Bloomberg: AI Firms Redesigning Data Center Electrical Architecture — Sidecars and Solid-State Transformers Target 20–27% Efficiency Gains for 500kW Racks
Summary
Bloomberg's June 1 longform examines two competing electrical redesigns now reaching commercial scale inside AI data centers: sidecar AC-to-DC converters improve energy efficiency 20% and enable racks to reach 500 kilowatts — roughly 10× the pre-AI norm — while solid-state transformers targeting a 27% efficiency gain are in active commercial competition. The redesigns are driven by Nvidia's escalating chip power budgets, which require data centers to draw 'many times more power than their predecessors,' putting pressure on electricity prices, carbon footprint, and the pace of the AI buildout itself.
Originally reported by bloomberg.com
Read the original article →Original headline: Bloomberg: AI Firms Redesigning Data Center Electrical Architecture — Sidecars and Solid-State Transformers Target 20–27% Efficiency Gains for 500kW Racks