> ## Content Index > Fetch the complete content index at: https://bytevyte.com/llms.txt > Use this file to discover other available public pages before exploring further. # ABB's Infinitus DC Power Portfolio Targets Power, the Real Limit on AI Capacity - URL: https://bytevyte.com/abbs-infinitus-dc-power-portfolio-targets-power-the-real-limit-on-ai-capacity/ - Published: 2026-09-21T17:26:08.000Z - Updated: 2026-09-21T17:26:08.000Z - Description: ABB's Infinitus DC power portfolio targets GPU-dense AI data centers, cutting conversion stages and copper as power availability limits new capacity. - Author: Bytevyte Editorial - Tags: ai-beats **ABB** announced the **Infinitus DC power portfolio** on September 21, 2026, according to the company's launch statement. The line of direct current power products targets data centers where racks draw so much power per square metre that delivering high-voltage alternating current to the hardware stops being routine engineering and starts limiting how much compute a site can hold. According to ABB, the portfolio is a way to manage rising AI data center power demand and to accelerate DC architectures across the industry. The product list matters less than the framing around it. According to ABB, power availability rather than compute supply increasingly sets the pace of new AI capacity. The Infinitus DC power portfolio is the commercial expression of that argument, and the argument is the more consequential of the two. For anyone planning capacity rather than shopping for accelerators, the reframing is the part worth acting on. It moves the question from how many GPUs a site can host to how many megawatts the site can deliver to silicon. The technical case for DC rests on arithmetic that data center engineers already know. Power arrives from the grid as high-voltage alternating current, then passes through transformers, switchgear, rectifiers and the power supplies inside each rack. Every conversion stage loses energy as heat, and removing that heat costs more energy again through the cooling plant. Direct current distribution removes stages from that chain. That reduction is the mechanical core of ABB's pitch. The second argument is physical rather than electrical. Higher-voltage direct current carries the same power through less copper and needs less room for the equipment between the utility connection and the server hall. On a campus drawing hundreds of megawatts, floor space is not cosmetic. Square metres given to switchgear are square metres not given to racks. ABB frames the portfolio as a response to rising AI data center power demand. The demand signal to track is whether hyperscale operators, who model every watt of loss and every kilogram of copper before committing capital, adopt the designs in new builds rather than in pilot halls. ## Inside the Infinitus DC Power Portfolio According to ABB, the launch extends its existing data center electrification business into a segment where decisions are made early, often during site design and long before servers are ordered. That timing carries commercial weight. The vendor that shapes the power architecture influences which switchgear, busway, protection and monitoring equipment gets specified for the life of the building, across every phase of expansion rather than just the first. Rack-level power density has risen steadily for years, and server designs have already pushed toward higher-voltage direct current inside the rack itself. Extending DC further up the chain, from the rectifier toward the grid connection, is the logical next step once the rack stops behaving like a conventional IT load. Voltage choice is where this turns practical. Direct current at data center scale has to settle on a voltage class that equipment makers and operators both accept, and that agreement decides how much of the savings ABB describes actually reaches a buyer. Until a specification is common, each operator's design is a bespoke exercise in protection, isolation and fault handling. None of this makes DC distribution free. It changes which components matter and where the engineering risk sits, moving complexity from the rack power supply toward central conversion and protection equipment. That shift favours vendors with deep switchgear and protection portfolios, which is the business ABB already runs. If power availability caps new AI capacity, capital moves with it. The marginal dollar on a constrained campus buys megawatts, copper and cooling capacity before it buys accelerators. The Infinitus DC power portfolio is therefore competing for a budget line that is growing faster than the one for compute hardware. Hyperscale operators have the engineering staff and the capital to run DC pilots. Smaller colocation providers tend to wait for standardized, off-the-shelf products with published performance data before they move. ## Where the Skeptics Have a Point The strongest objection to treating this as a turning point is structural. Data centers worldwide are built around alternating current, and an installed base that large does not change because one supplier ships a portfolio. Retrofitting distribution inside a live facility costs money and downtime, and operators rarely accept either without a hard efficiency number attached. Greenfield campuses are the realistic beachhead, and that limits how quickly the shift can run. A second objection concerns evidence. The numbers a buyer needs to model this decision are voltage classes, efficiency measured at the rack, reference deployments and price. Those details emerge through specification cycles and pilot projects rather than through a launch. For now the strategic case rests on ABB's argument about conversion losses rather than on published results from a named site. The skeptics still lose this argument, and the reason is arithmetic. If power availability caps new AI capacity, every watt saved in conversion and every square metre returned to the server hall compounds across an entire campus. The operators who run DC pilots are chasing schedule. The marginal megawatt is harder to secure than the marginal accelerator, and a design that needs fewer megawatts wins on delivery even when it loses on component price. ## What to Watch Four signals will show whether the launch changes procurement or stays in marketing material: - Whether ABB publishes rack-level efficiency figures and names reference sites instead of leaving buyers to model savings themselves. - Whether DC designs appear in new-build specifications or remain confined to pilot halls. - Whether published campus designs show lower copper content and a smaller electrical footprint per megawatt. - Whether competing electrification vendors respond with their own DC lines, which would confirm the segment rather than any single supplier. For infrastructure and procurement leads, the practical response is to ask power vendors for full-path efficiency, measured from utility connection to rack, next to the floor space and copper each design consumes. Comparing portfolios on headline capability tells a buyer little when the real currency is watts and square metres per rack. ## Why This Matters The AI buildout is usually framed as a compute race, and ABB's launch is one more sign that it has become a power race. Companies that treat electricity as the scarce input will schedule new capacity faster than those still counting chips. For readers making infrastructure bets, the decisive question is who controls the path between the grid and the silicon. ## Related Articles - [The Flex EPC Power Acquisition: A $4.4 Billion Bet on AI's Electricity Chokepoint](https://bytevyte.com/the-flex-epc-power-acquisition-a-4-4-billion-bet-on-ais-electricity-chokepoint/) - [Pure DC Secures $2.7 Billion to Expand AI Infrastructure Across EMEA](https://bytevyte.com/pure-dc-secures-2-7-billion-to-expand-ai-infrastructure-across-emea/) - [Nvidia Puts Tokens per Megawatt at the Centre of the AI Factory Pitch](https://bytevyte.com/nvidia-puts-tokens-per-megawatt-at-the-centre-of-the-ai-factory-pitch/) ✔Human Verified --- *Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.*