ABB Launched Direct Current Power Portfolio for Data Centers
The Infinitus line enables 800 VDC power distribution to support high-density AI computing racks.
Updated on Sept. 28, 2026 in Data Centers
Live Poll
Should data centers be required to adopt high-efficiency power infrastructure to support increasing AI demands?
ABB has launched its Infinitus direct current power portfolio designed to improve data center energy efficiency by over 5 percent. The system is specifically engineered to meet the rising power demands of next-generation AI workloads.
Why it matters
As AI chip power requirements surge from current levels of 200 kW to 1 MW per rack, existing AC power infrastructure faces severe density constraints. This shift toward direct current architecture allows for higher efficiency and capacity scaling in the next generation of data centers.
The Infinitus system operates at 800 VDC to reduce infrastructure footprints by eliminating multiple power conversion stages. This architecture supports a five-fold increase in rack power density, scaling from current 200 kW deployments to 1 MW per rack.
The players
ABB
A multinational technology company specializing in electrification and automation systems with 25 years of experience in direct current technologies.
NVIDIA
A designer of graphics processing units and high-performance computing platforms that currently dominates the market for AI chip hardware.
The details
The portfolio integrates solid-state transformer technology, static medium-voltage UPS systems, and solid-state circuit breakers to manage energy flow without traditional mechanical bottlenecks. By utilizing DC-connected variable-speed drives and motors, the system optimizes cooling energy consumption. The architecture removes the conversion stages necessary in conventional AC systems, effectively reducing the physical and electrical infrastructure requirements.
Timeline
October 2025: ABB and NVIDIA began collaborating on 800 VDC power architectures.
Next 12 months: Several main components of the Infinitus portfolio enter production.
Within two to three years: First DC-native facilities are expected to be operational.
2030: Projected market usage of DC distribution in 25 to 40 percent of new data center capacity.
The Tech Race
This launch places ABB at the forefront of the transition to DC-native power grids within the high-performance computing sector. It represents a significant move to standardize the infrastructure needed to support the extreme electrical loads of next-generation AI hardware.
The transition to 800 VDC systems will primarily change site power design for massive data center operators and high-density compute providers. While the first native facilities are not expected for two to three years, the deployment of components will begin within the coming year.
The takeaway
The industry is moving toward high-voltage DC distribution as the only viable path to support the next generation of 1 MW compute racks. Stakeholders should track the production rollout of these components over the next 12 months as a bellwether for the adoption of DC-native data centers.
What happens next
Production of main components is scheduled for the next 12 months, which will serve as the primary indicator for industry adoption rates.
Further reading
For more on the evolving requirements of infrastructure, see our section on Data Centers.
More information
For more details on the portfolio specifications, visit the ABB corporate information portal.
Source note: This article includes information reported by Engineering.
Live Poll
Should data centers be required to adopt high-efficiency power infrastructure to support increasing AI demands?
