Meta CEO Mark Zuckerberg recently highlighted that as artificial intelligence technology rapidly advances, electricity supply is poised to become the "new bottleneck" in the AI sector, potentially succeeding computational hardware. This perspective suggests a shift in the focus of AI infrastructure development.
Concurrently, the previously tight supply of GPUs (Graphics Processing Units) in data centers, a long-standing concern for the industry, is gradually easing. While this indicates a reduction in hardware supply pressure for AI training and inference, new challenges such as power consumption are emerging on the horizon.