
Ken Sullivan, co-founder and CEO of Bay Compute, addresses the critical power bottleneck hindering AI growth and the prevalence of "stranded capacity" within datacenters. While hyperscalers have historically optimized their infrastructure, colocation facilities often operate at only 30-50% average power utilization due to the spiky nature of AI inference workloads. As new grid connections now face six-to-seven-year delays, these facilities must shift from simply leasing power to actively optimizing it. Sullivan advocates for a holistic, software-driven orchestration layer that manages compute demand alongside diverse power sources like Battery Energy Storage Systems (BESS) and behind-the-meter natural gas. This systemic approach aims to free up 10-20% of additional capacity, moving beyond the diminishing returns of stacking individual hardware improvements like denser racks or more efficient cooling.
Sign in to continue reading, translating and more.
Open full episode in Podwise