The availability of software tools that align with modern hardware designs has been an important enabler of AI deployment, reducing the time taken for software development and allowing easier access to GPUs. Open-source development has been central to the success of this approach, with software published under flexible BSD-licenses. Maintaining this culture of open-source software development and code sharing will be vital for the continuing success of the field. In the context of a trend for larger models, effective management of access to compute will become increasingly important, balancing the demand for increased compute with the need to ensure energy efficiency in AI. To chart a path towards delivering both these outcomes, ELISE has created a roadmap for compute infrastructure in Europe.