Significant improvements in the energy efficiency of AI hardware are a prerequisite for the growing spread of this computationally intensive technology. Science and business, nationally and internationally, are currently researching such solutions at great expense. On the one hand, to counteract the rapidly increasing energy demand in the field of information and telecommunications technology and, on the other hand, to address certain applications, e.g. B. in mobile devices. Mobile end devices that are to offer their full functionality without a data connection to a data center require efficient AI hardware. But scenarios in which information cannot leave the local context for data protection reasons also benefit enormously from corresponding developments, for example in medical technology. Industry 4.0 is also relevant. Automation solutions that are to become more agile and intelligent require the use of AI-supported control systems. If these are compact and energy-efficient, they can enable real-time AI close to the action. The following three main objectives are being pursued as part of this project: 1 The methods and algorithms for evaluating energy-efficient AI solutions that are required for the implementation of the pilot competition for the Federal Agency for Jump Innovations are to be systematically researched, implemented and tested, whereby - On the one hand, methods for the classification of EKGs are examined and criteria for the quantitative recording of a minimum performance are to be derived, and - On the other hand, methods for determining the energy efficiency of AI platforms are to be developed. 2 Based on the methodology developed above, the procedures proposed in the competition are to be evaluated and assessed. Furthermore, by comparing and identifying synergies, comprehensive approaches that can improve energy efficiency are to be identified. Basic recommendations for the implementation of AI in terms of energy efficiency, data protection, manageability, platforms and architecture are to be derived. Research needs in the field of AI microelectronics are to be identified. 3 Using the example of the analysis of ECG data, the fundamental problem of evaluating the performance of AI solutions is to be examined. In view of the current efforts to include AI solutions in standards, the fundamental question of how the performance of AI solutions can be recorded quantitatively and how the minimum performance specifications required in standards can be meaningfully defined.