AI ambition is colliding with physical reality.

As data centers push from 20-kilowatt racks to 120-kilowatt systems—with megawatt-per-rack infrastructure on the horizon—the limits of power and cooling have become impossible to ignore.

In this episode of Innovation Heroes, Ed McNamara explores the collision between AI ambition and energy reality with two guides who see the problem from opposite ends of the timeline.

Brian Smith of Idaho National Laboratory takes us inside the nuclear renaissance, where small modular reactors are being designed to deliver massive, carbon-free power on dramatically shorter timelines—"like building airplanes, not airports."

Meanwhile, Ryan Hotchkin of SHI International brings the conversation back to today's constraints: liquid cooling becoming mandatory, power capacity maxed out, and cloud costs rising for everyone.

Together, they map what IT leaders need to understand now—and what's coming next—if AI initiatives are going to scale instead of stall.

Key Takeaways:

• Why AI workloads are overwhelming traditional data center power and cooling


• How small modular reactors could reshape long-term energy infrastructure


• The math problem: power density doubling every generation of GPUs


• What IT leaders must do now—even if deployment is three years away

Listen on Apple Podcasts, Spotify, or watch on YouTube shi.com/heroes

#AI #DataCenter #NuclearEnergy #ITInfrastructure #Innovation

• Brian Smith — Nuclear Reactor Development Director, Idaho National Laboratory


• Ryan Hotchkin —Sr. Director, Advanced Solutions Group, SHI International

Related Resources:

• SHI’s Next Gen Infrastructure Solutions: https://www.shi.com/solutions/next-generation-infrastructure


• SHI GenAI Solutions: https://www.shi.com/solutions/generative-ai