Data Center Power Conversion, Thermal Management and Safety

AI data centers are projected to consume hundreds of terawatt-hours annually, demanding power conversion technologies that are not only energy-efficient but also cost-effective, scalable, and reliable. To support the rapid growth of AI workloads, innovative power electronics solutions are urgently needed. This presentation explores the development of a novel megawatt-scale, medium-voltage (MV) power conversion system based on a split-air-gap solid-state transformer architecture. The approach aims to reduce energy loss, footprint, and cost while enabling flexible, modular deployment and achieving more reliable operation.

AI data centers are pushing power-conversion systems toward higher switching frequencies, higher power density, and tighter efficiency targets. This presentation examines the system-level tradeoffs of GaN-based high-frequency conversion, including semiconductor losses, magnetics, parasitic effects, EMI/EMC, thermal behavior, and reliability. Rather than focusing on device specifications alone, it shows how switching frequency and converter architecture must be co-optimized with magnetics and electromagnetic performance to achieve meaningful system-level gains.

AI data centers are scaling faster than the water systems that support them. This presentation examines water as a critical infrastructure layer connecting cooling demand, municipal supply, reclaimed water, stormwater, wastewater, recharge, and watershed conditions. Attendees will learn how earlier water planning can reduce project risk, expand reuse and repurposing options, improve community acceptance, and create measurable local water benefits. The session also explores how monitoring and AI-enabled decision support can improve treatment, allocation, reuse, recharge, and long-term watershed performance.

Every data center is a physics problem in disguise. Power, heat, water, and space compete for the same margin, settled by spreadsheets and intuition. Scott Charter, Head of GTM – Data Centers at Physical Superintelligence (PSI), shows how AI "virtual physicists" are rethinking infrastructure from the molecule up, on Earth and in orbit, by using PSI's Emmy engine to reason across these interconnected design challenges, proving the math before capital is committed.

AI rack densities are rapidly moving beyond 100 kW, with emerging systems approaching MW-scale loads. At these densities, simply scaling conventional data center electrical infrastructure creates new challenges in power distribution, redundancy, protection, power conversion, and maintainability. This presentation examines how electrical design assumptions change as rack density increases from 100 kW toward 1 MW and identifies the architectural decisions engineers must reconsider for the next generation of AI infrastructure.

AI is changing what data centers demand from batteries. Higher power density, evolving UPS architectures, and increasingly complex operating profiles create new questions around cell performance, lifetime, reliability, and safety. This presentation will examine those challenges from the cell level up, connecting lithium-ion degradation and failure mechanisms to the engineering decisions that determine whether battery systems deliver the resilience expected in next-generation AI data centers.

AI is reshaping data center design, where site selection has become a strategic differentiator. As HPC, AI, and GenAI compute-intensive workloads surge, operators are racing to develop facilities capable of supporting unprecedented rack-power density, energy consumption, and advanced cooling requirements. The market is seeing rapid growth in AI-dedicated data centers, from MW to GW-scale campuses across hyperscalers, co-location providers, and emerging neoclouds. We examine the evolving criteria that matter most—including power availability, sustainability, and proximity to energy and telecom ecosystems—and emphasize how disciplined due diligence and cross-functional alignment enable faster, more effective decisions in an increasingly competitive landscape.