Traditional NIR systems are stand-alone and difficult to model. In order to share resource of existed NIR model, a NIR analysis system based on cloud computing is proposed. The NIR software analysis system is designed on a high performance server instead of the host of stand-alone version, and the architecture and design procedures of the NIR cloud analysis system are described in detail. The system has the functions of preprocessing of near infrared spectral data, quantitative analysis, qualitative analysis, spectral model search and spectral model transfer. The identification results of waste edible oil between NIR cloud analysis system and stand-alone version are compared. The overall rate of correct identification is 86.21% for the 50 samples of waste edible oil by the NIR cloud analysis system. This result is fully consistent with the analysis results in stand-alone. The experimental results show that the cloud NIR analysis system is low cost, easy modeling, access to flexible, enabling resource sharing and remote access, etc.