ABB has deepened its collaboration with Applied Digital, a U.S.-based developer of high-performance data centers, to supply the power infrastructure for the company’s second large-scale AI Factory campus in North Dakota. The expanded partnership underscores the rapidly accelerating demand for electrical systems capable of supporting the extreme power densities associated with artificial intelligence (AI) and GPU-intensive workloads.
The new order – booked in Q4 2025 – covers medium-voltage electrical architecture for Applied Digital’s Polaris Forge 2 development near Harwood, North Dakota. The 300-megawatt campus is being built in phases across two 150-MW buildings, scheduled to enter operation in 2026 and 2027, with additional expansion capacity planned.
As the industry shifts toward AI factories – large, power-intensive facilities designed specifically for AI training clusters and high-performance compute – operators increasingly require more efficient, scalable, and resilient power systems.
ABB’s technologies will serve as the electrical backbone of the Polaris Forge 2 site, enabling high energy efficiency and supporting what Applied Digital projects to be an industry-leading low PUE (Power Usage Effectiveness).
Building the power foundation for AI-intensive workloads
Applied Digital says the new campus represents the next evolution of its AI Factory model, which aims to combine high power density with optimized energy efficiency and operational flexibility.
“Our partnership with ABB reflects Applied Digital’s commitment to redefining what is possible in data center scale and performance,” said Todd Gale, Chief Development Officer at Applied Digital. “Polaris Forge 2 begins with two 150-megawatt buildings that can scale further, reinforcing our leadership in delivering high-performance, energy-efficient AI infrastructure.”
ABB will supply both low- and medium-voltage systems, continuing its focus on raising power density while reducing overall energy waste and deployment complexity. The collaboration builds on ABB’s Smart Power division’s ongoing efforts to modernize electrical infrastructure for AI workloads, which in many cases exceed the power demands of earlier cloud and enterprise data centers by multiples.
“As AI reshapes data centers, ABB is working with leading digital infrastructure innovators to introduce a new generation of power system solutions,” said ABB Smart Power President Massimiliano Cifalitti. “Developing medium-voltage architecture with Applied Digital marks a significant step forward for large-scale AI facilities, enabling higher efficiency, performance, and reliability at lower cost.”
Medium-voltage UPS at hyperscale
Both Polaris Forge 1 and Polaris Forge 2 use ABB’s HiPerGuard medium-voltage UPS technology and advanced medium-voltage switchgear. By shifting from traditional low-voltage to medium-voltage architecture, operators gain greater power density, reduced cabling complexity, improved reliability, and the ability to deploy capacity in larger 25-MW blocks – an increasingly critical feature for hyperscale AI deployments.
Applied Digital recently reported that the first 200 MW of the new Polaris Forge 2 site will be leased to a U.S.-based hyperscaler, reflecting strong market demand for AI-optimized facilities.
Executive Insights FAQ: Power Infrastructure for AI Factories
Why do AI factories require fundamentally different power architectures?
AI training clusters operate at extreme, rapidly scaling power densities – often 2–5× those of cloud data centers – requiring medium-voltage distribution, advanced UPS systems, and higher-efficiency conversion.
What role does medium-voltage UPS technology play?
Medium-voltage UPS reduces energy loss, increases reliability, and enables data centers to scale power delivery in much larger increments, critical for AI clusters exceeding 100 MW.
How does power infrastructure impact PUE?
More efficient electrical distribution and cooling systems reduce overhead power consumption, allowing operators to achieve lower PUE values even at very high rack densities.
Why is scalability so important for AI data centers?
AI training workloads expand rapidly as models grow. Scalable power blocks (e.g., 25 MW units) allow operators to add capacity without re-engineering electrical systems.
How is power reliability maintained at such large megawatt levels?
Technologies like medium-voltage switchgear, redundant UPS systems, and grid-interactive controls ensure continuous operation even during fluctuations or outages.
What is driving hyperscaler demand for AI factories?
The rise of generative AI, LLM training, and GPU-intensive workloads is pushing hyperscalers to secure multi-hundred-megawatt campuses purpose-built for AI.
How do AI factory power systems support sustainability goals?
Higher-efficiency electrical architecture, reduced cabling losses, and optimized cooling lower total energy consumption and help operators integrate renewable energy more effectively.
