AI workloads are rapidly redefining what a modern data center must deliver, from rack densities and power distribution to network architecture and deployment timelines. As hyperscalers, neocloud providers, and enterprises race to stand up GPU capacity, operators and suppliers alike are confronting a new class of design and operational tradeoffs. This interactive roundtable will explore what’s actually changing inside AI facilities today and what remains unresolved. Participants will examine real-world deployment lessons, emerging design patterns, and the practical constraints shaping next-generation AI infrastructure.