Power and Compute Convergence

Overview

Power and Compute Convergence refers to the integration of power delivery, cooling infrastructure, and IT hardware into a single co-designed stack — in contrast to the traditional data center model where these three disciplines are handled by separate, sequentially collaborating teams (or separate business units that barely interact). As AI rack densities reach 1 MW, the constraints imposed by each discipline directly interact: power architecture determines available voltage for compute, voltage determines cooling requirements, and cooling capacity caps compute density. Designing each in isolation is no longer viable.

The organizational dimension is equally important. At many large data center operators, power, cooling, and IT teams are completely siloed — separate business units that don't converge until four levels up in the org chart. Flex argued at OCP Global Summit 2025 that this structure must change, citing a customer anecdote where an IT-side engineer realized late in the process that not engaging the power team early meant not getting power at all. The benefits of convergence — time-to-market, scale, safety, efficiency, and resilience — all depend on cross-functional collaboration starting at project inception.