Not just local AI inference, plenty of broader workstation workloads will benefit from a modern APU with a lot of VRAM-equivalent.