Tech companies heavily invest in AI factories and enterprise intelligent agents

Artificial intelligence (AI) is fully moving from the training phase into the inference and execution phase, and technology companies in both China and abroad are accelerating their deployment of AI factories and enterprise agents, jointly driving AI development from both the infrastructure and application sides, as well as the cost and productivity sides.

Recently, Nvidia announced a collaboration with industrial software company AVEVA Jianzhi Software, integrating AVEVA’s engineering design and operational optimization software into an AI factory called Omniverse DSX Blueprint.

According to the announcement, the industry views this collaboration as a strong alliance, which will help maximize GPU productivity and energy efficiency through optimized computing architecture. Nvidia will rely on AVEVA Jianzhi Software’s full lineup of products, including the CONNECT industrial intelligence platform and industrial digital twin capabilities. By leveraging simulation for specific domains, digital visualization, and collaborative design tools, it will maximize GPU utilization efficiency, speed up AI factory deployment, and is expected to shorten the AI factory’s “time-to-token” cycle (the period from construction to outputting computing results).

AVEVA Chief Product Officer McGready said that AI factories are rapidly becoming the industrial-grade engines of the global digital economy. To promote this transformation, the two sides will jointly build a new digital twin deployment approach, constructing a new type of digital twin at scale. Through design, construction, and AI optimization, they will build a future-oriented intelligent industrial system. Troy, vice president of AI infrastructure at Nvidia, said: “With the rapid rise of AI factories at the gigawatt scale, the industry needs a new kind of industrial intelligence to optimize the entire lifecycle of large-scale data centers—from initial design to real-time operations. Through this collaboration, we are providing developers with a unified digital twin architecture, thereby accelerating AI infrastructure deployment and improving its efficiency.”

AI factories can significantly reduce the cost of running AI. And large-scale AI factories and data center operations place very high demands on energy supply. In addition to Nvidia, tech giants are also making big moves to invest in the energy sector to support the operation of large data centers. Amazon has reached a long-term agreement with Talen Energy, which will supply 1,920 megawatts of electricity to Amazon Web Services’ data centers from the Susquehanna nuclear power plant under Talen Energy. The two sides are also considering collaborating to build small modular reactors and expand capacity for the nuclear power plant. Google has also announced the restart of a nuclear power plant located in Iowa, USA, to power its AI infrastructure. In Europe, according to a data center expansion strategy plan published by the German government this month, by 2030, the computing power of Germany’s general-purpose data centers will be at least doubled compared with 2025.

At the industrial application level, Chinese technology companies are also accelerating their deployment, and are expected to achieve a paradigm breakthrough in enterprise-agent application and empowerment on a global scale.

Li Kaifu, CEO of Zero One Wanwu and chairman of Innovation Works, said that artificial intelligence has moved beyond a mere stage of technical exploration and has officially entered a new phase of deep industrial application and full-scale development. He said that bringing these world-class AI solutions to China is precisely to accompany Chinese enterprises as they move toward the world with a stance of technological leadership and independent, controllable capability, so that domestic traditional industries undergoing digital and intelligent transformation and upgrading can truly embrace this wave of “intelligent economic dividends.”

Li Kaifu said that 2026 will be the “year of enterprise multi-agent onboarding.” AI is evolving from the auxiliary role of “one person, one tool” into the organizational core of “one person, one team.”

Li Kaifu believes that real industrial transformation is not about purchasing a standardized set of AI software, but a “organizational transformation” that must be driven from the top down. “That is why Zero One Wanwu has always insisted that an enterprise’s AI digital and intelligent transformation must be a ‘top executive project.’ Managers need to think about and practice ‘how to use AI to reshape business.’”

He also pointed out that China has the world’s most complete range of industrial categories and an enormous market at scale. Each production workshop and every step in the supply chain process is an excellent laboratory for the evolution of multi-agents. With China’s deep industrial foundations coupled deeply with multi-agent technologies, China will evolve steadily from the “world’s factory” into a “global multi-agent factory” driven by intelligence.

Keenly focused on its AI business, Kingdee International recently said that only by processing massive amounts of effective data that fit specific business scenarios can the value of applications such as enterprise-level AI agents be realized.

Executives at Kingdee International said that for over 30 years, the company has served more than 7.4 million enterprises, and it has a very deep understanding of business logic in complex scenarios such as finance, supply chains, manufacturing, and R&D. “For developing enterprise-level AI applications, what matters is deep understanding of real scenarios. The agents that Kingdee has developed are closer to customers’ actual needs.”

However, for the software services industry in which Kingdee operates, AI applications may replace older business models, exposing related service companies to the risk of performance declines. In the industry, it is widely believed that accelerating adoption of AI agent services is the direction to break the deadlock.

Xu Shaochun, chairman of the board and CEO of Kingdee International, said: “The AI era is accelerating. On the one hand, we have a sense of crisis; on the other hand, we also have a strong sense of excitement.” He revealed that since the end of last year, the AI software development ecosystem and software engineering methodologies have become increasingly mature, and the development frameworks and infrastructure related to large models are also being gradually improved. All of this provides technical support for developing agents. And based on years of accumulated understanding of business logic and processes of software application customers, Kingdee can develop agent products more quickly.