4-model average
building category history
loading
7pt spread
The three supporting readings tell you how much weight to put on the probability: confidence reflects category-level track record, stability tracks how the estimate has moved over time, models shows whether the four agree.
Loading belief trajectory...
As of 2023, many utilities are facing increasing pressure to balance power supply with demand, with a notable trend towards renewable energy integration. For instance, the U.S. Energy Information Administration (EIA) projects a substantial rise in electricity demand driven by data centers, estimated to increase 4% annually by 2026. Recent statements from utilities, such as Pacific Gas and Electric, indicate concerns over capacity limitations in the face of ongoing infrastructural upgrades and regulatory challenges, suggesting that at least one major utility might deny a new AI data center power contract due to insufficient capacity.
The rapid expansion of AI data centers is placing unprecedented strain on electricity grids, with numerous reports detailing utilities struggling to meet projected demand in some regions. For example, the PJM Interconnection, which serves 13 states and the District of Columbia, has over 260 GW of generation and load requests pending, far exceeding its available capacity. This trend, coupled with the long lead times for grid upgrades and the increasing power draw of AI workloads, makes it highly probable that some utilities will be forced to decline new contracts by 2026.
AI data center power demand is projected to grow 160-300% through 2026 according to major utility filings (NextEra, Duke Energy, Southern Company), while U.S. grid capacity additions are tracking 15-25% annual growth—a structural mismatch. Regional transmission constraints are already causing renewable curtailment in high-demand areas (California, Texas), and at least 3 major utilities (PG&E, ERCOT participants, Duke Energy Carolinas) have publicly acknowledged capacity limitations in forward planning documents. The combination of explosive AI load growth, multi-year transmission build timelines (5-7 years), and utilities' stated capacity concerns creates high probability that at least one major utility formally declines or significantly delays an AI data center contract specifically citing capacity constraints by 2026.
Duke Energy’s 2025 IRP shows a 9 GW shortfall by 2030 against 35 GW of announced data-center load, and Dominion’s 2024 IRP already deferred 2.3 GW of new load because transmission cannot be built before 2028; with FERC Order 2023 queue delays averaging 40 months and transformer lead-times now 3-4 years, at least one utility will likely reject or defer a signed PPA on capacity grounds in 2026.