The Physical Limits of AI: Grid Power and Data Center Realities
Summary: Scott Donahue, VP of Grid Power Services at Sargent & Lundy, leads a closed-door session on the physical limits of scaling enterprise AI. It strips away software hype to confront the real bottleneck: power grids and data center capacity, not code. Focus: how CTOs and boards must factor energy constraints, grid stabilization, and utility delays into AI strategy.
The boardroom conversation around Artificial Intelligence is currently dominated by software: token limits, foundational models, and compute costs. But the actual bottleneck for the AI revolution is not code—it is electricity and concrete. As enterprises rapidly scale their AI workloads, they are colliding with the brutal physical limits of the national power grid and data center capacity.
You cannot run exponential AI algorithms on a linear, legacy power grid.
This closed-door session strips away the software hype to tear down the physical realities of scaling enterprise AI. To anchor this discussion, we are locking the doors with Scott Donahue. Scott is a rare executive who understands both the digital scale of the Fortune 1 and the hardcore engineering required to physically power it. As the Vice President of Grid Power Services at Sargent & Lundy, he currently designs and delivers high-voltage utility interconnections, capacity upgrades, and resilient backup power systems for the world's largest hyperscalers and Fortune 100 technology clients.
Bringing a foundation in commercial nuclear power engineering , combined with his experience as the Vice President of Digital Transformation at Walmart—where he spearheaded the migration of hundreds of on-premise server applications to cloud-native platforms —Scott will break down exactly how CTOs and Boards must factor energy constraints, grid stabilization, and utility delays into their cloud and AI strategies.