Applied Digital’s Data Center Design for a 100 MW AI Factory Built from the Ground Up
AI workloads are redefining the limits of data center design and infrastructure. Legacy data centers, built for traditional co-location, cannot handle the density, thermal demands, or power dynamics of accelerated computing. The AI boom has upended the data center sector, forcing a rapid shift to liquid-cooled racks as facilities pivot from sub-10kW racks to 120kW racks. According to NVIDIA’s 2025 GTC roadmap, the company is now targeting 600kW racks by 2027—a dramatic escalation that demands a complete rethinking of infrastructure design. In this climate, starting fresh isn’t a luxury—it’s a requirement.
So, how do you design an AI-first facility that can scale from 100 to 400 megawatts and stay relevant as the pace of innovation accelerates?
Welcome to Episode 1 of Architects of Acceleration—a five-part series from Applied Digital that takes you behind the scenes of one of the most ambitious AI infrastructure projects in the U.S. In this premiere episode, host Philbert Shih, founder of Structure Research, sits down with Todd Gale, the Chief Development Officer, and Brian Hasemann, the Design Services Manager at Applied Digital, to unpack the full design strategy behind Polaris Forge 1, a greenfield AI campus in Ellendale, North Dakota. Together, they walk through the blueprint, site selection, power procurement, and thermal architecture that shaped Applied Digital’s first 100 MW AI Factory.
What You’ll Learn in This Episode:
-
Why Applied Built from Scratch: A greenfield approach allowed the team to build around 1,000–1,500 watts/sq. ft. workloads and eliminate retrofitting compromises.
-
Why Rural North Dakota: Ellendale offers scalable land, renewable wind energy, free cooling climate conditions, and power pricing advantages ideal for AI workloads.
-
How It Supports AI Workloads: The facility is optimized for liquid cooling and purpose-built to host dense GPU clusters, including evolving rack-level demands.
-
How the Campus Will Scale: The site is designed to grow from 100 MW to 400 MW, with ELN03 and ELN04 incorporating lessons from the first build for even greater flexibility.
Todd Gale is a veteran data center engineer and executive with over four decades of experience designing mission-critical infrastructure at a global scale. As Chief Development Officer at Applied Digital, he leads the strategy, design, and construction of next-generation AI data centers, including the Ellendale Campus, purpose-built for GPU workloads. Previously, he held senior roles at Flexential, Equinix (formerly Terremark), and AT&T, where he pioneered innovations in low Power Usage Effectiveness (PUE) architecture, Tier IV-certified facilities, and hyperscale cooling systems.
Brian Hasemann is Director of Data Center Design at Applied Digital, where he leads the development of mission-critical infrastructure for high-density AI workloads. With a background in mechanical engineering and project management, he has overseen major design efforts and facility assessments across data centers, hospitals, and large-scale commercial properties. Before joining Applied Digital, he held engineering and commissioning roles at EBI Consulting and Shire, bringing hands-on experience in HVAC, utility systems, and CMMS optimization across complex environments.