Companies have been talking about gigawatt-scale demand response for last couple of years (ie. reducing individual data center load based on grid strain), but Google is the first hyperscaler to actually do it in for real and at scale. It probably related to the growing share of AI workloads, these workloads are easier to move around compared to traditional cloud workloads that are more latency-sensitive and harder to touch (5x9's SLAs for availability & latency).
Companies have been talking about gigawatt-scale demand response for last couple of years (ie. reducing individual data center load based on grid strain), but Google is the first hyperscaler to actually do it in for real and at scale. It probably related to the growing share of AI workloads, these workloads are easier to move around compared to traditional cloud workloads that are more latency-sensitive and harder to touch (5x9's SLAs for availability & latency).