Could your home become a mini AI data center? For @0x7SUN, CEO of @flock_io, inference opens up that possibility. Large-scale training needs tightly connected GPUs in one place. Inference can be less demanding about where it runs. His example is @gmcbhutan and its hydropower supply: run inference when power is available, and reduce computing hours during lean periods. “There's an uptime and a lean time. But for inference, that's actually tolerable.” He sees that opening up locations previously considered unsuitable for AI data centers - potentially even our homes. The question is no longer just where to build the next giant data center. It’s how small and distributed AI compute could become.