International Journal of Science, Engineering and TechnologyAuthors: Samuel N Nimaful, Augustine Hanyabui, Faith Esther Holison, Laureta Tatenda Nyamsutswa, Joel Holison
Abstract: U.S. data center growth—driven by cloud services, artificial intelligence (AI), and digitalization—has become a power-system planning problem as much as a real-estate or information-technology problem. Recent federal analysis indicates data centers consumed about 176 terawatt-hours (TWh) in 2023 (about 4.4% of total U.S. electricity use) and could reach roughly 325–580 TWh by 2028 (about 6.7%–12%), depending on broader economic and electricity demand growth (U.S. Department of Energy [DOE], 2024). [1] These levels of load growth are material at a bulk-power-system scale and are arriving faster than most traditional generation, transmission, and distribution (T&D) investment cycles—especially in regions already constrained by transformer lead times, siting and permitting timelines, and limited interconnection and construction capacity. [2] Interconnection bottlenecks—historically framed as a generator problem—now collide with large-load connection timelines and cost responsibility debates. At the transmission level, DOE reports the interconnection queues have expanded from under 500 gigawatts (GW) in 2010 to roughly 2,600 GW “today,” with more than 95% associated with zero-carbon generation and storage; DOE also notes that time-to-interconnect has more than doubled nationally. [3] Independently, Berkeley Lab’s queue synthesis reports that, as of end-of-2024, thousands of projects and thousands of gigawatts remain in queues, and median durations from interconnection request to commercial operation have risen to over four years for recently completed projects in regions with available data. [4] Even when data centers are not themselves “in the queue” (because loads often follow different or less transparent processes than generation), the same constrained equipment, studies, and network upgrade construction resources govern whether new load can be served on feasible timescales. [5] Power infrastructure constraints are now observable in reliability performance. NERC identifies “large loads”—including data centers—among the most significant near-term reliability challenges, citing observed events in which approximately 1,500 MW of data centers disconnected simultaneously and unexpectedly from the bulk electric system (BES) in 2024 after a transmission fault, creating balancing and stability challenges analogous in magnitude to a large nuclear plant changing output unexpectedly. [6] This operational reality re-frames “speed-to-power” strategies: getting connected quickly is not sufficient if poor load observability, protection coordination, and flexibility arrangements raise system risk or shift costs to other customers. [7]
DOI: https://doi.org/10.5281/zenodo.19422330
