Predictive maintenance reduces downtime and maintenance costs by early detection of equipment failures using sensor data and machine learning techniques. Despite its potential, the progress of this technique is usually affected by limited access to real-world industrial data. Industrial datasets are often not shared due to confidentiality concerns, competitive interests, and regulatory constraints such as the GDPR. This article discusses the challenges of industrial data access, explores alternative approaches to enable data-driven collaboration, and outlines common technical challenges associated with real-world datasets. Addressing these challenges through collaborative, privacy-preserving data sharing is essential for enhancing maintenance strategies and enabling closer collaboration between industry and academia.