AI Data Center Investment Outpaces Power Supply and Public Support, Shifting Focus From Optimism to Practical Feasibility

Opposition to data centers spreads as power and water concerns compound job insecurity
Industry struggles to win over residents, raising concerns that expansion delays could hinder investment recovery
With expansion still necessary, industry faces task of assessing supply conditions and demonstrating benefits

Growing conflict over the construction of artificial intelligence (AI) data centers in the United States is putting the industry's expansion strategy to the test. As residents concerned about electricity and water supplies and job losses call for construction restrictions, companies must both recoup their investments and win over local communities. Meanwhile, evidence of AI-driven cost savings and productivity gains in business operations continues to fuel demand for more computing facilities. Industry observers say construction must be paced to match power and equipment availability, while the benefits AI can offer residents and workers must be demonstrated in concrete terms.

Schmidt: “Data Centers Should Avoid Residential Areas and Secure Their Own Power”

According to U.S. news outlet Axios on the 9th (all dates hereafter are local time), former Google CEO Eric Schmidt proposed locating data centers away from densely populated areas and establishing their own power supply systems to reduce residents' opposition to construction. He explained that the United States has ample vacant land available, leaving no reason to concentrate large facilities near residential areas. In particular, he said data centers should secure the electricity they need themselves to ease pressure on the existing grid. The proposal is seen as a call to consider both residents' living environments and electricity supply and demand from the site-selection stage.

Local opposition to data centers was also evident in the latest U.S. midterm elections. In an Associated Press voter survey, about three-quarters of respondents opposed the construction of new data centers in their communities. Among related local ballot measures tracked by polling firm Redbird Consulting, construction restrictions passed in 44 of the 45 votes for which counting had been completed. Residents' concerns include higher electricity bills and water shortages. In Texas in particular, fears emerged during the election campaign that data centers' heavy water consumption would strain supplies for farming and livestock operations already reduced by drought.

Schmidt blamed the technology industry's communication practices for deepening public distrust. His assessment was that the industry had discussed AI's negative aspects without adequately explaining its benefits, directing voters' attention toward the risks. Companies are also moving to spell out the benefits and resident protections they will offer communities hosting data centers. On the 5th, OpenAI unveiled a plan outlining commitments to communities where data centers are developed and called on federal, state and local governments to establish clear construction rules. The plan called for institutional standards to protect residents and ensure that the benefits of AI infrastructure are shared with local communities.

Debt-Heavy Data Center Industry Faces Pressing Need to Win Over Residents

The industry's efforts to win over residents are also tied to the finances of data center businesses that have expanded through borrowing. According to Reuters, financial institutions have sharply increased their exposure as lending has poured into data centers and emerging cloud providers targeting demand for AI computing. JPMorgan's outstanding income-producing real estate loans, which include data centers, stood at USD 29 billion at the end of the third quarter, with their share of total commercial real estate lending rising from 8% three years earlier to 12%. More recently, concerns about construction delays and tighter regulation have prompted lenders to scrutinize facility startup deadlines and the conditions under which customers can scale back or terminate contracts. Failure to bring facilities online on schedule could disrupt anticipated revenue and loan repayments, meaning project delays caused by local opposition also burden companies and financial institutions seeking to recoup their investments.

Opposition to data centers has intensified further as fears grow that the spread of AI will threaten livelihoods. According to the Associated Press, Kardal Coleman, who leads the Democratic Party in Dallas County, Texas, cited job displacement, air pollution and noise in residential areas in an email criticizing data centers. The perception that data centers support the spread of AI that will replace workers has become part of the opposition to development. Complaints that construction is being pushed through without adequately reflecting residents' wishes have also led to nationwide collective action. On July 18, the civic group Humans First organized 142 protests across 42 U.S. states, with participants in Atlanta displaying messages saying residents had never consented to the construction of massive AI data centers.

Dispersing Sites Emerges as a Solution for Expansion

As conflict over data center construction grows, calls are mounting for development approaches that account for both AI investment and the burdens on local communities. Schmidt's proposals to disperse sites and secure independent power supplies are also intended to enable continued expansion of AI infrastructure. At an AI event in Washington, D.C., in October 2024, he said preparations were insufficient to meet climate goals and argued that investment in AI development was needed to find solutions to environmental problems. He also judged that improvements in battery and transmission technology alone would struggle to meet AI's rapidly growing energy demand. His latest proposals on siting and power supply are likewise seen as ways to reduce friction with local communities so that data center investment can continue.

For development to continue, the first step is to align data center construction schedules with the timing of power and equipment supplies. In a recent analysis of AI capital expenditure, JPMorgan identified securing reliable electricity as the biggest constraint on expansion, noting that the time and cost required to reinforce power grids are delaying facility startup schedules. In particular, lengthy procurement times for key equipment, including graphics processing units (GPUs), memory and cooling systems, are holding up construction. Data centers typically take 18–30 months to build, and completion can be delayed further by shortages of skilled workers and permitting delays. Even with buildings and servers in place, operations must wait until grid connections and cooling system installations are complete.

Table 1. Data Center Development Challenges and Responses

The Case for Phased Data Center Investment

Financial market analysts have also suggested that moderating the pace of expansion could improve investment efficiency. In an analysis released on the 9th, asset manager Invesco said slower data center construction could help companies apply greater scrutiny to capital expenditure and sustain the AI investment cycle for longer. Its assessment was that, with corporate AI adoption and computing demand continuing to grow, companies could reduce excessive spending by staggering investments across projects. JPMorgan likewise suggested that a post-midterm slowdown in construction could give AI service monetization time to catch up with upfront investment. The rationale is to make greater use of existing facilities while adjusting expansion schedules, then proceed with follow-on investment as customer demand and revenue are confirmed.

After the midterm elections, state-level regulatory revisions and the relocation of project sites are seen as variables that could determine expansion schedules. Goldman Sachs expects some states to temporarily halt new construction while revising the relevant rules. Texas Governor-elect Gina Hinojosa has in fact said she would pause new projects until safeguards for residents are in place. In a letter to the governor-elect, CoreWeave also requested discussions on completing already permitted projects and establishing clear standards for future development. The new state administration and the industry will need to coordinate on whether construction can continue on previously approved projects and on approval conditions for new ones.

AI Delivers Major Cost Savings, but How Its Benefits Are Communicated Must Change

Even as expansion schedules and approval conditions are negotiated, companies still have strong economic incentives to adopt AI. The technology is cutting costs by reducing time spent on repetitive tasks and bringing work previously outsourced to specialist firms in-house. According to Reuters, U.S. consumer goods company Kimberly-Clark used an AI platform to shorten advertising content production time from 24 days to 2 hours. The platform also helps select suitable influencers for advertisements and adapt content for individual markets. As companies handle more advertising work internally, they are reducing both the work outsourced to agencies and the time required for production.

In customer service, AI has also improved productivity by helping employees handle their work. A study by researchers at Stanford University and other institutions analyzing 5,172 customer service agents found that employees assisted by generative AI resolved an average of 15% more customer inquiries per hour. AI helped agents respond by suggesting answers tailored to customer inquiries, with particularly strong improvements among less experienced or less skilled employees. By sharing the response practices of skilled employees through AI, the technology helped new hires adapt to their work and increased the volume of inquiries handled.

The results demonstrated in business operations also offer lessons for how AI should be explained to the public. Messaging that emphasizes replacing workers risks causing even AI's ability to ease workloads and improve skills to be perceived as a threat to employment. On the Futurology podcast last month, Stanford professor Fei-Fei Li also criticized the practice of associating AI with “replacement,” stressing that it should be described as a technology that expands human capabilities. In particular, she proposed human dignity, agency and community as guiding standards for AI development and highlighted potential uses in education, drug discovery and energy innovation. This suggests that addressing the inadequate explanation of benefits identified by Schmidt requires communicating proven results, such as shorter working hours and better use of specialized knowledge, in concrete terms. For an industry seeking support from residents and workers, messaging that highlights improvements in employee skills and public services is also likely to gain traction.