Coal Reenters Debate Over AI-era Electricity Needs

A May 22, 2026 op-ed by energy-industry advocate Amy Oliver Cooke argues the U.S. must keep coal plants running to meet AI-driven electricity demand, citing NERC projections of 967,000 megawatts of peak summer demand by 2030 (an 18% increase) and a 1,810% year-over-year surge in gas-plant proposals with turbines backlogged to 2029-2030. The piece says Xcel Energy wants to keep its four Colorado coal units open through 2030 and that Southern Company is tracking over 50 GW of new demand. Independent Colorado Sun reporting complicates that framing: Xcel's own regulators already rejected similar capacity-shortfall arguments once, calling them unpersuasive, and environmental groups say the crisis is overstated since recent extended coal-unit outages did not cause blackouts.
The load-bearing claim in this piece, that keeping coal online is the only way to meet AI-driven power demand, is presented as settled fact but is actively contested by the very regulators it cites. That gap between an energy-industry advocate's framing and the regulatory record matters more for practitioners assessing power-availability risk than the piece's headline numbers alone.
What happened
In a May 22, 2026 RealClearEnergy op-ed (widely syndicated, including to Watts Up With That), Amy Oliver Cooke, founder of Always On Energy Research, argues that AI, data centers, advanced manufacturing, and electrification are driving power demand growth the U.S. grid cannot meet with intermittent generation alone. She cites NERC's 2025 Electricity Supply and Demand data projecting U.S. peak summer demand near 967,000 megawatts by 2030, an 18% increase from today. She reports that gas projects surged 1,810% year-over-year after PJM reformed its interconnection queue, that major turbine manufacturers are backlogged to 2029 or 2030, and that Southern Company is tracking more than 50 GW of potential new customer demand, which led it to reverse planned coal retirements in Georgia and Mississippi. She also cites Xcel Energy's push to keep its Colorado coal plants (Comanche 2, Comanche 3, and two Hayden units) online through 2030, Duke Energy delaying North Carolina coal retirements after the state legislature dropped a 2030 interim carbon target, and Tri-State Generation calling its Craig Unit 1 back into service in April 2026 after a Southwest Power Pool resource advisory.
Industry context
Independent Colorado Sun reporting on the Xcel case adds context the op-ed omits. Xcel's March 2, 2026 filing to the Colorado PUC proposed three options, one of which was running all four coal units to 2030, framed as informational rather than a final request. Critically, the Colorado PUC had already rejected the same underlying arguments (supply chain issues, geopolitical factors, and peak demand growth) when Xcel used them to justify extending Comanche Unit 2, calling the reasoning unpersuasive and finding that only Comanche Unit 3's unplanned turbine outage justified the extension. Consumer and environmental advocates (350 Colorado, the Western Clean Energy Campaign, and the Colorado Office of Utility Consumer Advocate) argue the capacity shortfall is overstated, noting reserve-margin cushions absorbed a period when all three disputed coal units were simultaneously offline without blackouts, and suggest Xcel's all-coal option may be a negotiating position ahead of further settlement talks rather than an operational necessity.
For practitioners
The underlying data-center demand numbers (NERC's 967,000 MW/18% growth projection, Southern Company's 50 GW pipeline, gas turbine backlogs to 2029-2030) are independently sourced and not in dispute; the contested part is the policy conclusion that coal retirements must be delayed to meet that demand. Teams evaluating power-availability risk for data center siting should treat utility capacity-shortfall filings as one input among several contested by regulators and consumer advocates, not as a neutral technical baseline, especially in jurisdictions with active coal-retirement litigation.
What to watch
Xcel's follow-up regulatory filing (due after its March 2026 report) laying out specific proposed actions, and how the Colorado PUC rules on the coal-extension options for Comanche and Hayden given its prior rejection of similar reasoning.
Key Points
- 1An energy-advocate op-ed argues coal must stay online to meet AI-driven demand, citing NERC's 967,000-megawatt 2030 peak projection.
- 2Colorado's PUC already rejected similar capacity-shortfall arguments from Xcel as unpersuasive when extending a different coal unit.
- 3Practitioners assessing power-availability risk should treat utility shortfall claims as contested, not as a neutral technical baseline.
Scoring Rationale
The underlying AI/data-center demand figures (NERC projections, gas turbine backlogs, Southern Company's pipeline) are independently verified and remain a real practitioner-relevant infrastructure constraint. Slightly down-adjusted from 6.8 after verification showed the piece's central policy claim, that coal retirements must be delayed, is a contested industry-advocate position that Colorado's own regulators have already called unpersuasive, not a neutral consensus finding, which the original framing did not make clear.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems


