Data centers meet democracy: The backlash to the AI build-out

Data centers represent the bedrock of the AI build-out, providing the computing power to process large language model (LLM) requests from the industry’s biggest players. That’s why any speedbumps are important for investors to consider—including the current political environment and its ramifications for access to resources.

With the accelerated build-out, those demands are substantial. Data centers consume enormous quantities of electricity and require new power transmission and generation capacity. One 100-megawatt data center has the electricity consumption of approximately 83,000 average U.S. homes.⁴ That has kicked off competition among developers to secure land, electricity, water and cooling technology, among other inputs.

As the build-out ramps up, supply chains and energy solutions have lagged. One of the key hurdles is access to power. Estimates suggest the United States will reach a power shortfall as early as 2029 as demand grows at 3.2% per year over the next decade—driven by data centers, industrialization and electrification.

For consumers, the growing imbalance between power supply and demand is already pushing prices higher as data centers tap into local electricity grids. One example is a capacity auction by grid operator PJM, which supplies nearly half of U.S. data center power demand. In July 2025, the auction—which creates a forward market that pays for generator availability in the future—hit a record $329.17 per megawatt per day.⁵ That auction marked a 1,000% rise in prices over the last two years. Given affordability’s position as the top voter concern in the upcoming midterm elections, the election’s verdict could set the tone.

The pushback

Data centers are by far the most tangible component of the AI build-out, drawing the public’s questions, concerns and pushback. This trend has ramped up in the past few months, just in time for election season.

Polling data shows this is a top issue for Americans with three quarters of the country saying they would oppose the construction of a data center near their home. And it’s more than just talk: Disputes over data center construction have more than tripled this year.⁶

Politicians are listening. The surge in public interest and advocacy around data centers has already made an impact on campaign rhetoric. In line with the negative shift in public opinion, the loudest voices on the campaign trail tend to be neutral or negative on data centers, creating an unusual amount of bipartisan agreement in today’s politically polarized environment.

Until recently, Democrats were the main critics of the AI build-out. The party’s traditional focus on labor and consumer safety aligns with skepticism towards a technology that some fear will displace large numbers of workers. If Democrats win control of the House and Senate, they could pursue stricter environmental reviews, and place a greater responsibility for data center costs on developers—although any bill would likely require the support of the executive branch.
 
As recently as a few months ago, Republicans supported faster permitting and AI infrastructure expansion, so a Republican majority in the House and Senate was the most bullish scenario for the data center build-out. But now, a growing number of Republicans have staked positions further from the Trump administration’s enthusiasm for AI.

Growing voter backlash and signs of bipartisan consensus could still force developers to absorb more of the infrastructure-related expenses. These signs suggest the AI infrastructure build-out will continue, but who bears the costs will change. And that rulebook is likely to vary by state.

The policy divide

In the absence of a national framework on the data center build-out, state-level legislation is falling into three broad approaches.

• The Texas approach: Initially an attractive destination for the data center build-out because of its independent electricity grid, the Electric Reliability Council of Texas (ERCOT), greater than expected backlog of proposed electricity demand has resulted in Texas halting new data center grid connections as the state audits the requests and ensures the costs don’t fall on ordinary ratepayers.
• The New York approach: The state has created a temporary moratorium on new projects, temporarily stopping the build-out while the state establishes rules to navigate the growth. Proposals have included shifting the incremental electricity and infrastructure costs to the data center developer rather than equally with households who share the same access to power. Other controls include noise limits, water restrictions, environmental regulation and zoning concerns.
• The Pennsylvania approach: Using a conditional incentive structure, data center developers in the state can receive government support for the build-out, but only if they demonstrate benefits to communities, workers, ratepayers and the environment.
  

The larger political question has begun to shift from “Should America build data centers?” to “Who should pay for them, where can they be built and what protections must communities receive in return?”

Shifting policies could mean rising costs of building and operating data centers, squeezing profit margins for both tech giants and the companies contracted to construct these facilities. Given the reliance of data centers on power grids, utility stocks in particular have become tied to the AI trade, benefiting from the rallies, but also underperforming as concerns on the build-out arise.

The announcements of data center moratoriums from New York and Texas have created a particular overhang on the utilities sector, with its share price underperforming the broader S&P 500 since the announcements. This is especially significant in Texas where widespread support for the data center build-out, supported by affordable land and an independent electricity grid, faced an unexpected roadblock. The move is evidence that investors are beginning to price in a reality where the AI data center build-out could face hurdles beyond just capacity and supply restraints. The politics may matter too for financial markets more directly than they have in previous cycles.

Conclusion

Public opinion can impact future policy even if it doesn’t immediately change activity or regulation. Americans are starting to think about what data centers are doing to their energy bills with a sensitivity akin to the way they watch gas prices or the housing market. The question that still needs an answer is how this issue will determine voters’ choices. Because of this, politicians and parties strategizing for future elections will be watching the results very closely. For investors, the most important measure to look for on November 3 may not be whether the Democrats or Republicans notch more wins, but how well candidates do on either side of the data center issue.

If politicians who bash the AI build-out win races across the board, more politicians will take similarly strict positions in the future. This outcome could prompt an uptick in proposed regulatory legislation or even lead to federal regulation becoming a priority for the next Congress.

If the results are mixed, or exit polls show that voters prioritized other issues, politicians will have a different takeaway. They may see it as too risky to stake out a firm position on such a rapidly evolving issue without the guarantee of public support. The result would be a majority of both state and federal lawmakers who put additional regulations on the back burner.

The 2026 midterm elections won’t determine whether the U.S. builds AI infrastructure, but may shift timing, who pays, and where it gets built.

We can help

For more information about what the AI build-out and the evolving regulatory landscape might mean for your portfolio, contact your J.P. Morgan team.