The Scale of the Project
OpenAI's Georgia data center in Effingham County is a roughly 20 billion dollar development that will draw about 3.2 gigawatts of power from Georgia Power under a 25-year contract. Those numbers are worth pausing on. A single campus consuming 3.2 gigawatts is comparable to the demand of a mid-sized city, and it represents the kind of concentrated load that grid operators historically planned decades ahead to accommodate. The project sits within the wider Stargate expansion that OpenAI is pursuing alongside Oracle and SoftBank, a program that has added multiple gigawatt-scale sites across the United States as the company races to secure the compute its roadmap requires. Georgia is one node in a national buildout of unprecedented energy intensity.
For enterprise leaders, the significance lies in the pattern this campus represents. The AI infrastructure race has crossed into a phase where the binding constraint is no longer chips or capital, it is power and the physical capacity to deliver it. Deals like this one are being struck at a velocity that outruns the planning cycles utilities and communities are built around, and the 25-year term reflects how long these commitments now stretch. When a single AI company signs a quarter-century power contract for a load this large, it reshapes the energy landscape of an entire region, and it does so on a timeline set by the company's compute ambitions rather than the grid's comfortable pace.
The Backlash That Followed
The way the Georgia project was approved is the part that should concern anyone planning large infrastructure. According to accounts from the county, the 20 billion dollar development had been negotiated for months but was announced without any public comment or vote, and it surfaced only shortly before nearly 1,000 residents turned out to confront OpenAI and Effingham County officials at a subsequent meeting. That is a striking number for a local infrastructure hearing, and it reflects a community that felt a decision of enormous consequence had been made over its head. The anger was not only about the data center itself, it was about the process that delivered it as a done deal.
This dynamic is becoming a defining feature of hyperscale AI buildouts. Communities are increasingly alert to what these facilities mean for their power bills, their water, and their local environment, and they are organizing faster than developers anticipate. A project negotiated in private and unveiled without meaningful consultation is now almost guaranteed to draw exactly this kind of response. The reputational and political cost of that approach is real, and it can translate into permitting delays, litigation, and rate-case fights that push timelines and budgets well past their original assumptions. The era in which a data center could be quietly sited in a rural county with minimal friction has ended.
Who Pays for the Power
The 25-year Georgia Power contract raises the question that sits underneath every one of these deals: who ultimately pays for the grid expansion that a 3.2 gigawatt load demands. Delivering power at this scale requires new generation, new transmission, and substantial upgrades to existing infrastructure, and those costs land somewhere. If they flow through to the general rate base, ordinary residents and existing businesses end up subsidizing the energy appetite of an AI company, which is precisely the grievance animating the Effingham County turnout. Regulators across several states are now wrestling with how to structure these arrangements so that the data center operator, rather than the public, bears the true cost of the capacity it consumes.
For enterprise technology leaders, this ranks as a direct input to the cost and reliability of the AI services they depend on, well beyond an abstract policy debate. The economics of these mega-deals will shape cloud pricing, the geographic availability of AI compute, and the political durability of the buildout itself. If communities and regulators push back hard enough, some projects will be delayed or restructured, tightening the supply of the capacity that enterprises are counting on. The companies planning their own AI infrastructure, or negotiating large cloud commitments, should factor in that the energy foundation of this boom is contested and that contestation can move timelines and prices in ways a purely technical plan would miss.
Energy Is the New Bottleneck
The Georgia project crystallizes a shift that has been building for two years. The scarce resource in AI infrastructure is now electricity and the physical means to deliver it, having overtaken silicon and funding. That reality is reordering where data centers get built, sending developers toward regions with available generation, favorable regulation, and grids that can absorb enormous new loads. It is also driving the industry toward long-term power contracts, on-site generation, and in some cases dedicated nuclear and renewable projects, because securing energy has become as strategic as securing chips. The companies that lock up power capacity early are buying themselves a durable advantage in the compute race.
For CTOs and CIOs, the practical implication is that AI compute availability will increasingly be shaped by energy geography and grid politics rather than by pure technology decisions. The cost and location of AI capacity depend on where power can actually be delivered and on whether local communities and regulators permit the buildout. That introduces a layer of physical and political risk into AI capacity planning that most enterprise technology strategies have never had to model. Leaders making multi-year bets on AI should understand that the infrastructure underneath those bets rests on an energy foundation that is stretched, contested, and unevenly distributed, and that those constraints will influence what compute is available, at what price, and when.
The Lesson for Infrastructure Planners
The clearest takeaway from Effingham County is that community consent has become a gating factor for large infrastructure, and treating it as an afterthought is a costly mistake. Projects announced as faits accomplis, negotiated in private and unveiled without consultation, invite the kind of organized opposition that can derail timelines and inflate costs. The developers who navigate this environment successfully are the ones who engage communities early, are transparent about power and water impacts, and structure deals so that local residents see genuine benefit rather than only cost. That approach is slower and harder, and it is now the price of building at this scale without a fight.
We read the Georgia episode as an early warning for the entire AI infrastructure boom. The physical and social constraints on hyperscale computing, energy availability, grid capacity, water, and community acceptance, are tightening at the same moment demand is exploding. That collision will define which projects get built, where, and how fast, and it will feed directly into the cost and availability of the AI capacity enterprises rely on. The companies that internalize this now, and plan their buildouts with energy economics and community engagement as first-order concerns, will move faster in the long run than the ones still treating a hostile town-hall crowd as a surprise.
