Technology industries are becoming increasingly competitive, and in such environ-ments, the veracity of companies’ decision-making directly affects the future development of enterprises. Therefore, the way in which an enterprise formulates and constructs a set of appropriate decision-making systems, to accurately predict future market trends, is a particularly important issue. In the study presented here, an artificial intelligence-based prediction system was used to estimate manufacturing capacities and client demands, which can provide manufacturing managers with a point of reference for inventory ar-rangements, so that stockholdings can be adjusted appropriately to avoid excessive inven-tory levels. In recent years, neural networks have been widely and effectively applied to many prediction problems. A key reason for their popularity is that backward neural net-works can be used to construct non-linear models. Here, we propose a prediction model combining grey correlation and a neural network, which can be used to establish a high-accuracy prediction system for integrated circuit (IC) production. Firstly, grey correlation analysis was used to screen for the most relevant factors. These were then inputted into the neural network prediction model for training and prediction. We found that the training prediction error and the empirical error value were about 14%, indicating good prediction ability and the suitability of the proposed prediction model for use in the case of IC sub-strate production. Our findings can serve as a point of reference for the design of other predictive systems and support accurate, convenient and fast decision-making that will enhance companies’ competitiveness.