In this decade it is expected that the semiconductor market reaches USD 1 trillion. This growth is especially driven by electrification and renewable energy, connectivity and artificial intelligence, which pushes industry4.0 worldwide. Due to the immense request for semiconductor devices, the microelectronic industry is under extreme pressure. An important trend in microelectronics is miniaturisation and system integration. This means integration of more functionality in smaller volume. Especially heterogeneous system integration, which means integration of different chips into one package, demands different materials and creates complex challenges to understand the interaction between these materials.

Failure analysis is becoming a major enabler to support the development of new devices, characterize new materials and understand their reliability and quality, analysing material interfaces, as well as understand problems within production and react quickly on field returns.

On the one hand, the failure analysis process requires a lot of different physical measurement methods, therefore development of new methods is crucial. On the other hand, measurement data and images must be analysed correctly to get the required understanding of the results. In this field artificial intelligence (AI) offers new possibilities to support the failure analysis process.

Thus, in the Eureka Penta/Euripides project FA4.0 started a first approach to apply methods of AI in failure analysis tools and data analysis methodology. In this presentation based on examples we show the impact of AI for failure analysis. These include image-based applications, e.g. void recognition in X- ray images, signal-based applications, e.g. the AI-gate within SAM, as well as automated understanding of FA-databases. We also learned within this project that standardization is crucial for daily usage of AI within failure analysis, which will be shown based on the example of integrated workflows. A follow- up project for FA4.0 is close to start to develop methods for AI-ready data infrastructure systems and improved AI-methods. Main ideas and motivation for this project will be shown.