TTI IP&E technology development engineer, Serge Lafontaine, explains why power delivery is emerging as a key constraint in artificial intelligence infrastructure
The conversation around artificial intelligence tends to focus on processors. Every new generation of AI hardware arrives with headlines about compute performance, model size and processing speed. Yet as organizations race to deploy AI infrastructure, another challenge is moving to the forefront: power. As rack densities rise and power requirements continue climbing, power conversion efficiency, distribution losses, thermal management and system reliability are becoming central considerations throughout the data center ecosystem.
Capacitors, magnetics, shunt resistors and other passive technologies supporting those functions are receiving renewed attention as a result. For decades, passive components have quietly enabled advances across the electronics industry. The AI buildout, however, is placing them at the center of one of the industry’s most significant infrastructure expansions.
Much of the discussion surrounding artificial intelligence focuses on what happens inside a processor. The larger story may be what happens outside of it. Modern AI accelerators consume substantially more power than the computing platforms that preceded them, creating new demands on the systems responsible for converting, delivering and managing energy throughout a server and across an entire facility.
Higher currents, tighter efficiency targets and greater thermal loads are reshaping decisions surrounding voltage regulation, power distribution and system monitoring. Capacitors play a critical role in managing rapid transient (high di/dt) loads and maintaining voltage stability on low-core-voltage power rails. Magnetics support sophisticated power-conversion architectures, while shunt resistors provide visibility into system performance and energy consumption. As power levels continue to rise, these supporting technologies are becoming more consequential to overall system performance.
The AI infrastructure buildout is altering demand patterns across far more than the semiconductor market. Historically, demand for passive components often tracked broader industrial, transportation and consumer electronics cycles. AI infrastructure introduces a different dynamic, driving concentrated demand around technologies associated with power conversion, power delivery and energy efficiency.
The International Energy Agency projects that the electricity generation needed to supply data centers will grow from roughly 460 TWh in 2024 to more than 1,000 TWh by 2030. That growth carries implications throughout the electronics ecosystem. Every incremental improvement in power efficiency becomes more meaningful when multiplied across thousands of servers and entire fleets of data centers.
The consequences extend beyond system design. Technologies associated with power conversion, power delivery and efficiency improvements can experience market conditions that differ significantly from adjacent product categories. In many applications, these components are not readily interchangeable, making early awareness of technology trends more important as architectures continue to evolve.
While processors may continue to dominate headlines, the technologies enabling efficient power delivery often determine how effectively new infrastructure can be deployed, scaled and supported over time. As AI infrastructure expands, passive components are becoming a more important part of the conversation, shaping efficiency, reliability and long-term scalability across the ecosystem.
