On October 14, the U.S. Department of Agriculture announced a $3.3 billion loan to Oglethorpe Power Corporation to help finance a combined-cycle natural gas plant under construction in Monroe County, Georgia. It is the largest loan ever made through USDA Rural Development’s Rural Utilities Service, and the size alone would make it news. What made it a political story came two days later, when the department’s rural development chief told Reuters the plant could potentially benefit data centers in the state.
That single word — potentially — is doing a great deal of work, and it deserves to be unpacked rather than repeated. Oglethorpe, for its part, has been explicit that the facility is not dedicated to serving a data center. Both statements can be true at once, and the space between them is where the data-center load story now collides with federal rural-utility lending. A plant built for cooperative members can still end up feeding the AI buildout, because that is how a shared power system works. The question is not whether the electrons can be traced to a particular server hall. They cannot. The question is who pays for the capacity, who decides what load it serves, and what a federal loan program designed for rural electrification is now being asked to carry.
What the loan actually funds
Start with the project itself, because the financing announcement is easy to misread as a new project announcement. It is not. The Smarr Combined-Cycle facility is a two-unit, 1,425-megawatt natural gas plant in Smarr, in Monroe County, about an hour south of Atlanta. Oglethorpe broke ground on May 7, 2026, after site work that began in September 2025, and expects commercial operation in 2029. The USDA loan helps finance construction already under way; it does not conjure the plant into existence, and the plant will not generate power until the end of the decade.
Oglethorpe is a generation-and-transmission cooperative owned by 38 electric distribution cooperatives that together serve nearly 5 million people across Georgia. That structure matters to everything that follows. Oglethorpe does not sell electricity to households directly. It builds and owns generation, and its member cooperatives buy wholesale power from it to serve their own retail customers — families, farms, businesses, and any large commercial or industrial load that connects within their territories. When Oglethorpe says the Smarr plant will serve 29 of its member cooperatives, that is the relevant detail: the plant’s output is subscribed within the cooperative system, not sold on the open market to a single buyer.
The plant’s scale is substantial by any measure. Oglethorpe puts its output at enough electricity to serve about 800,000 homes, a figure that should be read as an equivalence rather than a count of customers, since a combined-cycle plant running at high utilization serves a mix of residential, commercial, and industrial load around the clock. The facility will use GE Vernova 7HA.03 gas turbines, and Oglethorpe says it will use air-cooled condensers instead of traditional cooling towers, significantly reducing on-site water consumption compared with most combined-cycle plants. That matters in a debate where water has become a flashpoint, even as fuel costs and emissions over a multi-decade life remain open questions.
The financing terms are where the federal role becomes concrete. The Rural Utilities Service Electric Loan Program lends to rural utilities at lower cost than they could typically obtain on their own, and Oglethorpe says the RUS financing will save its member cooperatives more than $25 million a year across the loan’s 30-year term. That savings claim is Oglethorpe’s, and it rests on the difference between federal financing and the market alternative — a difference the cooperative says will flow through to member bills. Whether it fully materializes depends on construction costs, fuel prices, and how the plant is actually dispatched over thirty years, none of which are settled today. But the mechanism is straightforward: cheaper debt for the generation cooperative means lower wholesale power costs for the distribution cooperatives, which is how a loan made in Washington reaches a monthly bill in rural Georgia.
How a plant not built for data centers can still serve them
Oglethorpe spokesperson Blair Romero told Reuters in an email that the facility is not dedicated to serving a data center, and that it will help member cooperatives meet growing electricity demand across their communities. Take that statement seriously, and also take it for exactly what it says. A dedication is a contractual concept: a plant dedicated to a customer has its output committed to that customer. Smarr has no such commitment. Its output goes to the cooperative system, and the cooperative system serves whoever connects and takes service within member territories.
That is precisely why the “may benefit data centers” line cannot be dismissed as a gaffe. USDA Under Secretary for Rural Development Glen Smith, in his October 16 interview with Reuters, said the plant will serve the 38 rural utility cooperatives and support growing industry in the state, including potentially data centers. He also noted that Google purchased nearly 1,000 acres in Monroe County last year for a data center. A land purchase is not a power contract, and it is not a built facility — the desk’s standing distinction between announced, executed, and built applies to land deals as much as to power plants — but it is not nothing either. It is the reason the hypothetical is concrete rather than abstract: the county where the plant is rising is also a county where a hyperscaler has assembled a data-center-scale parcel.
Once a large load connects within a cooperative’s territory, it draws from the same system the Smarr plant feeds. There is no separate grid for data centers and another for households. New capacity added to a system with fast-growing large load relieves pressure that the large load helped create, whether or not anyone intended it that way. Supporters can truthfully say the plant serves everyone on the system. Critics can truthfully say that without the data-center load, the case for building 1,425 megawatts on this timeline would look different.
The siting logic reinforces the point. Oglethorpe chose a site it already owned, near its existing Smarr Energy Facility, with existing transmission access, water resources, and available acreage. Those are the same attributes that make a site attractive to a data center developer: transmission headroom, land, and water. When generation and large load cluster in the same county, the line between “infrastructure for the community” and “infrastructure that enables the data center” blurs, not because anyone misled the public, but because infrastructure is shared.
Who pays, and who decides
The cost-allocation question that runs through every data-center rate case in America runs through this loan too, in a different form. In a conventional utility rate case, the fight is over whether a large new load pays its full cost of service or shifts costs to residential customers. Here, the financing subsidy is federal: taxpayers, through the RUS program, provide lower-cost debt, and the benefit flows to cooperative members in the form of lower wholesale costs. That is the program working as designed — rural electrification has always involved federal credit support — but the scale is new, and the beneficiary question is newly contested.
If Smarr’s output ends up serving a meaningful share of data-center load over its life, then a federal program created to bring affordable power to rural communities will have subsidized, in part, the power costs of some of the world’s most valuable companies. If it does not — if the plant’s output is absorbed by household, farm, and small-business growth, and the data-center load in Monroe County is served under separate arrangements that pay their own way — then the loan is simply what it appears to be: cheap financing for rural infrastructure. The honest answer today is that nobody can say which future arrives, because the load has not yet connected and the plant has not yet run. What can be said is that the cooperative governance structure, not a federal press release, will decide: member cooperatives set the terms under which large loads take service, and those terms — minimum bills, contract lengths, cost allocation — are where the subsidy question will actually be answered.
That is also why the local economics deserve a skeptical look. Oglethorpe says the plant will create at least 30 permanent, highly skilled local jobs and generate county revenue through a 4 percent local sales tax on the natural gas that fuels it. Thirty permanent jobs for a 1,425-megawatt plant is consistent with how modern combined-cycle plants operate — they are capital-intensive, not labor-intensive — and it sits awkwardly beside the employment expectations that often accompany billion-dollar announcements. The tax revenue and construction employment are real, but they should not be confused with a manufacturing-scale jobs story. The plant’s economic case rests on reliable, affordable power for the system, not on its payroll.
There is also ordinary construction risk between now and 2029, and it cuts both ways. A plant expected in service in 2029 is being financed against load forecasts made in 2026; if the large load arrives faster than the plant, the system leans on existing resources in the interim, and if it arrives slower, the cooperative carries new capacity ahead of the demand that justifies it. The loan fixes the financing cost, not the load forecast — the risk cooperative boards live with on every major build.
A loan queue far bigger than the program
The most consequential number in the Reuters interview may not be the $3.3 billion at all. Smith said he plans to meet with congressional appropriators and request more money for such loans, adding, “We have almost $8.6 billion in requests, and we’re only into our fiscal year one month.” The federal fiscal year began October 1, so one month of demand has already produced a pipeline more than two and a half times the size of the record loan just announced.
That queue reframes the Smarr loan from a one-off record to a test case. If RUS lending at this scale becomes routine — multi-billion-dollar generation loans to cooperatives facing data-center-driven load growth — then the program’s politics change with its scale. A program that once financed distribution lines and small generation for sparsely populated territories would become a significant financier of the AI era’s power buildout, with appropriators asked to fund it accordingly. Smith’s planned trip to Capitol Hill is the leading edge of that argument: the administration wants more money because demand is already there.
Whether Congress agrees will depend in part on how the Smarr story plays in the midterm campaign. Reuters reported that Democrats are attempting to capitalize on community discontent over data centers — over water use and utility costs — in November’s midterm elections, while the administration has supported rapid data-center expansion. A record rural loan that a senior official links, even potentially, to data centers hands both sides a talking point. Supporters will call it rural investment that keeps power affordable and attracts jobs. Critics will call it a federal subsidy for the AI buildout dressed as rural development. Both framings will survive contact with the facts, because the facts genuinely support both: it is rural investment, and it may enable data-center growth.
What to watch
The Smarr story will not be resolved by the next press release. It will be resolved, slowly, by a series of observable events that are worth tracking rather than predicting. First, whether a data center in Monroe County moves from land assembly to a power-service agreement with a cooperative, and on what terms — the contract, when and if it appears, will say more about cost allocation than any interview. Second, whether Congress increases RUS electric lending in response to Smith’s request, and whether future loans carry conditions about large-load service. Third, how Oglethorpe’s member cooperatives handle large-load interconnection requests as they arrive: the tariffs and minimum-bill provisions they adopt will determine whether Smarr’s low-cost financing benefits households or is absorbed by the fastest-growing load on the system.
There is also a narrower lesson for how to read announcements in this cycle. A financing announcement is not a construction start — Smarr’s start came months earlier. A land purchase is not a data center. A statement that a plant “could potentially benefit” an industry is not a commitment, and a denial that it is “dedicated” to that industry is not a guarantee that it will never serve it. The data-center load story has always rewarded the discipline of separating those categories, and the arrival of record federal lending makes that discipline more valuable, not less.
The $3.3 billion loan is real, the plant is real, and the load growth it is meant to serve is real. What remains uncertain is the share of that growth that will belong to data centers, and whether the federal government, having now financed the capacity at record scale, will have any say in the answer. That uncertainty is the story — and it will still be the story when Smarr enters service in 2029.