ABB has launched Infinitus, a source-to-rack direct-current power portfolio built for AI data centers moving toward megawatt-class server racks.

The architecture targets one of the largest constraints in modern data-center infrastructure: moving enormous amounts of electricity from the grid to processors without wasting valuable power and floor space on repeated conversion stages.

800 VDC from source to rack

At the center of Infinitus is ABB’s solid-state transformer technology, designed to convert medium-voltage AC power directly to 800 VDC. The wider portfolio combines medium-voltage powertrains, DC power conversion and quality equipment, DC distribution, DC protection and cooling optimization.

ABB says next-generation AI chips are expected to push individual rack demand to 1 MW and beyond. The company estimates 25% to 40% of new data-center capacity installed in 2030 could use DC distribution.

More compute from the same megawatts

ABB cites analysis indicating that DC distribution can improve end-to-end energy efficiency by more than 5%. At a 500 MW campus, a 5% gain could make roughly 25 MW more power available for revenue-generating IT equipment.

The architecture also reduces conversion stages and heat losses while shrinking the physical footprint of power equipment. That matters as AI facilities compete for grid capacity and attempt to fit more GPUs into each available megawatt.

UPS, protection and cooling move into the DC architecture

Infinitus includes medium-voltage UPS technology, switchgear, DC distribution and solid-state circuit protection. ABB’s HiPerGuard platform can connect directly at 34.5 kV, while its DC protection equipment is designed to interrupt faults without taking unrelated equipment offline.

The portfolio extends into power and cooling systems through DC-connected variable-speed drives and motors. ABB says the goal is to coordinate the entire electrical path rather than optimize isolated components.

The 800 VDC race is accelerating

The announcement follows a broader move toward 800 VDC architectures across next-generation data centers. BitcoinVersus.tech recently covered Huawei’s 800 VDC-ready grid-interactive AI data-center architecture and Mitsubishi Electric’s chip-to-grid NVIDIA reference designs.

ABB says components of Infinitus will enter production over the next 12 months, with complete DC-native facilities using the portfolio expected to become operational within two to three years.

Sources: ABB announcement and ABB Infinitus technical overview.

BitcoinVersus.Tech Editor’s Note:
We volunteer daily to ensure the credibility of the information on this platform is Verifiably True. If you would like to support to help further secure the integrity of our research initiatives, please donate here: 3C9o19EH5HSiwEPyCTmEKzxhNCbo2X6TTb