Amazon has decided to abandon a data center construction plan in the United States after encountering opposition from residents. Earlier this year, Amazon Web Services (AWS), Amazon's cloud division, purchased a George Washington University campus site in Loudoun County, Virginia, intending to build four data center buildings.
However, residents opposed the proposal over concerns about large-scale electricity use and effects on their living environment, while the local government also took a negative view of data center development at the site.
Ultimately, Amazon and the Loudoun County government provisionally agreed on October 20 to revise the development plan. Instead of proceeding with its data center construction application, Amazon agreed to withdraw it and transfer 100,000 square meters of the approximately 490,000 square meters it purchased to the local government at no charge.
The local government agreed to review options for sports and cultural facilities on the land and expedite rezoning procedures to allow housing and commercial facilities on the remaining land Amazon will retain. Amazon separately offered to donate $5 million to a local housing fund.
A final agreement has not yet been signed. Nevertheless, it is noteworthy that a company that purchased land to expand its AI and cloud businesses had to change its development plan because of residents' opposition. Although the US government is encouraging data center expansion to secure the AI industry's competitiveness, opposition is growing in host communities over the burdens on electricity supply and the living environment. Even when investment is needed nationally, the benefits and burdens for local residents are a separate matter.
The US federal government has accelerated data center expansion. In July 2025, President Trump signed an executive order simplifying federal permitting and making suitable federally owned sites available for data center construction. The White House described it as an infrastructure policy supporting US AI leadership and national security.
However, describing facilities as necessary for national competitiveness does not by itself secure local consent. Residents are concerned with more than how much AI will contribute to the US economy. They also care about how electricity bills and their living environment will change once a facility arrives, and who will bear the additional burden. Alongside securing electricity and land, sharing costs is becoming an issue in the US AI competition.
Electricity bills are a more immediate concern than AI's future
In a joint NPR, PBS News and Marist poll released on October 27, 66% of likely voters opposed a data center being built in their community. Support stood at 27%. In the same survey, 59% said data centers had a negative effect on the United States as a whole.
Opposition is also difficult to explain solely as political hostility toward the Trump administration. A separate survey released by the University of Massachusetts Amherst in September found that 65% of Americans opposed AI data center construction in their communities. Among Republican supporters, opposition at 52% far exceeded support at 18%. Opposition was stronger among Democratic supporters, but Republican supporters did not welcome local construction either.
That does not mean opposition to data centers can readily be interpreted as rejection of AI technology itself. When University of Massachusetts researchers asked respondents to explain their reasons freely, they mainly cited environmental and resource use concerns, land use and effects on communities. Distrust of AI and related industries overall was mentioned less often. The potential of the technology and the location of the facilities needed to run it were separate matters for judgment.
Perceptions of electricity bills have also worsened. In a survey released by Pew Research Center in September, 50% said data centers negatively affected household energy costs. That was 12 percentage points higher than the 38% recorded in January 2026.
This does not mean data centers actually caused bills to rise by that amount. It does, however, show that residents see data centers not only as investments in future industries but also as facilities that could affect their living costs.
These concerns have also led to legislative debate. A September 15 tally by the National Conference of State Legislatures (NCSL) included 16 states that had considered bills to temporarily suspend or restrict data center development. Some bills were defeated or blocked by governors' vetoes.
This does not mean 16 states actually banned construction. Nevertheless, there is a clear shift: facilities once courted with tax incentives have become the subject of discussions about suspending development.
Investment attraction and residents' benefits are not the same
Attracting data centers does bring economic benefits. In a 2024 report, the Virginia General Assembly's Joint Legislative Audit and Review Commission (JLARC) assessed that data center capital investment and local tax revenue had positive effects on local economies.
In five local jurisdictions with established data center industries, related revenue accounted for as much as 31% of total local government revenue.
However, the report distinguished construction investment from long-term employment. According to industry representatives cited in the report, a typical data center employs about 1,500 people at peak construction, but its full-time workforce after operations begin, including contract workers, is about 50. Operational jobs tend to be well paid, but employment is lower than during construction.
Under this structure, corporate investment, local government tax revenue and household benefits cannot readily be assumed to move in the same direction. Higher tax revenue can create room to expand public services or reduce other tax burdens. Meanwhile, residents near facilities directly experience changes to their living environment, and grid expansion costs may be allocated to users across a wider area.
Large investment announcements therefore do not automatically translate into local consent. Even when total investment is substantial, how residents' perceived benefits and burdens are distributed is a separate issue.
Additional costs arise even when electricity bills are paid properly
The scale of electricity demand is also changing. A report released by Lawrence Berkeley National Laboratory in June 2026 estimated that data centers would account for 11.8% of total US electricity consumption in 2030 under its reference scenario. The projected range, using different assumptions about equipment deployment and utilization, was 9.5% to 15.3%. Although this is not a confirmed level of consumption, it is a scale that cannot readily be disregarded when planning generation and transmission facilities.
It is not accurate to explain the problem as data centers failing to pay their electricity bills properly. JLARC's review of the rate structure in place in 2024 found that existing supply costs were being appropriately allocated among customers, including data centers.
Nevertheless, the report noted that rising data center demand could also increase costs for other customers. Additional generation and transmission facilities are needed, and electricity prices may rise if supply cannot keep pace with demand growth.
Paying for the electricity currently used is different from taking responsibility for all the grid costs needed in the future. Even if customers pay their prescribed share under the existing rate structure, substantial new demand can change the conditions of the electricity market itself. This is why the cost issue surrounding data centers is difficult to assess solely from a particular company's electricity bill.
There is also a risk in the opposite direction. If a utility expands facilities based on a company's demand forecasts but the data center is not built as planned or ceases operations, the costs of facilities without users remain. JLARC also noted the risk that these costs could fall on existing customers. Demand growing too quickly is a problem, but problems can also arise when anticipated demand fails to materialize.
The issue therefore extends beyond whether to supply electricity to data centers. Contracts have also come to address who will pay for facilities after a company's investment plans change.
Virginia's State Corporation Commission (SCC), which oversees public utility rates and other matters, approved a proposal to place Dominion Energy's large electricity customers in a separate rate class, GS-5. The measure aims to reduce the risk of costs shifting to other customers by separately calculating supply costs for customers that use large amounts of electricity, such as data centers.
The key provisions concern contract duration and minimum charges. Applicable large customers entering new contracts from January 2027 onward will be required to use and pay for electricity service for at least 14 years. Even if their actual consumption declines, customers subject to the rules will pay at least 85% of the transmission and distribution charges calculated on the basis of contracted demand. Exceptions apply to some existing customers.
The arrangement links companies' plans to consume large amounts of electricity to long-term payment obligations. If supply capacity was secured at a company's request, that company must bear certain costs even if actual use falls below expectations.
Companies have also begun emphasizing cost sharing. Announcing its Community-First AI Infrastructure plan in January 2026, Microsoft said it would ensure that the electricity and infrastructure costs required for its data centers were not passed on to residential customers.
The plan also included a promise not to seek lower property tax rates from local governments hosting its data centers. The company explained that even if it believed in AI's benefits, making the general public bear the additional electricity costs needed by the industry was neither fair nor politically realistic.
The federal government is also pursuing more than relaxed permitting. In March, the White House announced a Ratepayer Protection Pledge involving major companies including Amazon, Google, Meta and Microsoft. It calls for companies to pay for new electricity supplies and necessary grid improvements, and to negotiate separate contracts under which they pay for electricity and facilities secured for them regardless of actual use.
The actual burden depends on contracts with utilities, rate structures approved by regulators and whether additional supply facilities are completed. Even with separate rates, if electricity supply itself is insufficient, it is difficult to eliminate price pressure affecting all customers.
The challenge facing the United States goes beyond explaining why AI development is needed. It involves supplying the necessary electricity on time and establishing a structure capable of covering costs as businesses expand or contract.
Attracting large numbers of data centers and securing the conditions for their reliable operation are different achievements. This is why electricity bills and long-term contracts have become issues alongside investment competition for AI leadership.
Original Korean article: futurekorea.co.kr
English brief by Korea Wind Intelligence, koreawindintel.com
