This interview examines, through a dialogue, how industry leaders are responding to the phenomenon of power grids being strained by the rapid increase in power demand within data centers caused by the rapid advancement of artificial intelligence (AI).

According to President Adam White, Infineon prioritizes power efficiency from the early stages of AI system design.

Infineon stated that power efficiency is a competitive factor just as important as performance in modern AI servers where watt-per-compute performance is emphasized, and to this end, it reduces total cost of ownership (TCO) by minimizing power semiconductor losses and increasing component lifespan and reliability.

President Adam White [mentioned] the 'Power first' and 'Grid to Core' initiatives that ensure an efficient and reliable energy flow.He stated that he is focusing on the Chick.

'Grid to Core' is an approach that consistently optimizes the entire power supply chain, from the grid to AI cores such as GPUs and TPUs.

220V AC power is converted in multiple stages inside the data center from 48V to 12V to 1V DC.

Infineon reduced energy loss by about half using SiC (silicon carbide) and GaN (gallium nitride)-based wide bandgap (WBG) semiconductors and advanced packaging technology that increase conversion efficiency at each stage to up to 98%.

The amount of computation required to train the latest AI models has been doubling every 3.4 months since 2012.

As a result, the share of power consumption by data centers is expected to surge from the current approximately 2% of global power to 7% by 2030.

It is predicted that by 2030, data centers in major data hub regions in the United States will use up to 16% of total power, and in Ireland, 32%.

To increase energy efficiency in AI data centers, Infineon is focusing on key areas such as advanced semiconductor materials, optimized power architecture, advanced packaging technology, and smart control and monitoring.


State-of-the-art WBG material minimizes losses even in high-temperature and high-voltage environments as a SiC-GaN power switch.

Advanced packaging maximizes the thermal performance and efficiency of power modules through chip embedding and integrated magnets.

The smart control system optimizes PUE (Power Usage Efficiency) through workload-based dynamic power regulation.

Power and cooling requirements per AI server are up to 30 times higher than those of conventional servers, causing TCO to rise sharply.

Infineon's high-efficiency, high-reliability power components significantly reduce power loss and downtime risks, lowering AI data center operating costs and preventing millions of dollars in losses due to unexpected shutdowns.

Looking at key innovation examples, the TDM2254xD power module supports 160A-class high-density power conversion through the integration of OptiMOS™ MOSFETs.


The 12kW high-efficiency PSU significantly reduces server rack power loss by maintaining an efficiency of over 97.5%.

"As AI has become deeply rooted in all aspects of our lives, power supply solutions must also constantly evolve," said President Adam White. "Infineon is collaborating with global customers and ecosystem partners to pursue both efficiency and sustainability, and is accelerating the development of cutting-edge power semiconductor technologies to lead the era of digital transformation and decarbonization."

In addition, it was mentioned that “energy efficiency innovation in global AI data centers is key to securing environmental sustainability and economic feasibility,” and that “Infineon’s ‘Power First · Grid to Core’ strategy is regarded as the solution to this very challenge.”

Meanwhile, President Adam White'sThe full text of the interview can be downloaded from the Infineon website .