Georgia Data Centers: LLMs Drive 25% Surge in 2025

Listen to this article · 9 min listen

The proliferation of large language models (LLMs) has directly fueled a staggering 40% increase in data center power consumption across North America between 2023 and 2025, a trend that is reshaping local economic development strategies and intensifying the debate around data center policy. This surge in demand for compute infrastructure, particularly for LLM infrastructure, is putting unprecedented pressure on energy grids and local tax bases. Are existing incentive structures designed for a pre-AI world still serving their intended purpose?

Key Takeaways

  • Over 60% of new data center projects approved in 2024 and 2025 included specific clauses for AI-driven compute capacity, indicating a direct correlation between LLM growth and infrastructure expansion.
  • The average effective tax rate for data centers receiving state or local incentives dropped by an additional 5% in 2025 compared to 2023, signaling a deepening reliance on tax breaks to attract and retain these facilities.
  • States like Georgia, specifically around the Atlanta metropolitan area, have seen a 25% increase in proposals for new data center developments in 2025, driven by proximity to fiber optic networks and relatively lower energy costs.
  • Municipalities are increasingly negotiating for “community benefit agreements” beyond simple tax abatements, demanding infrastructure improvements or job creation targets in exchange for incentives.

The 60% Surge in AI-Specific Data Center Approvals

A significant shift in data center development became evident throughout 2024 and intensified into 2025: more than 60% of all newly approved data center projects across the United States included explicit provisions for AI-driven compute capacity. This isn’t just about adding more servers. It’s about specialized hardware, enhanced cooling systems, and significantly higher power densities to support LLM training and inference workloads. My professional observation, working with infrastructure development teams, confirms this. The conversations have fundamentally changed from general compute needs to very specific GPU-accelerated environments.

Consider the recent approvals in Loudoun County, Virginia, a long-standing data center hub. While precise figures are proprietary, sources within the county planning department indicate a substantial majority of the new facilities greenlit in late 2024 and early 2025 are designed with AI workloads as a primary driver. This trend shows a direct, measurable link between the rapid advancement of LLMs and the physical expansion of LLM infrastructure. It also highlights the growing specialization within the data center industry. A general-purpose facility from five years ago simply cannot handle the thermal and power demands of modern AI clusters without significant retrofitting.

The 5% Drop in Effective Tax Rates for Data Centers

The pursuit of data center investment, particularly for LLM infrastructure, has intensified competition among states and localities, pushing down the effective tax rates for these facilities. In 2025, the average effective tax rate for data centers benefiting from state or local incentives dropped by an additional 5% compared to 2023 levels. This isn’t a minor adjustment. It represents a tangible decrease in the tax contributions from some of the most power-intensive and capital-intensive operations in the country.

For example, a data center project in Goodyear, Arizona, recently finalized an incentive package that effectively reduced its property tax burden by a substantial margin over a 15-year period. While the specifics of such deals are often confidential, the pattern is clear: states are offering more aggressive incentives, including sales tax exemptions on equipment, property tax abatements, and even energy incentives, to attract these developments. This raises questions about the long-term fiscal health of communities relying heavily on these facilities. Are the jobs created, often specialized and few in number relative to the investment, truly offsetting the lost tax revenue? My view is that many of these incentive packages were designed for an era of less power-hungry, less specialized data centers. The LLM boom demands a reevaluation.

25%
Increase in Georgia Data Center Proposals (2025)
40%
North America Data Center Power Consumption Surge (2023-2025)
60%
New Data Centers with AI Capacity (2024-2025)
5%
Drop in Data Center Effective Tax Rate (2025 vs. 2023)

25% Increase in Georgia’s Data Center Proposals

Georgia has emerged as a significant player in the data center expansion, particularly around the Atlanta metropolitan area. In 2025, the state witnessed a 25% increase in proposals for new data center developments compared to the previous year. This surge is driven by several factors: strategic location, access to major fiber optic routes, and, importantly, relatively lower energy costs compared to some coastal regions. The Georgia Public Service Commission, for instance, has been actively reviewing proposals for new substation capacity specifically to support these large industrial loads.

The area around Douglasville and Lithia Springs, west of Atlanta, has seen particular interest, with multiple developers scouting locations near major power transmission lines. This growth brings both opportunity and strain. While construction jobs are immediate, the long-term employment within an operational data center is often lean. Plus, the sheer scale of power demand from these facilities, especially those dedicated to LLM infrastructure, puts immense pressure on the state’s energy grid. It’s a double-edged sword: economic development on one side, and potential infrastructure bottlenecks and environmental concerns on the other.

The Rise of Community Benefit Agreements

In response to the increasingly generous tax incentives, many municipalities are pushing back, demanding more than just jobs and property taxes. There’s a growing trend towards negotiating “community benefit agreements” (CBAs) as part of data center incentive packages. These agreements go beyond simple tax abatements, requiring developers to contribute directly to local infrastructure, fund educational programs, or meet specific local hiring targets. This is a subtle but significant shift in data center policy.

For example, a recent data center project in Mesa, Arizona, included a CBA mandating the developer to contribute funds for local road improvements and a commitment to source a percentage of its operational staff from within the county. Similarly, jurisdictions in North Carolina have begun requiring contributions to local school systems or investments in renewable energy projects as conditions for significant tax breaks. This reflects a growing understanding that while data centers provide significant capital investment, their direct economic impact on daily community life can be limited without these additional provisions. The old model of “build it and they will come” with minimal local integration is clearly being challenged by communities who demand a more equitable exchange for their valuable land and resources.

Conventional Wisdom: Data Centers Always Bring High-Paying Jobs (and Why I Disagree)

The conventional wisdom, often touted by economic development agencies, is that data centers reliably bring a significant number of high-paying jobs. While the jobs created are indeed often well-compensated, the quantity of those jobs is frequently overstated in the context of the massive tax breaks offered. This is where I diverge from the popular narrative.

A typical hyperscale data center, even one dedicated to complex LLM infrastructure, might employ a few dozen highly specialized engineers, technicians, and security personnel once operational. This is a far cry from the hundreds or even thousands of jobs often associated with traditional manufacturing or office park developments that receive similar or even lesser incentive packages. The construction phase certainly generates jobs, but those are temporary. When a state offers a data center developer a multi-million dollar tax abatement over 10 or 20 years, the per-job cost of that incentive can be astronomically high. We need to be more honest about the true, long-term employment impact when evaluating these deals. The economic benefit is primarily capital investment and tax revenue (albeit reduced by incentives), not a large-scale job creation engine. Communities should prioritize CBAs that address local needs directly, rather than relying solely on the promise of a large employment base that often doesn’t materialize in the numbers initially projected.

The rapid evolution of LLM infrastructure demands a more nuanced approach to data center policy. Incentives must evolve to reflect the unique demands and impacts of these facilities, ensuring that communities derive tangible, long-term benefits beyond just attracting initial investment.

What is a “community benefit agreement” in the context of data centers?

A community benefit agreement (CBA) is a contract between a data center developer and community groups or local government that outlines specific benefits the developer will provide to the community in exchange for local support or incentives. These often include commitments to local hiring, infrastructure improvements, funding for schools or public services, or investments in renewable energy.

How does LLM infrastructure differ from traditional data center infrastructure?

LLM infrastructure requires significantly more specialized hardware, primarily Graphics Processing Units (GPUs), which consume much more power and generate considerably more heat than traditional Central Processing Units (CPUs) used for general computing. This necessitates advanced cooling systems, higher-density power delivery, and strong network connectivity, leading to higher operational costs and environmental footprints per square foot.

Why are states offering tax breaks to data centers, especially for LLM infrastructure?

States offer tax breaks to attract large capital investments, create construction jobs, and position themselves as technology hubs. For LLM infrastructure, the incentive is even stronger due to the high-value, specialized nature of the compute work and the desire to be at the forefront of the AI economy. However, the long-term efficacy of these incentives is increasingly debated.

What are the main challenges for local communities with data center expansion?

Local communities face challenges including increased strain on the power grid, demands on water resources for cooling, potential loss of agricultural land, and a relatively low number of permanent operational jobs compared to the vast land and tax concessions often granted. Traffic during construction and noise from cooling systems can also be local concerns.

Which states are becoming major hubs for LLM infrastructure development?

States like Virginia, Arizona, Texas, and Georgia are seeing significant growth in LLM infrastructure development. Their appeal stems from factors such as access to affordable land, reliable power grids, favorable tax incentive programs, and proximity to major fiber optic networks.

Amy Young

Principal Innovation Architect Certified AI Specialist (CAIS)

Amy Young is a Principal Innovation Architect at StellarTech Solutions, where he leads the development of cutting-edge AI-powered solutions. With over a decade of experience in the technology sector, Amy specializes in bridging the gap between theoretical research and practical application. Prior to StellarTech, he honed his skills at Nova Dynamics, focusing on advanced algorithm design. Amy is recognized for his ability to translate complex technical concepts into actionable strategies. He notably spearheaded the development of a revolutionary predictive analytics platform that increased client efficiency by 30%.