Published on Dec 31, 2025

The Aluminium Corporation of China (Chinalco) has again shared details of its artificial intelligence deployment with industry peers and the public. The release of version 2.0 of the Kun'an Large Model on December 26, 2025 has been accompanied by the publication of 34 additional concrete use cases, reported the China Metallurgical News (in Chinese).

The updated list of 52 vertical models reveals a comprehensive attempt to embed AI into the entire value chain from geological exploration and mineral extraction to smelting, processing and administrative functions like procurement.

Since the state-run Chinalco Group started to deploy its specialised large model built on a cloud stack provided by Huawei, many of its subsidiaries have followed suit. For example, in 2025, electrolysis workers at Yunnan Aluminium Wenshan began to use prediction models for their process parameters.

They are now able to "combine data from cell control, measurement, and laboratory analysis to predict and output key control parameters for real-time regulation of electrolytic cell production", China Nonferrous Metals News writes (in Chinese).

Making AI work in this way can make a real difference. The intricate chemical reactions during the smelting of aluminium, copper, lead, zinc, gallium, and germanium demand precise control over process parameters. Even slight errors can reduce quality or unnecessarily increase the use of energy.

Now such processes are often no longer just driven by the experience of workers and managers, but are increasingly becoming data-driven. At the same time, the nonferrous metals industry association that Chinalco is cooperating with hopes to spread this know-how throughout the industry.

Ge Honglin, President of the China Nonferrous Metals Industry Association (CNMIA), spoke of a ‘development pattern’ taking root and cited early examples of productivity gains:

It remains to be seen, of course, how fast large models can be deployed throughout China's nonferrous metals industry. Not all players have the same resources and talent available as a big state-run conglomerate like Chinalco. The association, supported by Chinese government agencies, is showing a clear intent to promote widespread deployment, however.

One goal of the new model release at the end of December was "the establishment of an industry-wide AI development system", Ge Honglin said. Through the publication of these vertical models Chinalco and the association also want to "lower the barriers for small- and medium-sized enterprises to adopt AI".

The CNMIA also promotes the development of standards and "industry-wide specifications for training, evaluation, and security of large models", it repeated at the launch ceremony for the "Kun'an 2.0 Large Model".

Another noteworthy effort with regard to the practical use of these AI models is Chinalco's development of "eight industry-grade data sets covering key areas such as ore body identification, the optimization of energy consumption, equipment predictive maintenance, and improved metal recovery rates", as reported by the Chinese platform AI Base. (In Chinese).

All of these developments in the nonferrous metals space in China are relatively new, and engineers and AI experts in China continue to wrestle with the same problems of siloed and inconsistent datasets, legacy control systems, or distrust of black-box recommendations as seen elsewhere globally. The pace of AI deployment is clearly accelerating, however.