Power has become the gating factor for AI infrastructure, and the industry is responding by fundamentally rewriting how data centers are planned, powered, and delivered.

Interconnection delays, grid congestion, and speculative queue dynamics are forcing a shift away from traditional utility-dependent models toward power-first development strategies. In response, developers and operators are assembling a new energy stack combining on-site generation, combustion-free baseload technologies, large-scale battery storage, and advanced cooling architectures, the better to accelerate timelines and ensure operational certainty. What’s emerging is not a single solution, but a coordinated system. Power, cooling, and compute are no longer designed independently; they’re being engineered together as an integrated architecture capable of supporting ultra-high-density AI workloads. At the leading edge, this means rethinking not just how power is delivered, but how it is consumed. New approaches to system design, thermal management, and infrastructure integration are reshaping what “power-first” actually looks like in practice.

This panel brings together leaders across on-site generation, fuel cells, energy storage, and next-generation data center architecture to examine how projects are being built in spite of grid constraints, and how power and infrastructure are being co-designed to deliver capacity at AI scale. The discussion will explore how these systems come together in practice, where they succeed, and what it takes to make them repeatable.