ABB has unveiled Infinitus, an integrated DC power solution tailored for high-density AI data centers. The company is also collaborating with Nvidia to develop an 800V DC architecture that minimizes power conversion stages and reduces overall energy consumption.
Overview of ABB's Infinitus Launch
ABB has officially unveiled Infinitus, a comprehensive integrated direct current (DC) power solution specifically engineered to address the escalating high-density power demands of modern artificial intelligence (AI) data centers. Announced on September 23, 2026, this strategic launch marks a significant paradigm shift in how power infrastructure is conceptualized and deployed. Infinitus is not merely a component upgrade; it is an end-to-end integrated solution that spans the entire data center power infrastructure, seamlessly connecting the primary power source directly to the server racks. By rethinking the traditional power delivery model, ABB aims to provide a robust foundation for the next generation of computing workloads.
Technical Architecture and Solid-State Transformers
At the core of the Infinitus ecosystem is the utilization of advanced solid-state transformers (SSTs) alongside cutting-edge DC distribution and protection technologies. Traditional alternating current (AC) data center architectures rely on multiple power conversion stages, which inherently introduce energy losses and require substantial physical space for transformers and switchgear. By leveraging SSTs, Infinitus drastically reduces these power conversion stages. This technical refinement not only minimizes the overall energy consumption but also significantly shrinks the installation footprint. Furthermore, recognizing that immediate, full-scale transitions are complex, ABB has designed Infinitus to be highly flexible. The solution can be seamlessly deployed in hybrid configurations that combine conventional AC systems with DC, allowing facilities to modernize incrementally without disrupting existing operations.
The 800V DC Collaboration with Nvidia
A critical pillar of ABB's new power strategy is its ongoing collaboration with Nvidia. Since October of the previous year, the two industry giants have been jointly developing an 800-volt (V) DC power architecture. This partnership is specifically targeted at gigawatt (GW)-scale AI data centers, which are becoming the new standard for training and deploying massive AI models. The primary objective of this 800V DC architecture is to substantially improve power delivery efficiency and scalability. Next-generation AI chips demand unprecedented power density, and the 800V framework is engineered to meet these rigorous requirements by streamlining the electrical pathway and reducing thermal losses associated with lower voltage distributions.
Market Context: The Power Density Challenge in AI
The introduction of Infinitus arrives at a critical juncture for the data center industry. As AI models grow exponentially in parameter size, the power density required per rack has skyrocketed, pushing traditional AC infrastructure to its physical and thermal limits. Data center operators are increasingly constrained by power availability, cooling capacities, and physical real estate. The shift toward DC power distribution is emerging as a vital industry response to these bottlenecks. By eliminating the repetitive AC-to-DC and DC-to-AC conversions that occur at the rack level in conventional setups, DC architectures like Infinitus offer a more direct, efficient, and scalable pathway to deliver the massive wattage required by modern GPU clusters.
Implications for Data Center Infrastructure
The deployment of Infinitus carries profound implications for data center design and operations. By consolidating the power infrastructure from the source to the rack, operators can expect a dramatic reduction in the physical footprint of power equipment. This reclaimed space can be reallocated to house additional compute racks, thereby maximizing the return on investment for the physical facility. Additionally, the integration of DC protection technologies enhances the overall safety and reliability of the power grid within the data center. Giampiero Frisio, president of ABB's Electrification business, emphasized the versatility of the platform, noting that Infinitus is a solution that supports the power delivery required by next-generation AI chips, ranging from pure DC to AC-DC hybrid configurations.
Future Outlook and Production Timelines
Looking ahead, ABB has outlined a clear and aggressive roadmap for the commercialization of Infinitus. Key products within the Infinitus portfolio are scheduled to enter production within the next 12 months, signaling a rapid transition from development to market availability. The company anticipates that the first fully DC-based data center utilizing the complete Infinitus product portfolio will be constructed within two to three years. This timeline underscores the urgency with which the industry is moving toward high-density DC solutions. As the collaboration with Nvidia matures and the 800V DC architecture is fully realized, ABB's Infinitus is poised to become a foundational technology for the GW-scale AI data centers of the future.