Skip to content
AI Usage3 min read

Infineon and Eaton Partner on 800VDC Power for AI Data Centers

Infineon and Eaton partner to deploy silicon carbide power devices in medium-voltage solid-state transformers to support 800VDC power delivery in high-density AI data centers.

AB

Author

AUG Bot

Published

Silicon carbide solid-state transformer technology powering an 800VDC AI data center rack

Infineon and Eaton Partner on 800VDC Power for AI Data Centers

Silicon carbide solid-state transformers target 1MW rack density efficiency

Semiconductor manufacturer Infineon Technologies and power management leader Eaton have partnered to supply silicon carbide power devices for Eaton's medium-voltage solid-state transformer platform. The collaboration aims to accelerate the transition to 800VDC power distribution architectures in hyperscale AI data centers, drastically reducing power conversion losses as rack power densities approach 1 MW.

Key details

Eaton will integrate Infineon's silicon carbide (SiC) power devices into its Medium-Voltage Solid-State Transformer (MVSST) 2.0 platform. Traditional data center electrical systems draw alternating current (AC) from the utility grid and route it through multiple AC-to-DC conversion stages before powering IT server racks. This multi-step conversion process causes significant electrical transmission and heat losses at high load volumes.

Eaton's MVSST 2.0 platform bypasses several intermediate conversion stages to deliver power directly to 800VDC busways. Silicon carbide devices offer higher switching speeds, superior thermal conductivity, and lower energy dissipation compared to traditional silicon components. The companies plan to explore higher-voltage 2.3kV and 3.3kV SiC power modules to support future ultra-high-density AI facilities and utility grid connections.

Why this matters

Next-generation AI supercomputers, such as Nvidia's Blackwell and Rubin platforms, are pushing individual rack power requirements toward 100 kW to 1 MW. Legacy 415V AC or 48V DC power distribution systems cannot efficiently transmit such massive current levels without experiencing heavy resistive heat losses, copper cabling bottlenecks, and severe PUE degradation. Shifting to 800VDC distribution combined with solid-state transformers enables higher efficiency power delivery, reducing electricity waste and operational thermal overhead across AI clusters.

Context

As hyperscalers expand AI compute capacity, electric grid capacity limits and internal facility power delivery are emerging as primary bottlenecks. Industry suppliers are shifting toward direct medium-voltage grid connections and direct-current topologies to maximize energy efficiency. This partnership reflects a broader hardware push toward 800VDC infrastructure, following recent power architecture announcements from major chipmakers and power equipment vendors aimed at supporting gigawatt-scale AI deployments.

What happens next

Infineon and Eaton will roll out SiC-based MVSST 2.0 units for commercial data center pilot deployments while co-developing 2.3kV and 3.3kV modules. Utility grid operators and data center engineers will evaluate these solid-state transformers as hyperscalers prepare to deploy 1MW server rack architectures in late 2026 and 2027.


Source: Data Center Dynamics Published on AI Usage Global, author: AUG Bot

Older post
Related

Read more

More posts that expand on the topics, companies, and AI trends covered in this story.

Digital representation of a grid-reactive AI data center in Oklahoma
AI Usage

Lambda Plans Grid-Reactive AI Data Center in Oklahoma

Lambda plans a closed-loop and grid-reactive AI data center in Chouteau, Oklahoma, promising $500M in local taxes while covering all power infrastructure costs.

Digital representation of data center site plans and power grid infrastructure in Sydney
AI Usage

Goodman Withdraws Application for 90 MW Sydney AI Data Center Campus

Goodman Group drops its AU$1.2 billion, 90 MW Project Mars data center application in Lane Cove, Sydney, citing evolving state policy frameworks and local community feedback.

High-tech enhanced geothermal energy plant generating clean electricity for data centers
AI Usage

Fervo Energy Achieves First Power at 900 MW Utah Geothermal Plant

Fervo Energy reaches first power at its Cape Station enhanced geothermal project in Utah, supplying firm zero-carbon electricity to support Google's expanding AI data centers.