Gas-Fired Power Plants for Data Centers Will Raise Bills for Georgians

Executive Summary

Georgia Power Company has pledged that "Residential customers are not subsidizing data center growth" (their emphasis). In December 2025, before Commissioner Hubbard took office, the all-Republican Georgia Public Service Commission approved Georgia Power's request to add approximately 10,000 MW of capacity from new gas-fired power plants and battery storage, primarily to serve data centers. Of this massive expansion, 3,889 megawatts (MW) (winter) of approved capacity are from combined-cycle gas turbines (CCGT) that burn fossil gas fuel.

In February 2026, Commissioner Hubbard moved to decertify these gas units because they were not justified by contracted load. More importantly, CCGTs require both new pipeline capacity and fuel and that makes them more expensive to operate over the life of the plant than solar, battery storage, and other options available today. These CCGT units will directly increase the cost to produce electricity in Georgia and add carbon emissions that also have a significant cost. Commissioner Hubbard's motion failed when Republicans voted to keep the expensive fossil gas (see dissenting opinion in Document #225550).

To demonstrate the additional costs to consumers, this analysis used ReEDS, the flagship capacity planning model for the power sector developed by the National Lab of the Rockies, to compare two portfolios for the years 2030 to 2050. One portfolio forces into the model approximately the same CCGT capacity that was rubber-stamped by the All-Republican PSC last December. The other portfolio is a cost-optimized solution that selects the lowest-cost resources (solar, battery storage, and gas peaking units) while still meeting full reliability requirements. Note: The gas peaking units are fueled with fossil gas and could be replaced with more solar and battery storage, albeit at higher cost. The very fact of the inclusion of gas peaking units in the cost-optimized portfolio demonstrates the honesty of this analysis. Compare this honesty to Georgia Power, who redacts their analysis.

The portfolio with the forced CCGT capacity costs at least $8 billion more from 2030 to 2050, based on a regional view that assumes any surplus power is efficiently sold to neighboring utilities. However, the costs for these new power plants are incurred in Georgia. The local increase in electricity production costs could rise as high as $29 billion over and above the cost-optimized portfolio for the years 2030 to 2050.

Now compare this cost premium of at least $8 billion for uneconomic CCGT capacity to Georgia Power's promise of $15 per month in downward pressure on average residential bills for three years (2029 to 2031). For 2.4 million residential customers, that totals only $1.35 billion and that pledge is only as good as the regulators who hold Georgia Power to that pledge. In Net Present Value terms , the added cost of the CCGTs is 3.3 times the pledged residential customer benefit. In other words, the company’s promise of future benefit is much smaller than the costs being incurred to build new CCGT power plants for data centers in Georgia.

The model assumptions (available at the GitHub fork, link below), input assumptions, and output results, including interactive charts and a spreadsheet with the output numbers, are published here. Georgia Power does not publicly publish its modeling because it cannot justify its analysis. 

The Bottom Line: Building fossil gas CCGT capacity for data centers—instead of clean power—will increase the cost of electricity for Georgians. Georgia Power's pledge of three years of relief against decades of future costs is much less than the costs being incurred today for new CCGT generation (and transmission) for data centers.

Key Results

Higher System Cost: Forcing CCGT capacity into Georgia raises the cost of serving system load across the region by $8 billion in the years 2030 to 2050 (nominal dollars). On a NPV basis, this additional cost is $4.1 billion, more than three times the present value of the pledged benefit for 2.4 million residential customers ($1.25 billion) from three years of $15 downward pressure. Neighboring utilities can buy some of the resulting surplus energy, which offsets part of the higher system cost. For context, building and running the CCGTs in Georgia costs $29 billion over 2030 to 2050. PSC staff estimated that the full package of nearly 10,000 MW could cost customers $50 to $60 billion over the life of the assets.

More CO2 Emissions: Forcing CCGT capacity into Georgia raises regional carbon dioxide emissions by an average of 2 million metric tons each year from 2030 to 2050, or 42 million metric tons in total. Emissions increase more in Georgia, by about 4 million metric tons per year on average, but are partly offset as displaced generation shifts across state lines. In 2040, the forced CCGT portfolio emits 5.8 million metric tons more carbon dioxide in Georgia than the cost-optimized portfolio. If a modest $30 per ton carbon pricing is included, the price tag would rise by more than $1 billion over the same time period.

Cost-Optimized Capacity = More Solar and Storage: This analysis forced 3,625 MW of new CCGT capacity into Georgia by 2031, where it remains for the study period. The only change imposed on the model was that requirement. Without it, the model's own cost-optimized portfolio builds, on average, 1,850 MW more solar power and 800 MW more battery storage, along with 2,100 MW more simple-cycle combustion turbines (CTs). The CTs are still gas-fired but operate much less often than CCGT units. More importantly, the CTs could be replaced with more battery storage and solar, albeit at higher cost.

Results are Robust to Reliability: ReEDS is coupled with the Probabilistic Resource Adequacy Suite (PRAS) to ensure system reliability, measured as Expected Unserved Energy (EUE) under one part per million of annual load. In both scenarios, the worst year reached 0.15 parts per million EUE, one-seventh of the allowable level. Stressful days with high load and low solar output are identified by PRAS and dispatched by ReEDS in each modeled year, and modeled stress days were increased beyond the SERTP default to ensure robust results.

Conclusion

New gas-fired power plants for data centers will increase the cost of electricity in Georgia compared with lower-cost solar power, battery storage and other options. Georgia Power's pledge to put $15 of downward pressure on monthly bills for just three years is much smaller than the additional cost of building combined-cycle gas plants by at least a factor of three.

Assumptions and Benchmarks

This analysis assumes the overnight capital cost of a new combined-cycle gas turbine in Georgia is between $1,700 per kW (EPC contractor-build) and $2,400 per kW (self-build). New CCGT units also incur a fixed O&M cost of $2,000 per MW-week, including pipeline costs, and a variable O&M cost of $2 per MWh, plus delivered fuel costs that begin at $4 per MMBtu in 2030 and rise to $9 by 2050 (nominal). The CCGT heat rate is 6,700 Btu per kWh. Lower gas prices would narrow the cost gap and higher prices would widen it. The $8 billion figure also prices carbon emissions at zero. As a benchmark, the northern Indiana utility NIPSCO uses $3,500 per kW for the build cost of a CCGT in its 2026 resource plan public advisory meeting materials (October 2026, page 58). Accordingly, this analysis is conservative, and costs could be higher.

Model Description

Modeling was performed with the Regional Energy Deployment System (ReEDS), the open-source capacity expansion model developed by the U.S. Department of Energy National Laboratory of the Rockies. This analysis used the Georgia CES fork, branch GA_IRP. Results reflect two scenarios defined in cases_georgia.csv: SERTP_gpc, the cost-optimized counterfactual, and SERTP_gpc_force, which forces 3,625 MW of new combined-cycle capacity into Georgia. The two scenarios are identical in all respects except the forced-capacity CCGT requirement.

The analysis used a modified version of ReEDS branched from the public repository (ReEDS-Model/ReEDS) at commit 62f6381, dated June 23, 2026. Four sets of modifications were made. First, a new constraint (eq_newcapforce) was added to require a minimum level of cumulative new capacity investment by technology, state and year; this is the mechanism used to force Georgia Power's planned combined-cycle additions into the model. When the associated switch is disabled, the constraint is inactive and model behavior is identical to public ReEDS. Second, current estimates for capital cost, fixed O&M and heat rate for new combined-cycle capacity in Georgia replaced the default NREL ATB 2024 values, phased in from 2026 to 2029. Third, a one-line defect in the public code's representative-period clustering was corrected; it prevented clustering at the balancing-area level in subset-region runs. Fourth, reporting templates were added to produce Georgia, Southern Company and SERTP results. 

No existing model equation, cost assumption or default switch value was otherwise altered: all changes to the model's GAMS source are additions (61 lines added, none removed), and the clustering fix is the only line of existing model code modified. The model reflects the 2025 federal law that ended solar and wind tax credits early. It does not add costs for federal restrictions on foreign-sourced batteries.

To download a PDF of this white paper, click here.

To download the modeling outputs in Excel format and a graphical .html format, click here.