Over the past few years, there has been a swift advancement in deep learning, which has seen widespread adoption across various domains. Cryptographic device side-channel information analysis, a technique used in side-channel attacks, significantly imperils key security. In this paper, the research status and main methods of side-channel attack are investigated and analyzed. At the same time, the current application status of artificial intelligence technology in side-channel attack is explored. We compare and analyze the specific performance of CNNbest, MLP and Transformer in side-channel attack. The evaluation criteria used include the average ranking of the correct key, model accuracy under the same training time, and the time required to achieve the same accuracy. The influence of different epochs on the attack schemes based on different models is also explored. Through experiments, we prove that among the three models, CNNbest has higher efficiency and accuracy in the attack. Finally, we implement the system based on artificial intelligence technology through coding. In the system, we provide the above three models to choose from to realize the AI enablement of the side-channel attack technology.