Dafna Granot on solid-state transformers
Dafna Granot from SolarEdge explains how SSTs and full-DC systems enable high-density AI computing efficiently.
Image for illustrative purposes
The push for more computing power per square metre is putting solid-state transformers (SSTs) at the centre of AI data centre design. Dafna Granot, senior manager for strategy and innovation at SolarEdge, says the next-generation racks, moving from 10 kW to 100 kW and eventually 1 MW per unit, cannot be supported by traditional AC infrastructure, prompting a shift to full-DC architectures.
Currently, power arrives as AC and is converted multiple times before reaching DC chips. SSTs combine transformer and converter functions in a single device, enabling higher voltage DC directly to racks – around 800 V – improving efficiency and reducing space requirements. SolarEdge, using advanced silicon carbide components from Infineon Technologies, is developing compact, highly efficient SSTs for these high-density AI workloads.
Granot notes efficiency is crucial: even small gains allow more compute per unit of electricity. Designing for medium voltages up to 34.5 kV adds complexity but increases flexibility. Unlike traditional custom transformers, SSTs use modular, mass-producible components for faster deployment.
SolarEdge plans to launch its AI-ready SST product by 2027, coinciding with the next generation of AI systems. Full-DC architectures also facilitate solar and battery integration, replacing legacy AC UPS systems and improving simplicity and efficiency. Granot describes SSTs and DC-based data centres as the natural evolution to meet the growing energy demands of AI.
Source: PV Magazine
#AI data centres#Dafna Granot#power efficiency#SolarEdge#solid state transformers