had a certain failure rate over the years, leading to unscheduled shutdowns at different frequencies.
PetroChina Lanzhou Petrochemical Yulin Chemical Co., Ltd.
"In the past, the optimization of ethylene plants was advanced on a 'yearly' basis. Now, with our TPT 2, according to our previous cases, the optimization cycle has been shortened to about one month," Wu Yucheng, the Vice - President of SUPCON Technology and the CEO of the Industrial AI Business Group, told 36Kr. This means that the new technological capabilities of industrial AI Agents bring subversive efficiency improvements to industrial scenarios and help solve complex problems.
China National Coal Group also entrusted SUPCON Technology to optimize its complex core coal gasification equipment. It should be noted that the coal gasification process (converting coal into effective gases such as carbon monoxide and hydrogen) is extremely complex and often "invisible and intangible". Traditional mechanism - based modeling methods are difficult to apply accurately due to excessive ideal assumptions. SUPCON Technology's TPT large - scale model combines AI and mechanism and conducts pre - training based on a large amount of historical time - series data, which can better handle this problem.
As of the end of the release, 36Kr learned that TPT had a cumulative of 518 pre - ordered enterprises, 112 sets in sales volume, and 3,268 online registered applicants. This series of data fully reflects the high recognition of TPT 2 by the market and customers, marking that this model has officially entered the stage of large - scale application and implementation.
The Era of Large - scale Application of "Large - scale Model + AI Agent" in the Industrial Field Has Arrived
In the previous era, the improvement of quality and efficiency in industrial scenarios was driven by digital transformation, but the scenarios it truly reached were very limited. Industrial AI Agents have brought the efficiency improvement in industrial scenarios into a truly intelligent - driven era.
Before this, a core problem needed to be solved. In the era of industrial AI, the data accumulation ability of a single enterprise is no longer sufficient to support the development of industrial AI. Coupled with the complexity of industrial scenarios, high industry barriers, data sensitivity, privacy issues, etc., industrial AI Agents have never been applied on a large scale. Now, it seems that TPT 2 is a turning - point product in the industry, marking that industrial AI Agents have entered the stage of large - scale application. Behind this is the vision and courage of SUPCON Technology.
At the release site, SUPCON Technology also joined hands with partners in the industrial ecosystem to establish the "Industrial AI Data Alliance". This is not only a technological solution but also represents a new type of industrial cooperation model. It is reported that this alliance brings together leading enterprises, design institutes, general contractors, and service providers in the industry, with the core of jointly building an industrial data value ecological system based on trust.
The inauguration ceremony of the "Industrial AI Data Alliance"
Of course, data security is still the top priority. SUPCON Technology said that it will adopt privacy - computing technologies such as Federated Learning and multi - party secure computing to realize the value circulation of device data across scenarios and working conditions while ensuring data ownership and privacy security. On the absolute bottom line of security, stronger industrial AI will be refined, ultimately benefiting the entire industrial industry.
The IDC 2025 China Industrial Enterprise Survey shows that the proportion of industrial enterprises that have applied large - scale models and AI Agents has significantly increased from 9.6% in 2024 to 47.5% in 2025. Among the enterprises that have applied large - scale models and AI Agents, more than 73.7% have between 10 and dozens of application scenarios.
The industry is actively embracing the era of industrial AI, which also lays a good cognitive foundation for the release of SUPCON Technology's new AI Agent. The accelerated commercialization of AI Agents also enables leading enterprises that took the lead in the layout to enjoy a new round of dividends.
According to the views of industry experts such as Liu Jinsong, an engineer from the National Industrial Information Security Development Research Center, quoted by the China Industrial Economic Information Network, industrial AI Agents are the product of the deep integration of AI Agent technology and industry know - how. They can realize the intelligent control and optimization of production equipment, process flow, and logistics management, and are the key driving force for the digital transformation of the manufacturing industry to move from "information construction" to "value creation".
Conclusion: An Underestimated Leader in Industrial Intelligence
Currently, whether it is the technological leadership, the subversiveness of the business model, or the huge promoting effect it may bring to the industrial track, SUPCON Technology has once again stood at the new forefront of the industrial industry.
The financial report shows that in 2024, SUPCON Technology achieved an operating revenue of 9.139 billion yuan, a year - on - year increase of 6.02%; the net profit attributable to the parent company was 1.117 billion yuan, a year - on - year increase of 1.38%. However, the slowdown in net profit growth is mainly due to the investment in strategic businesses, especially the investment in the R & D of core AI technologies such as TPT. But similarly, the potential is emerging. In 2024, SUPCON Technology's industrial software revenue increased by 20.70% year - on - year to 2.653 billion yuan, with a growth rate significantly higher than that of traditional control businesses.
Its previous technological investment is now entering the harvest period. Especially after the release of TPT 2, its platform - type company attributes are more prominent, laying a solid