Along with the growth in unstructured data, the workloads being applied to that data are becoming less structured. This is because clouds have (re)taught us what HPC sites already knew, that compute should be shared for highest utilization. Shared compute requires shared storage and results in many storage workloads being applied at the same time. The explosion of AI/ML brings an entirely new workload that needs to be supported without buying a new storage system.
No one wants to buy special purpose storage. Instead, storage should serve all your needs without having to sync data across different systems for different uses.