ABB has launched Infinitus, a direct-current power portfolio designed for the rising electrical demands of AI data centers. The source-to-rack architecture brings together medium-voltage power, DC conversion, distribution, protection and cooling, with 800V DC architectures positioned as a way to reduce conversion losses and infrastructure footprint as AI rack power moves toward the megawatt range.
The next bottleneck for AI data centers may not be GPUs, networking or memory. It may be the electrical infrastructure required to power them.
ABB is targeting that problem with Infinitus, a portfolio of direct-current (DC) technologies designed to support high-density AI computing while reducing the physical footprint and energy losses associated with conventional power architectures.
The launch comes as AI accelerators push data-center rack power sharply higher. ABB says next-generation AI chips are expected to drive approximately 1MW per rack and beyond, compared with almost 200kW today. The company argues that this level of power density requires data-center operators to reconsider how electricity is converted and distributed from the grid to the compute rack.
At the center of Infinitus is ABB’s solid-state transformer technology, which forms part of an integrated source-to-rack architecture. The portfolio is designed to support both DC-native facilities and hybrid AC/DC deployments rather than requiring operators to replace their entire electrical infrastructure.
The five building blocks cover medium-voltage powertrains, DC sources and power quality, DC power distribution, DC power protection and cooling optimization.
That integration addresses a fundamental inefficiency in traditional data-center power chains. Electricity may pass through multiple AC-to-DC and DC-to-DC conversion stages before reaching processors and other IT equipment. Moving more of the distribution architecture to DC can eliminate some conversion stages and associated heat losses, although the actual benefit depends on system design, equipment efficiency and operating conditions.
ABB is specifically promoting 800V DC distribution for next-generation AI facilities. Higher-voltage DC can move large amounts of power with lower current for a given power level, potentially reducing conductor requirements and allowing more compact electrical infrastructure.
The company cites a recent BCG/ABB analysis that found DC distribution can deliver efficiency gains of more than 5% while increasing space available for compute racks. ABB estimates that, for a 500MW data center, a 5% efficiency improvement could make 25MW more power available to IT loads. The company also calculates that such an improvement could represent more than $300 million in potential additional annual revenue, depending on how the additional compute capacity is monetized.
Those economic figures are scenario-based estimates rather than guaranteed financial outcomes.
The pressure behind the transition is substantial. The International Energy Agency’s latest outlook projects global data-center electricity consumption will roughly double from 485TWh in 2025 to about 950TWh in 2030. It also estimates that electricity consumption by AI-focused data centers will triple during that period.
The IEA’s analysis highlights why power delivery is becoming an AI infrastructure issue rather than simply a facilities-management concern. AI server power density has risen rapidly, while grid connections, transformers and other electrical equipment are already facing supply and deployment constraints.
ABB’s strategy also reflects a broader convergence between power and compute infrastructure. The company has been working with NVIDIA on 800V DC power architectures for gigawatt-scale AI facilities since 2025. The Infinitus launch extends that work into a broader portfolio covering the electrical path from incoming power to the rack.
This places ABB alongside a growing ecosystem focused on the power requirements of AI infrastructure. Hyperscalers and chip vendors are increasingly looking at power delivery, cooling, networking and compute as interconnected engineering problems rather than independent layers.
The cooling component of Infinitus is therefore significant. ABB’s portfolio includes DC-connected variable-speed drives and motors intended to optimize cooling energy use across data-center infrastructure. As rack power rises, cooling becomes another major electrical load that can influence the overall efficiency of a facility.
ABB expects the first DC-native facilities using the complete Infinitus portfolio to become operational within two to three years, with several major components scheduled for production over the next 12 months.
The transition will not necessarily mean the end of AC distribution. Hybrid architectures are likely to remain important because existing facilities, grid infrastructure and electrical equipment are predominantly built around AC. The more immediate change could be the introduction of DC closer to high-density compute, where reducing conversion stages can have greater impact.
For AI infrastructure developers, that makes 800V DC more than a power-electronics specification. It is becoming part of the architectural discussion around how future gigawatt-scale computing facilities can deliver enough electricity without allowing conversion losses, heat and electrical equipment footprint to consume an increasing share of the available capacity.
Market Landscape
AI is turning data-center electricity infrastructure into a strategic component of compute scaling. The IEA projects global data-center electricity consumption to reach around 950TWh by 2030, roughly double the 2025 level, while AI-focused data-center consumption is expected to triple.
At the same time, the IEA says AI is pushing server power density toward levels that will test the supply chain for transformers, power electronics and other electrical technologies.
This is creating several competing approaches, including higher-efficiency AC systems, 800V DC distribution, onsite generation, advanced UPS technologies, battery storage and direct integration between power systems and AI hardware. ABB’s Infinitus addresses the DC portion of that broader infrastructure transition.
Top Insights
- ABB Infinitus combines DC power technologies from medium-voltage input through rack distribution for next-generation AI data centers.
- The portfolio targets 800V DC architectures as AI rack power moves toward 1MW and beyond.
- ABB says DC distribution can reduce conversion losses, shrink infrastructure footprint and make more electrical capacity available for compute.
- Infinitus supports DC-native and hybrid AC/DC facilities rather than requiring a single data-center power architecture.
- ABB expects the first complete Infinitus DC-native facilities to become operational within two to three years.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
