ABB launches Infinitus portfolio for 800 VDC AI data-center power architectures
Редакция ЦОД:Главное
ABB has introduced Infinitus, a Source-to-Rack power portfolio for 800 VDC AI data-center architectures. The company expects selected key components to enter production within 12 months and estimates that the first native DC data centers using the full portfolio could be built and commissioned in two to three years.
- Infinitus combines SSTs, medium-voltage equipment, power conversion, DC distribution and protection, and cooling optimization in one Source-to-Rack portfolio.
- ABB expects selected key components to enter production within 12 months; full-portfolio native DC deployments are projected for two to three years.
- ABB and NVIDIA have collaborated since October 2025 on an 800 VDC architecture for gigawatt-scale AI data centers.
- ABB projects DC distribution could account for 25% to 40% of newly added data-center capacity by 2030.
- The cited efficiency and revenue figures are modelled estimates for a 500 MW data-center example, not operating-project results.
ABB has launched Infinitus, a Source-to-Rack portfolio intended for native DC data centers and hybrid AC/DC facilities using 800 VDC architectures. The offering combines solid-state transformers (SSTs), medium-voltage systems, AC/DC conversion, DC distribution and protection, and cooling optimization.
The portfolio addresses the increasing electrical and thermal demands associated with high-density AI infrastructure. The source cited by ABB indicates that current AI rack power is approaching 200 kW of IT load and that next-generation racks could exceed 1 MW of IT load. At those densities, distribution topology, conversion stages, cable infrastructure, cooling integration and equipment footprint become central considerations in data-center design.
ABB expects several key Infinitus components to enter production over the next 12 months. It estimates that the first native DC data centers deploying the complete portfolio could be constructed and commissioned within two to three years. These are ABB expectations, rather than confirmed deployments or operating projects.
ABB has also been working with NVIDIA since October 2025 on an 800 VDC power architecture for AI data centers at gigawatt scale. The available information does not identify a specific site, project capacity or construction status for such a facility.
ABB estimates that DC distribution could be used in 25% to 40% of newly added data-center capacity by 2030. Separately, an ABB and BCG model suggests that DC distribution could improve energy efficiency by more than 5%. In the model, a 500 MW total-project data center with a 5% efficiency improvement could make about 25 MW of additional power available to IT equipment and support more than $300 million in potential annual revenue. These figures are illustrative model outputs, not reported performance from a specific project.
If ABB meets its stated commercialization timetable, Infinitus could support the emergence of an integrated DC-infrastructure segment for high-density AI data centers. That could create competition around architecture-level standards spanning SSTs, conversion equipment, DC protection and cooling systems, rather than solely around discrete electrical products.
As AI rack power density rises, power distribution, cabling, cooling and electrical-equipment footprints become more consequential design constraints. ABB’s integrated approach could shift competition from individual power components toward end-to-end DC power architectures for high-density AI campuses, although its commercialization and market-adoption projections remain unverified.
- TechNews 科技新報 / КитайскийПервоисточник →
Материал подготовлен редакцией ЦОД: Главное на основе указанных первоисточников.