AI workloads are forcing data centers to rethink established priorities. Rising chip power, increasing rack densities and growing energy demand are putting pressure on operating costs and sustainability goals, while conventional air cooling is reaching its limits. This raises a central question: what matters most—maximizing performance within a given power envelope, minimizing energy-related costs, or building greener IT? And do data centers really have to choose?
This talk explores that question from the perspective of cooling and infrastructure design. It compares air cooling, single-phase liquid cooling and dual-phase liquid cooling, and discusses how these approaches affect efficiency, power overhead, achievable density and sustainability potential. The presentation argues that performance, cost efficiency and green IT do not always have to remain competing objectives. With the right cooling architecture, they can be aligned more effectively—while still requiring careful consideration of implementation effort, regulation, maintenance and site-specific constraints.
