Digital Foundations: The Essential Guide to Data Centers and Their Growth

Turner Loesel — Policy Analyst, Center for Technology and Innovation, The James Madison Institute

Josh Smith — Energy Policy Lead, Abundance Institute

Executive Summary
What Are Data Centers?
Background on Data Center Operations
Why Are They Essential to AI and Tech?
Water and Energy Use by Data Centers Is Manageable and Often Exaggerated
Economic Impact of Data Centers
Widespread Interest in Attracting Data Centers Exists
Speed To Power Is the Real Challenge
A Public Utility’s Core Mission Is To Provide Stable and Reliable Power
Enabling Data Centers and Protecting Consumers
Conclusion: Data Centers Are a Digital Foundation for the Modern World

Executive Summary

Data centers are foundational infrastructure for the modern economy. In short, they are the computers we use but don’t touch. They power essential services ranging from cloud computing and online commerce to artificial intelligence and secure financial transactions. As the scale and scope of digital services continue to expand, so too does the need for reliable, efficient data centers capable of supporting global connectivity, computation, and storage.

Today, over 5,300 data centers operate in the U.S., anchoring a fast-growing segment of both digital infrastructure and economic development. These facilities generate thousands of jobs during construction and offer high-wage careers once operational. They also deliver substantial tax revenues, often exceeding their demands on public infrastructure. In regions like Loudoun County, Virginia, data centers have driven broad fiscal benefits while enabling the county to lower residential tax rates.

Growing energy and water usage has led to public concern, yet these concerns often stem from misunderstandings about scale and context. While data centers are large electricity consumers, their relative share remains modest compared to industrial and transportation sectors. Moreover, the industry has achieved remarkable gains in efficiency and sustainability, reducing water usage, enhancing cooling systems, and investing heavily in renewable energy sources.

In this way, the last 30 years are a better guide to understanding the likely electricity demands of data centers than the last 30 months. The emergence of the internet and personal computing was replete with exaggerated energy estimates. That reality was moderate, even small, relative to these exaggerations is a useful caution against excesses in this new era of increasing demand. There is little doubt that the rise of artificial intelligence will increase demand for electricity and water. But this has been represented as unmanageable rather than concentrated on enabling solutions.

Still, a regulatory mismatch persists. Utilities are designed to avoid risky investments, while data center developers prioritize speed. This disconnect creates unnecessary delays and risk. States can address this by enabling competitive power procurement for large new loads, insulating residential customers from speculative utility costs while accelerating economic growth.

Ultimately, data centers are not a burden but a critical engine of prosperity. States that modernize their regulatory frameworks will attract high-value investment and secure leadership in the digital economy.

Policy opportunities:

Avoid blanket restrictions or moratoriums on data center development based on exaggerated concerns about energy or water use.
Enable retail electricity choice options for new data centers to bring their own power and to avoid burdening existing ratepayers.
Emphasize speed to power in legislative efforts to enable data center growth.
Streamline permitting for both data center construction and associated power infrastructure.

What Are Data Centers?

Though only recently entering mainstream consciousness, the modern data center traces its lineage to the dawn of the computing age. In 1946, the University of Pennsylvania unveiled ENIAC (Electronic Numerical Integrator and Computer), a groundbreaking machine that laid the foundation for today’s data centers. Originally commissioned by the U.S. Army to calculate artillery firing tables during World War II, ENIAC occupied a massive 1,800 square feet of space and required dedicated facilities for power, cooling, and maintenance—infrastructure elements that remain essential in modern data centers. From these military roots, data centers have evolved into technological cornerstones of our digital age.

Modern data centers bear little resemblance to early computers like ENIAC. These specialized facilities house vast arrays of servers, storage systems, and networking equipment that power our digital economy. They enable businesses to host websites, run e-commerce platforms, and manage critical internal operations from accounting to human resources. Both private companies and public institutions rely on data centers to store and process the massive amounts of information that power their daily operations.¹ Without data centers, much of our modern digital economy would not be possible.

Data centers are the computers we use but never touch. The real computing power foundational to much of our electronic world isn’t actually in our hands when we use a phone application or type away at a laptop—it’s in a data center. While smartphones and laptops serve as interfaces, the vast majority of processing, storage, and computation for internet services now takes place on remote servers.

The design and scale of a data center can vary significantly to meet diverse operational demands, ranging from small local facilities to massive warehouse-sized installations. Each type of data center serves a distinct purpose in our digital world.

Enterprise data centers serve as private, in-house computing hubs for individual organizations, containing “an organization’s routers, switches, servers, firewalls, and other components that enable critical applications and store and process data.”²

Enterprise data centers are typically built by an organization for the exclusive use of that entity. These facilities offer organizations maximum control over their data and systems, making them particularly valuable for organizations that handle sensitive information, such as healthcare providers or financial institutions.

Organizations unable to afford the high upfront cost of building their own data center often rent server space at a colocation data center. In doing so, businesses can seamlessly scale their computational needs based on seasonal demand and utilize pre-audited, colocated facilities to meet compliance requirements without maintaining in-house IT and compliance teams.

Hyperscale data centers operate at the largest scale, providing foundational services for social media platforms, cloud computing operations, and artificial intelligence computations. These facilities are typically at least 10,000 square feet, however, the largest facility is roughly 10.6 million square feet, equivalent to 165 U.S. football fields.³

Background on Data Center Operations

The landscape of global data center infrastructure is complex and rapidly evolving, making precise estimations difficult to establish. However, the United States maintains clear dominance in this sector, housing approximately 5,300 operational data centers, representing 45% of the global total of 11,800 facilities.⁴ In the U.S., 80% of data center energy demand is concentrated in just 15 states, primarily in California, Texas, and Northern Virginia.⁵ Northern Virginia particularly emerged as a landing spot for data center development due to its reliable power infrastructure, location near a critical internet exchange point, and strategic proximity to Washington, DC.⁶ This global concentration in data centers reflects the nation’s early adoption of digital infrastructure and continued leadership in cloud computing and technology services.

While media attention often fixates on hyperscale facilities operated by major technology companies, this focus obscures the true composition of the data center market. Colocation facilities, which provide shared infrastructure for multiple organizations, comprise over half of the marketplace.⁷ The U.S. is responsible for approximately a third of colocation data facilities.⁸ The hyperscale segment, despite its outsized presence in news coverage, consists of only about 1,000 facilities globally. The United States accounts for over half of hyperscaler energy capacity.⁹ Within this hyperscale sector, three companies dominate: Amazon Web Services, Microsoft Azure, and Google Cloud Platform collectively operate 60% of these massive computing and storage environments.¹⁰

The data center market’s structure reveals surprising diversity beneath the surface. While six major colocation operators control 37% of the global market, the landscape is far from consolidated. According to Synergy Research Group, a remarkable 50% of the market consists of numerous medium and small-scale data center operators.¹¹ This fragmented composition challenges the common perception of a market dominated by tech giants, highlighting instead a vibrant ecosystem of diverse providers serving different niches and regions.

Why Are They Essential to AI and Tech?

Just as the Interstate Highway System of the 1950s transformed America by connecting its cities with ribbons of asphalt, today’s digital infrastructure revolution is reshaping society through networks of fiber optic cables and data centers. While President Eisenhower’s ambitious project facilitated the physical movement of goods and people, our modern challenge is building digital highways that carry everything from entertainment to critical financial and healthcare data across the globe at the speed of light.

Every digital action in our lives flows through the beating heart of a data center. When 300 million Netflix subscribers press play simultaneously, data centers orchestrate a symphony of streaming video across continents.¹² Each Instagram story shared, every Spotify song streamed, and all 140 billion WhatsApp messages sent daily pulse through these digital nerve centers.¹³ Behind each tap of your smartphone lies an intricate ballet of servers, processing millions of calculations per second to deliver seamless digital experiences. These facilities don’t just store our digital world—they breathe life into it, transforming billions of binary signals into social connections, entertainment, and essential services that power modern life.

One place this may not be obvious because it has become mundane is in the instant access to photos, songs, movies, and information of all kinds from anywhere. Each of these digital actions reflects humanity’s data explosion. Every day, the world generates over 402.74 million terabytes of information more data than was created in the entire year of 2003.¹⁴ The COVID-19 pandemic dramatically accelerated this digital transformation, as global shifts to remote work, online education, and streaming entertainment drove digital content creation up by 56% between 2019 and 2020.¹⁵ This surge shows no signs of slowing, with data generation expected to maintain an annual growth rate near 19%, pushing global data creation beyond trillions of gigabytes each year by 2025.¹⁶ As businesses and consumers migrate to cloud storage, which now hosts nearly half of the world’s digital assets, data centers are the warehouses of the information age.¹⁷

Data centers have revolutionized the financial landscape, powering a digital economy where money moves at the speed of thought. Every time someone taps their phone to split a dinner bill or a merchant swipes a customer’s card, data centers process these transactions in milliseconds, handling more than $22,000 per second of PayPal transactions alone.¹⁸ Beyond simple payments, they’ve enabled a financial. revolution: street vendors in rural areas now accept digital payments through QR codes, while global corporations orchestrate complex supply chains with real-time payment settlements across borders.¹⁹

In the realm of e-commerce, data centers serve as the central nervous system for retail giants like Amazon, processing over 8,000 transactions per minute during peak shopping periods.²⁰ These digital fortresses simultaneously track millions of items across vast warehouse networks, dynamically adjust prices based on real-time demand signals, and analyze billions of customer interactions to predict shopping patterns. When you click “buy now,” you trigger an intricate dance of algorithms that coordinate everything from fraud detection to inventory management, all within fractions of a second.

The relationship between artificial intelligence and data centers represents a technological symbiosis that’s reshaping the future of computing. Training a single large language model demands more computational power than launching a rocket to Mars, requiring specialized data centers that more closely resemble supercomputers than traditional server farms. These AI factories house dense arrays of GPUs and TPUs—processors specifically engineered to perform millions of simultaneous calculations, enabling them to compress decades of traditional computing time into mere months.

Unlike conventional software that processes tasks sequentially, AI workloads operate more like a digital brain, requiring constant, intensive computation across thousands of interconnected processors. Modern data centers have evolved to meet this challenge, transforming from simple storage facilities into sophisticated neural networks of hardware. These specialized facilities enable companies to develop increasingly sophisticated AI models that can understand speech, recognize images, and engage in complex reasoning.

The emergence of artificial intelligence poses new energy challenges for data centers. AI workloads are exceptionally power-intensive, with a single ChatGPT query consuming nearly ten times the electricity of a standard Google search.²¹ This dramatic increase in power requirements stems from the unique computational demands of AI systems, particularly in training large language models.²²

As AI advances toward more ambitious horizons, data centers stand as the foundation of this revolution. Their evolution from housing simple web servers to orchestrating vast AI computations mirrors the journey from early computers to today’s artificial intelligence. The next generation of AI breakthroughs will emerge from these digital crucibles, making data centers the factories of the intelligence age.

Water and Energy Use by Data Centers Is Manageable and Often Exaggerated

Data centers consume massive amounts of resources, driven by two critical requirements: maintaining uninterrupted server operation and implementing sophisticated cooling systems. The need for 24/7 uptime demands enormous electrical power while preventing hardware damage through temperature control requires extensive cooling infrastructure, often consuming as much energy as the servers themselves.²³ This dual demand creates a significant environmental and economic challenge, particularly as the digital economy’s growth drives the construction of ever-larger facilities in more locations worldwide.

The energy demands of data centers have reached staggering levels, driven by operational and cooling requirements. In 2023, these facilities consumed 4.4% of total U.S. electricity, equivalent to the power usage of several million households.²⁴ This appetite for energy is accelerating dramatically, with projections indicating data centers could account for 6.7% to 12.0% of national energy consumption by 2028.²⁵ The wide range in these estimates reflects uncertainty around artificial intelligence adoption, as workloads require significantly more computational power—and consequently more energy—than traditional computing tasks.²⁶ This surge in energy demand has raised alarm among utilities and regulators, who worry that concentrated data center development could strain local power grids, potentially leading to blackouts or increased energy bills for customers. These concerns, while understandable, are largely misplaced.

Despite consuming over 176 TWh in 2023²⁷—more than triple their 2014 consumption—data centers represent a tiny fraction of the United States’ total electricity production of 4,178 TWh.²⁸ While data centers may represent one of the fastest-growing segments of energy demand, their absolute contribution remains dwarfed by industrial sectors that continue to dominate the nation’s energy landscape. For comparison, data centers consumed less than 3% of the energy consumed by U.S. manufacturing.²⁹ Transportation and residential sectors also substantially outpace data centers, with the transportation sector alone consuming approximately 27% of total U.S. energy.³⁰

Examining the broader industrial landscape further diminishes these concerns. While larger data centers can consume between 20 MW to over 100 MW of electricity, this usage represents just one component of broader industrial demand growth.³¹ Efforts to reshore manufacturing is another important factor in load growth as factories are energy-intensive.³² Altogether, research from the
R Street Institute confirms that AI remains a “formidable but minority driver of electric demand growth expectations” when compared to industrial sources of load growth.³³ Even high-end projections indicate data centers will not dominate new load growth, but represent only about a quarter of new electricity demand.³⁴

Remarkable efficiency gains further offset the industry’s moderate share of energy consumption. Between 2010 and 2018, data center computing workloads increased by nearly 550%, yet electricity consumption rose by only 6%, demonstrating the industry’s ability to dramatically improve efficiency while expanding operations.³⁵ In 2007, data centers operated at a power use effectiveness (PUE) rating of 2.5, meaning only 40% of their energy directly powered IT infrastructure, while the remaining 60% supported cooling and other auxiliary systems.³⁶ By 2021, technological advances had driven the average PUE down to 1.57, enabling data centers to devote approximately 60% of their energy consumption to IT equipment.³⁷ This trend toward greater efficiency will continue as cooling and electrical infrastructure technologies advance. Furthermore, the industry’s shift toward colocation and hyperscale facilities instead of smaller edge facilities allows organizations to leverage economies of scale, promising even more efficient use of energy resources.³⁸

Water consumption has emerged as another critical environmental concern for data centers, especially in Western states where existing water scarcity issues already strain regional resources. Data center water consumption occurs through two primary channels: direct usage in cooling systems and indirect consumption through power generation.³⁹ This interconnection means that improvements in energy efficiency yield cascading benefits for water conservation as well.

The scale of water usage is substantial – a typical midsized data center consumes approximately 300,000 gallons of water daily, with requirements particularly intensive in regions where air cooling alone proves insufficient.⁴⁰ A common industry comparison is to golf courses. Google’s sustainability report, for example, says that their data centers and offices use the equivalent water as “43 golf courses on average in the southwestern United States.”⁴¹ Compared to the nearly 16,000 golf courses just in the US, these data centers represent a relatively inconsequential portion of overall water consumption.⁴²

The global data center industry has opportunities to improve its water usage reporting. A 2024 survey of global data center operators found that 43% of the operators “compile and collect” water usage for their data centers for sustainability purposes.⁴³ This suggests that many data centers are not located in areas where water usage poses an environmental or sustainability challenge, such as regions with abundant water resources. An earlier survey suggested that this was the likely explanation, as almost two out of three responses indicated that there was no business justification for collecting water usage data.⁴⁴ Another one in four also said that because the cost is “so low it is not routinely collected.”⁴⁵ However, there is still a data gap of some extent. This lack of comprehensive monitoring is particularly concerning in drought-prone regions, where water resource management is increasingly critical.

The absence of industry-wide water usage data presents a challenge for assessing and improving data center sustainability. Without accurate baseline measurements and consistent reporting across facilities, it becomes difficult to establish effective conservation strategies or set meaningful reduction targets. This information gap contrasts sharply with the industry’s more mature approach to energy monitoring and highlights an area requiring immediate attention and standardization.

Data centers are actively adapting their cooling strategies to optimize water consumption. Traditional evaporative cooling systems, while initially cost-effective, can result in significant water loss through evaporation. Recognizing these inefficiencies, the industry has moved toward more sophisticated cooling solutions. For example, Microsoft facilities are employing closed-loop cooling systems that have reduced water use by 29% since 2021, while others have adopted natural cooling solutions that leverage nearby ocean water or filtered cold outside air to reduce reliance on potable water.⁴⁶

The industry’s commitment to water conservation extends beyond natural cooling solutions. Rather than drawing from municipal drinking water supplies, which directly compete with community needs, industry leaders are increasingly tapping non-potable sources.⁴⁷ Amazon Web Services (AWS), for example, has begun using purified wastewater, rather than drinking water, in several data centers across the country to leave potable water sources unaffected.⁴⁸ This approach represents part of a broader industry commitment in which Google, Amazon, Microsoft, and Meta have committed to becoming water positive by 2030, replenishing more water than they use each year.⁴⁹

Furthermore, while consumption estimates often make headlines, they frequently overlook the significant portion of water that returns to municipal wastewater treatment cycles for future use, making the net impact on water resources substantially lower than raw usage figures suggest.

The water consumption caused by thermoelectric power generation is unavoidable. However, data centers have made significant efforts to source their energy from renewable sources, such as wind and solar, that do not consume water when available. In fact, Google has pledged to operate on 24/7 carbon-free energy (CFE) by 2030, with Google already reaching 67% of their goal in 2020.⁵⁰ Other larger data center providers are making similar investments in renewable energy to offset their resource consumption. Looking forward, as data center infrastructure improves its energy efficiency, water consumption from energy generation is expected to continue declining.

The ongoing development of data center infrastructure shows steady progress in resource management. While these facilities will continue to require significant energy and water resources, advances in cooling systems, the adoption of renewable energy, and improvements in computing efficiency are helping to mitigate their environmental impact.

Though many of the carbon-free or water restoration commitments stem from stated environmental concerns, policymakers should keep in mind the market incentives in the background as well. Because data centers are large resource users, they have sizable market-driven incentives to invest in energy efficiency and water efficiency improvements. This incentive also applies to their interest in having a reliable and abundant supply of both water and energy. The investments that data center developments make in public infrastructure of all kinds reflect this incentive.⁵¹

Economic Impact of Data Centers

Data centers serve as powerful economic catalysts in their host communities, generating both immediate and enduring benefits that ripple throughout local economies. During construction, these facilities employ approximately 1,700 skilled workers over an 18-24 month period, including electricians, engineers, and HVAC specialists.⁵² Once operational, data centers provide nearly 160 stable, high-paying positions for IT professionals, facilities managers, security personnel, and maintenance staff.⁵³

While this may appear modest compared to the job creation from large manufacturing plants, it understates the facility’s true economic impact. Research indicates that each additional job in the US data center industry supports 6.4 additional jobs elsewhere in the economy through increased consumer spending and growth of supporting industries.⁵⁴ Moreover, these positions garner salaries well above regional averages. A 2024 survey reports the average, annual data center salary of about $160,000 for permanent jobs.⁵⁵ This is around twice the 2023 national median household income of $80,610.⁵⁶ For example, a new $10 billion data center project in northeast Louisiana’s Richland Parish is expected to support at least 500 jobs with salaries at least 150% of the state’s per capita average, elevating disposable income in host communities.⁵⁷

Beyond fostering employment, data centers significantly boost local tax revenues, helping fund essential community services and infrastructure improvements. The facility’s extensive real estate footprint substantially expands the local property tax base, while the high-value computing infrastructure inside generates considerable personal property tax revenue. These facilities’ tax contributions often dwarf those of other commercial properties due to their dense concentration of valuable equipment that requires regular updating.

The transformative effect of this revenue is exemplified in Loudoun County, Virginia, a county that has become a hotbed of anti-Data Center activity. The county has, however, leveraged the data center industry to dramatically improve its fiscal position. The county’s data center tax base has enabled it to maintain the region’s lowest residential property tax rate —approximately 25% lower than surrounding jurisdictions— while simultaneously expanding public services and investing in community infrastructure. The county now reports that almost half of its revenues are generated by data centers. Combined with prudent budgeting, tax rates have actually fallen since 2008 in Loudoun–from about $1.28 per $100 assessed value to $0.80 per $100 of assessed value.⁵⁸

This robust revenue stream proves particularly valuable because data centers place minimal demands on public resources compared to other development types. They require relatively little public infrastructure, generate minimal traffic, and don’t burden school systems. In Loudoun County, data centers contribute $26 in tax revenue for every dollar of local public services required.⁵⁹ This favorable cost-benefit ratio makes data centers particularly attractive for local governments seeking to maximize their return on infrastructure investments.

Even when states offer tax incentives to attract data center developers, net fiscal gains remain strong. Pennsylvania’s projections of sales and use tax exemption generated a $110 million annual surplus.⁶⁰ In a retroactive analysis determining whether or not to revoke Wyoming’s tax exemption for data center purchases, the state found that they would lose approximately 60 jobs per year if the incentives were removed.⁶¹

Policymakers should consider how their tax incentives and economic development incentives operate. Policies that exempt data centers from some taxes obviously reduce the taxes collected. Sales tax exemptions, for example, are common. Broadly, these are exemptions that maintain tax neutrality (treating all industries the same) and prevent pyramiding (when something is taxed multiple times before it is purchased by the end user).

Complex and punitive taxation discourages investment. Jared Walczak of the Tax Foundation estimated that a data center in Kansas would effectively pay a rate 21 times what a new manufacturer would pay.⁶² Notably, Walczak points out that Nebraska only taxes data centers at an effective rate of 1 percent.

There can clearly be dramatic differences between states that attract or deter investment that go far beyond any specific tax incentive or economic development incentive.⁶³ Research suggests there are many possible reforms to improve the tax treatment of data centers that will benefit the communities where data centers invest alongside data center companies.⁶⁴ These include allowing or making permanent immediate depreciation of capital investments or lowering other state tax rates (such as excise tax rates, franchise taxes, and gross receipt taxes).⁶⁵

The presence of data centers often attracts complementary businesses and technology companies, creating clusters that further diversify the local economy. From 2016 to 2023,⁶⁶ the sector’s remarkable 60% employment growth reflects not just increasing digital infrastructure demands but how data centers have become vital anchors for regional economic development. Furthermore, many data center operators invest in community initiatives, including workforce development programs and partnerships with local educational institutions, helping build sustainable talent pipelines while providing career opportunities for residents.⁶⁷ This comprehensive economic impact transforms host communities into emerging tech hubs, positioning them for longterm growth in the digital economy.

Widespread Interest in Attracting Data Centers Exists

Inspired by these economic benefits and the potential for artificial intelligence, recent interest in data center development has skyrocketed. Industry projections suggest that meeting AI-driven demand will require approximately one new megascale data center to come online every week through 2030.⁶⁸ In 2025, at least 16 states have introduced legislation related to data centers.⁶⁹ There is an extensive and growing debate on how to provide the necessary energy infrastructure to enable basic internet services and the expanding uses of artificial intelligence.

These data center proposals are growing in size, with some approaching gigawatt scales. This would be several times larger than the demand of some cities. For example, the peak load of Garland, Texas, in 2023, was 514 megawatts.⁷⁰ So, a one gigawatt data center site would be two Garlands of electricity.

At the same time, there is extensive uncertainty. It’s unquestionable that additional data centers will be built. Where and what size are matters of debate. Previous moments of predicted load growth, such as with the original internet and the dot-com boom, overestimated how much demand there would be from the internet and personal computers.⁷¹

Speed To Power Is the Real Challenge

To simplify, a data center’s ultimate location depends on (1) their ability to build the data center and (2) their ability to then power that data center.

On the first condition, building the physical data center site is not particularly time-consuming, taking 18 to 24 months, on average. The process here is straightforward and requires obtaining legal control of the site and permitting through state and local jurisdictions. A simple way to improve this is by establishing byright processes for data center development.

On the second condition, there is more to consider. At one level, there is little question about America’s ability to power data center sites of this size. Data centers already operate today at the size and scale of hundreds of megawatts. Putting several of these sites together is not a great technical challenge. In addition, remember that much of the energy concerns around data centers are overplayed and exaggerated. They are large users, but many such energy-intensive industries already exist.

The problems arise from the need to build or find already-built generation resources to rely on. There’s a mismatch between electricity generation development timelines and the speed at which data centers can be built. Because of this, there’s extensive “queue shopping,” where would-be data center developers submit applications to be plugged in before they have built sites. Many of these are speculative attempts to find out where power might be available. Think about this process like calling multiple contractors to get quotes for a repair to your house. Once you have learned about the range of costs, you can choose which contractor to hire. Similarly, large loads, such as data centers, have often requested multiple utilities to connect and determine the speed to power and costs for their prospective development.

A Public Utility’s Core Mission Is To Provide Stable and Reliable Power

Difficulties in quickly powering new data centers arise from the regulatory rules governing the electricity system. Because states largely regulate their electricity system via a state monopoly, costs are often shared. This means that adding new generation sources or grid infrastructure does not move at the speed of commerce. New additions of either infrastructure or generation must be justified via proceedings with the public utility commission and related proceedings.

On the one hand, this slower pace makes sense for a public utility. In other markets, bad investments cause the company to go out of business. Customers of that company may suffer, but can turn to alternative providers. There aren’t alternatives for utility customers.

This means that bad investments in grid infrastructure or generation raise electricity costs. Imagine that a state utility invests in a new generator to power a potential 100 MW data center. This represents upfront costs and investments by the utility company. If no additional user of that electricity arrives, then that state’s citizens will pick up the costs in higher electricity rates. This is because approved investments are allowed to be “recovered” by electric utilities. Recovery is done through the electricity rates that public utility commissions approve. So, if state policymakers approve a new generation unit to be recovered, then customers will pay the price even if a new data center is not built.

The core mission of a public utility is to provide stable and reliable service. This is a critical role. But there is a fundamental mismatch between the institutional design of public utilities and today’s AI challenge. Public utilities rightly care first about avoiding bad investments that raise rates without need. New data centers care first and foremost about speed to power. Public utilities that speculate about which data center developments are likely to happen may inadvertently raise their electricity costs.

Enabling Data Centers and Protecting Consumers

Policymakers are rightfully interested in enabling artificial intelligence through continued data center development. However, they must also protect consumers from cost spillovers connected to poor investments by utilities.

The best way to protect ratepayers is to allow these customers to procure electricity through competitive service providers.⁷² Then, the state utilities can charge grid access fees just as clubs charge club dues. Any needed infrastructure supporting power needs for new data centers can be financed and managed independently of the rate base that is subject to recovery.

Broadly, arrangements like this are known as “retail choice.” Their central and largest benefit is that these remove the possibility that failed bets will be socialized. In contrast, when monopoly utilities build speculative assets for large loads, the risk falls squarely on the shoulders of all ratepayers. As NRG’s Travis Kavulla noted in technical comments about Virginia and data center integration, “The first, best, and only protection against stranded assets is to not allow a utility to build assets that may become stranded.”⁷³

The immediate problem with retail choice is explaining the need for average consumers to find and supply their own power. Though this is common in Texas, it is a large shift in practice for other states. However, this issue does not apply to sophisticated large customers, such as AI companies or technology companies. They already maintain and have teams to vet and negotiate contracts that procure power for their operations.

Competitive supply options are a win-win solution for the average citizen and large data center companies. First, it protects consumers from cost overruns or company failures. Although AI is destined to be a major contributor to economic growth and America’s standing, certain companies are likely to fail just as the technology field at large has had multiple bankruptcies. With retail choice options, average people are protected from paying for these mistakes just as they are in other industries. Second, retail choice options provide the ability to move quickly. Data center companies can negotiate for power supply with competitive suppliers.

For states worried about the effects on existing customers of adding new data centers , there are still options:

Limit the retail choice program to new electricity users.
Require companies to pay for investments made if they end up going unused.

For example, a state that creates a retail choice program could limit it to new developments. This would resolve worries about “stranded costs” – the idea that if a customer leaves a utility service, they might leave behind unpaid investments. Importantly, new large loads do not create stranded cost risk for existing ratepayers. These loads are incremental; they represent new demand above what the utility currently serves. If a 50 MW data center opts for a competitive supplier, the utility simply doesn’t invest in new power plants or contracts for that 50 MW under regulated rates. With no new investments dedicated to that load, there is nothing to “strand” on remaining customers. In short, when a new industrial load is served under new terms outside the ratebase, it carries its own costs and leaves legacy ratepayers unharmed. Legislators can be confident that allowing choice for new loads will not burden residential or small business customers with stranded asset costs – there are no legacy plants built for these yet-to-exist facilities.

Another option is to require that data centers pay for some portion of the infrastructure built to support them. One example is ongoing in Ohio. Since public utilities may still make investments in their wires system to power data centers, it is reasonable to charge the beneficiaries of those investments. To give certainty, data centers could be expected to pay some share of those costs even if they end up not using them in full. This insulates average ratepayers from the costs of data center development. However, they can also be designed to avoid kludgy processes. Companies are rightfully wary of being held responsible for costs or locked into losing business propositions. That said, they have signaled interest in such agreements by proposing their own.⁷⁴ Clearly, such requirements are controversial and require careful consideration.⁷⁵

Conclusion: Data Centers Are a Digital Foundation for the Modern World

Data centers are the digital foundation of the information economy. They support not just our social media or streaming habits but the core operations of finance, healthcare, logistics and, increasingly, artificial intelligence.

Concerns about their energy and water use are important, but they are often exaggerated or misunderstood. Data centers are among the most efficient and rapidly improving industrial operations in the country. Their footprint is smaller than manufacturing or transportation, and their innovations—from advanced cooling to renewable integration—are helping lead broader sustainability gains across sectors. Where there are real risks, market entrepreneurs can help meet the energy needs of data centers.

Policymakers should not view data centers as threats to local grids or water systems, but as opportunities to strengthen infrastructure and create high-paying jobs. States can welcome data centers without jeopardizing ratepayers by encouraging competitive procurement models. Allowing new industrial loads like AI firms to source power independently can reduce risk to residential customers, avoid unnecessary investments, and accelerate deployment. States can also design shared-cost models that ensure fairness without deterring investment.

Data centers are not a luxury—they are a necessity for a modern, thriving economy. States that act now to align policy with infrastructure realities will be best positioned to attract investment and share in the growth of the digital age.

Find the pdf version of “Digital Foundations: The Essential Guide to Data Centers and Their Growth” here.

D⁠i⁠g⁠i⁠⁠t⁠al Founda⁠t⁠⁠i⁠ons: The Essen⁠t⁠⁠i⁠al Gu⁠i⁠de ⁠t⁠o Da⁠t⁠a Cen⁠t⁠ers and The⁠i⁠r Grow⁠t⁠h

Table of Contents