Overview of ABB's Infinitus Portfolio

ABB has officially launched Infinitus, a comprehensive source-to-rack direct-current power portfolio explicitly engineered for the next generation of AI data centers. As artificial intelligence workloads drive server architectures toward megawatt-class configurations, the industry faces severe constraints in moving massive amounts of electricity from the grid to processors. Infinitus addresses this by eliminating the power and floor space wasted on repeated AC-DC conversion stages, establishing a streamlined 800 VDC infrastructure from the utility source directly to the compute rack. This represents a fundamental paradigm shift in facility design, moving away from fragmented alternating current systems toward a unified direct current ecosystem designed specifically for the unique thermal and electrical profiles of modern AI accelerators.

Technical Architecture and Solid-State Transformation

At the core of the Infinitus architecture is ABB’s advanced solid-state transformer technology. This critical component is designed to convert medium-voltage alternating current power directly into an 800 VDC distribution network. By bypassing traditional multi-stage conversion processes, the system significantly optimizes power delivery, reduces harmonic distortion, and minimizes the physical space required for heavy, low-frequency magnetic components. The broader portfolio encompasses medium-voltage powertrains, DC power conversion and quality equipment, comprehensive DC distribution, and specialized DC protection mechanisms. A notable feature is the integration of ABB’s HiPerGuard platform, which is engineered to connect directly at 34.5 kV, providing a robust medium-voltage interface that anchors the entire DC-native ecosystem.

Efficiency Gains and Footprint Reduction

The transition to an 800 VDC architecture yields substantial improvements in end-to-end energy efficiency. According to ABB’s internal analysis, deploying DC distribution can improve overall energy efficiency by more than 5 percent. To contextualize this gain, at a 500 MW data-center campus, a 5 percent efficiency improvement translates to approximately 25 MW of additional power made available strictly for revenue-generating IT equipment. Furthermore, the architecture drastically reduces conversion stages and associated heat losses. This efficiency directly translates into a reduced physical footprint for power equipment, a critical advantage as hyperscale AI facilities fiercely compete for limited grid capacity and strive to maximize GPU density per available megawatt while keeping operational expenditures strictly in check.

Integrated Protection, UPS, and Cooling Systems

Beyond primary power conversion, Infinitus integrates critical auxiliary systems directly into the DC architecture. The portfolio includes medium-voltage uninterruptible power supply technology, advanced switchgear, and solid-state circuit protection. ABB’s DC protection equipment is specifically designed to interrupt electrical faults instantaneously without taking unrelated, operational equipment offline, thereby ensuring maximum uptime for continuous, multi-week AI training workloads. Additionally, the system extends its DC-native philosophy into facility cooling through the integration of DC-connected variable-speed drives and motors. This holistic approach allows operators to coordinate the entire electrical and thermal path, rather than optimizing isolated components in a fragmented environment, ensuring that thermal management scales seamlessly alongside computational workloads.

Market Context and the 800 VDC Race

The launch of Infinitus accelerates the broader industry race toward 800 VDC architectures for next-generation data centers. The shift is driven by the exponential power demands of modern AI accelerators, which are pushing individual rack requirements to 1 MW and beyond. Recognizing this trajectory, ABB estimates that between 25 percent and 40 percent of all new data-center capacity installed by 2030 will utilize DC distribution. This strategic pivot aligns with parallel industry developments, including Huawei’s recently introduced 800 VDC-ready grid-interactive AI data-center architecture and Mitsubishi Electric’s chip-to-grid reference designs tailored for NVIDIA ecosystems. Together, these initiatives signal a definitive industry consensus that legacy infrastructure is insufficient for the AI era, necessitating a ground-up rethinking of power delivery.

Deployment Timeline and Future Outlook

Looking ahead, ABB has outlined a clear deployment roadmap for the Infinitus portfolio. Initial components of the system are scheduled to enter production over the next 12 months, allowing early adopters to begin integrating DC-native elements into their infrastructure upgrades. Complete, fully DC-native facilities utilizing the entire Infinitus portfolio are expected to become fully operational within two to three years. As AI data centers continue to scale toward unprecedented megawatt-class densities, the transition to 800 VDC architectures like Infinitus will become a foundational requirement, redefining how global compute infrastructure is powered, protected, and cooled for the foreseeable future, cementing DC distribution as the new standard for hyperscale compute.