Experts say utilities’ projected data center growth is speculative
Utilities in Alabama, Georgia, North Carolina, and South Carolina are planning for data-center-driven electricity demand growth that has a roughly 0.2% chance of actually occurring. Modeling from the report, “Impacts of Projected Data Center Growth and Emerging Uncertainties on Power Demand in the Southeast,” raises serious questions about the assumptions used to justify the methane gas buildout.
In this interview, Megan Gibson, Senior Attorney at the Southern Environmental Law Center, chats with Etan Gumerman, Director of Analytics at Greenlink Analytics and Amy Sharma, Executive Director of Science for Georgia on the data center scramble, utility over forecasting and the possible economic consequences for ratepayers in the Southeast.
Currently, the Southeast is a growing data center market. The states discussed in the report —Alabama, Georgia, North Carolina, and South Carolina — are facing an expansion of methane gas (often referred to as ‘natural gas’) infrastructure in large part to provide energy for new data centers. How does this impact energy resource planning across the Southeast?
Etan Gumerman: Utilities across the country have identified the opportunity in data centers for renewed load growth after a few decades of slow growth. While data center development has been growing rapidly, discerning the implications for localized future load growth is not straight forward.
The big questions are not whether developers want to build data centers rapidly, or whether utilities want to build gas infrastructure in response; they do. Rather, how many data centers will actually be built in a particular place, and how long will they stay in that place? And how much new power and related infrastructure is actually needed to meet their needs?
The load growth projections that are being used to justify expanding fossil infrastructure may be very speculative. Therefore, in this study we wanted to methodologically dig into the range of uncertainty for data center load growth.
Amy Sharma: Data center demand projections are fueling the push to expand methane gas infrastructure.
In Georgia, in Dec 2025, the Public Service Commission approved another 10GW of power generation, a good portion of which is for data centers. Most of this power generation capacity will be from gas turbine generation.
What is your biggest finding from the report?
Amy Sharma: That the Georgia, Alabama, North Carolina, and South Carolina utilities predicted asset needs of approximately 10GW only have a 1 in 500 chance in happening. Instead, our independent analysis found that the most likely expanded need will be between 3.5 and 5.5 GW. Thus, any ‘needed’ generation capacity is likely to be half of which has been requested.
For example, Georgia Power originally estimated a 93% likelihood that all data centers in its pipeline in technical review will eventually sign a contract for service in Georgia, and more than half of the projects in Georgia Power’s 2023 IRP model left the queue and will not materialize. Still, Georgia Public Service Commission (PSC) unanimously voted to approve Georgia Power’s request to acquire nearly 10 gigawatts (GW) of new energy capacity to meet in large part the generation needs of data centers. Most of this demand is expected to be filled by gas turbines.
It is worth noting that Southern Company is the parent company of Georgia Power and some of the relevant methane gas pipeline companies, including co-owning the proposed SSE4. Therefore, they have approval to build gas turbines (that they don’t have to pay for – due to guaranteed cost recovery) and then are going to try and sell the gas from this pipeline to power some of their own gas turbines. AND the gas prices will also be passed onto the Georgia ratepayers. They are highly incentivized to overestimate demand. They could make money in several different ways from this.
This decision goes against expert claims and the recent testimony of the PSC’s own staff stating that five Hoover Dams’ worth of energy is unnecessary. The approved plan will cost customers an estimated $50-60 billion over the life of the resources.
Can you explain why there is so much overspeculation in these utility load growth projections?
Etan Gumerman: Utilities have a financial interest in load growth and a history of over predicting load growth, so aggressive utility forecasts should not be surprising. There are many reasons that this new, unprecedented aggressive projected load growth may not materialize. I will mention a few.
First, developers consider building data centers in many locations, and the cost to shop around is low enough that there is double or triple counting for some anticipated builds.
Second, hardware and software efficiency improvements occur frequently in this field and may accelerate beyond conventional wisdom. Third, data center load growth could primarily be offset by load flexibility.
Why should this matter to the average electricity customer in the Southeast?
Etan Gumerman: If utilities are allowed to build unnecessary speculative load, somebody will have to pay for it. These stranded assets will not be paid for by data centers that were never built. Obviously, the more potential stranded assets, the more cost risk will be shifted to existing customers.
If the perceived data center demand never appears – if we build it and they do not come – the average electricity customer in the Southeast will be left picking up the tab.
Amy Sharma, Science for Georgia
Amy Sharma: I can’t speak for other utilities – but in Georgia – Georgia Power is guaranteed cost recovery and profit for asset creation. This means once a power plant comes online Georgia Power will shift the cost of building that plant, plus a profit margin, onto power bills.
Thus – they are actually incentivized to overbuild – there are no negative consequences for overestimating.
If the perceived data center demand never appears – if we build it and they do not come – the average electricity customer in the Southeast will be left picking up the tab.
What other themes emerged while conducting this research?
Amy Sharma: We also looked into what happens if technology improves – and if there is investment in improved technology: demand scheduling, better cooling, improved algorithms, new chip designs, etc. – data center power demand will remain almost flat for the next ten years.
Maybe instead of investing in an army of lobbyists – these data center companies could invest in better technology!
Etan Gumerman: Long-term, worldwide growth of data centers is speculative. We should take forecasts of this recently booming industry with a grain of salt. I remember when in the 1990s some extrapolations were predicting that internet and IT would consume 50% of electricity in short order.
There are plenty of examples of communities not supporting building any or additional data centers, which is an unresolved issue, even if the developers and utilities are enthusiastic.
Given utilities track record of over forecasting, and this new additional evidence, how should policymakers, regulators, and communities treat aggressive projections around data center demand?
Amy Sharma: They should be skeptical, curious (ask questions!), and take these with a grain of salt. The only people who think the data center train will not end are the power companies, everyone else thinks this upward trajectory will eventually level off.
Instead – they should seek to put the cost of speculative building on large load customers. Rules that prevent costs from shifting onto residential rate payers are a good start.
What issues still need more attention in order to create a system that works for everyone? How should utilities address those issues?
Etan Gumerman: Utilities could consider rules to reduce the likelihood of multiple counting or another approach to guaranteeing that large customers or the utility will accept all the risk of overly aggressive capacity building. Utilities can invest in alternatives like demand side management programs, and efficiency that would offset the need for peak load growth and save lots of money. Utilities can invest in alternatives like demand side management programs, and other efficiencies that would help offset the need for yet more power generation capacity (e.g., more gas power plants) and save costs.
I understand ERCOT has begun to adopt rules moving in that direction and allowing for curtailment of large loads.
Amy Sharma: Step 1 – Rules to prevent speculative overbuilding and shielding residential rate payers from costs.
Step 2 – Incentives to invest in power and water saving technologies. Building up the science and technology innovation ecosystem so that we can have AI without hyperscale data centers.
Step 3 – Investment in sustainable power systems. It is now faster and cheaper to meet demand with solar and wind – not with gas or coal. We should not be powering the future with dirty technology that is from the past.
“Impacts of Projected Data Center Growth and Emerging Uncertainties on Power Demand in the Southeast” was published by Greenlink Analytics and Science for Georgia, and commissioned by the Southern Environmental Law Center.
Amy Sharma, PhD is the Executive Director of Science for Georgia, a non-profit dedicated to getting science out of the lab and into service for society. Amy has worked in public, private, and academic spheres. Early in her career, she worked in chip design for IBM. She holds a BSE in Biomedical and Electrical Engineering and a PhD in Biomedical Engineering, both from Duke University.
Etan is Greenlink Analytics’ Director of Analytics, previously working at Duke University and Berkeley National Laboratory. He has 30 years of experience in energy policy, modeling, and systems analysis. At Greenlink, Etan leads a team evaluating clean energy transition policies and opportunities, including forecasting the associated energy, emissions, jobs, and public health impacts.