This episode of our AI CEO series looks at the potential for AI capabilities to address one of AI's largest challenges – energy availability. Dr. Varun Sivaram, founder and CEO of Emerald AI, joins host Eric Hanselman to explore how greater flexibility is AI workloads can unlock unused capacity. Traditionally, data center construction has focused on delivering peak capacity, even at those times when it may not be needed. That doesn't mesh well with a power grid that is already under stress. AI intelligence can shift workloads dynamically to reduce data center demands when the power grid needs it and to leverage excess grid capacity when it's available.

Flexing AI workloads to respond to grid conditions isn't simple, but it can be more attractive than alternatives. Adding generation capacity is a long-term and expensive process and community concerns about rate increases are creating headwinds for new data center builds. Adding Battery Energy Storage Systems (BESS) are costly and have their own capacity and durability limitations. Flexing workload demand could address grid integration problems on a much shorter timeline.

More S&P Gobal content:

• 2026 Trends in Data Center Services & Infrastructure
• Geopolitics of data centers: An AI showdown that will reshape the world
• AI's global resource race: Challenges and opportunities
• Hyperscaler earnings quarterly: What price inference?

For S&P Global subscribers:

• Data Center Services & Infrastructure Market Monitor & Forecast: US ISO Region
• BYOP emerges as a strategic decision for data center operators in response to grid challenges
• US Independent System Operators 101
• Service providers race to meet surging enterprise demand for AI infrastructure
• Survey Data Hub – VoCUL: Connected Customer, AI & Data Centers 2026

Host/Author: Eric Hanselman

Guest: Varun Sivaram

Producer/Editor: Dylan Scheible

Published With Assistance From: Feranmi Adeoshun and Sophie Carr