Workshop examines energy and grid choices for AI data centers

NC State University hosted the Workshop on Energy Needs for AI Data Centers (FREEDM Center) to examine energy demand, grid integration, technology pathways and policy choices for AI-scale data centers.

• Main announcement/action: The workshop presented concrete projections and site examples and discussed on-site generation (“bring your own generation“) as an interim solution while utility capacity expands. Key facts: Duke Energy projections show global AI data center demand rising from 485 TWh (2024) to 945 TWh (2030) (IEA base case, ~3% of global energy by 2030); hyperscale data centers range from 10 MW to 1 GW, with specific examples of Amazon (up to 400 MW, Richmond County, NC) and Microsoft (600 MW, Person County, NC); PowerSecure stated it can ramp up generation + storage behind-the-meter and repurpose assets to the grid after 5-10 years.

  • Date: March 2026
  • Location: NC State University, College of Engineering (FREEDM Center)
  • Agenda/subject: meeting energy demands, grid integration challenges, technology pathways, policy considerations for AI data centers

• Background and details: The discussion contrasted grid-centric plans (Duke Energy’s 2025 IRP weighted to natural gas) with onsite alternatives including engines/turbines, fuel cells, geothermal, thermal energy storage, CHP; legislative context includes Virginia HB323 prioritizing waste heat capture and reuse. Examples cited: Joule Energy data center (Millard County, UT) up to 4 GW and DataOne Vineland, NJ up to 300 MW; references include IEA (Energy and AI) and U.S. DOE guidance on AI.