Across the nation, states are taking matters into their own hands by setting regulations for artificial intelligence (AI). However, this decentralized approach has drawn criticism from former President Donald Trump, who is pushing for a unified federal stance on the issue.

In a post on Truth Social this Monday, Trump announced his intention to sign an executive order aimed at streamlining AI regulations. He argued that expecting companies to seek numerous approvals for every new AI-related initiative is unreasonable. His proposed “one rule executive order” seeks to simplify these processes and create a uniform standard.

This year has seen a flurry of legislative activity concerning AI across all 50 states, Puerto Rico, the Virgin Islands, and Washington, D.C. Lawmakers have introduced various bills addressing who owns AI-generated content, the use of AI in critical infrastructure, and employing AI for medical advice. The National Conference of State Legislatures reports that dozens of these measures have been enacted.

Specific to data centers, 21 states have passed laws, as noted by Multistate. Florida, for example, enacted HB 7031, which imposes limits on sales tax exemptions for energy purchases by data centers that are under 100 megawatts.

Last week, Florida’s Governor Ron DeSantis introduced an “AI Bill of Rights,” a comprehensive legislative package that aims to establish extensive guidelines on AI usage. This initiative not only addresses the implementation of AI but also tackles broader issues such as the impact of large data centers on consumers’ electricity bills.

[WATCH: Florida governor calls for artificial intelligence bill of rights]

How data centers power AI

As artificial intelligence and cloud computing drive soaring demand for raw processing power, data centers – the sprawling server farms powering these technologies – are rapidly reshaping communities and indirectly having a negative effect on some people’s wallets.

According to information from DataCenterMap.com, the U.S. now has over 4,000 data center and server farms (up and running or under construction) with more than a third concentrated in just three states: Virginia (667), Texas (417), and California (326).

Florida ranks fourth with 127 data centers, mostly around Miami (42), Tampa (23), Jacksonville (15), and Orlando (14). Large clusters of data centers also dot Atlanta, Chicago, Dallas, Des Moines, and Phoenix. Globally, there are slightly less than 10,000 data centers.

Companies like Amazon Web Services, Google, Meta, and Microsoft build and operate massive campuses equipped with state-of-the-art equipment to serve businesses, store data, and process AI requests.

For AI workloads, these are known as “hyperscale” operators, and the secret sauce that make AI server farms into literal super-computers: accelerators or custom chips like Nvidia’s Blackwell and DGX B200 (GPUs) or Google’s Cloud Tensor Processing Units (TPUs).

[WATCH: Your next boss might be a bot]

These specialized pieces of technology can perform not just millions or billions of operations per second – instead, they can take on quadrillions of operations all at once at mind-boggling speeds known as floating-point operations per second, or FLOPS.

AI data centers bring notable economic benefits to communities by creating hundreds – and sometimes thousands – of high-wage technical jobs as well as construction and local service opportunities. Property taxes offer the biggest upside for municipalities, sometimes funding a significant chunk of local revenues.

In Loudoun County, Virginia – nicknamed Data Center Alley – tax revenue from server farms contributed $1 billion to the county this year, and in 2024 added $16 billion in assessed value to its real property portfolio.

The influx of “AI dollars” funds schools, hospitals, emergency services, and libraries – all essential as core facilities and services for any thriving community.

But while AI server farms boost the local economy, America is discovering they make for bad neighbors.

The community cost to run a data center

Hyperscale data centers run thousands of servers 24 hours a day, 7 days a week. Ever hear your laptop or desktop computer’s hard drive or fan spinning up? Multiply that by 5,000. AI-powered server farms generate tremendous ambient noise. And when emergency operations occur, electricity from the grid is replaced by power from diesel generators – noisy, smelly, bad for the environment, diesel generators.

Beyond the noise, server farms guzzle enormous amounts of electricity and water to sustain normal operations.

According to the U.S. Energy Information Administration, the average household consumes about 10,500 kilowatt-hours of electricity per year. Accounting firm Deloitte estimates that a hyperscale data center – a specialized server farm with about 20,000 high-end servers for Generative‑AI queries, would consume roughly 1.14 terawatt‑hours of electricity per year – about as much as the electricity for 108,450 U.S. households.

Right now, Florida does not appear to have any active “large-load” data centers that require more than 100MW of electricity to operate.

Pew Research reports U.S. data centers accounted for about 4% of national electricity use in 2024 (183 terawatt-hours), up from 176TWh in 2023. Projections show consumption more than doubling by 2030, up 133% to 426 TWh. And in case you’re wondering, Deloitte also quotes an IEA report that says on average, a Generative AI–based prompt request consumes 10 to 100 times more electricity than a typical internet search query.

[WATCH: Amazon is implementing AI to downsize corporate ranks]

And because the data centers are taking up so much energy, companies have to find ways to generate more for everyone else. And that means costs are going up.

One Bloomberg analysis estimated that “electricity now costs as much as 267% more for a single month than it did five years ago in areas located near significant data center activity.”

Tech companies, however, are looking to trim power costs to the bone, which has revived interest in nuclear power.

Microsoft cut a deal with Constellation Energy to re-start Three Mile Island’s Generating Station Unit 1 – which shut down in 2019 – to power undisclosed data centers. TMI Unit 1 could be back online as early as 2028 producing enough energy to power 800,000 homes, though Microsoft would be its only customer.

And then there is water.

Hyperscale sites use vast amounts for cooling, far more than non-hyperscale data centers in two ways: on-site using evaporative cooling and to control humidity, and indirectly when power plants produce more electricity (those facilities rely on water as well to keep things cool).

Pew reports hyperscale sites annually consume 84% of the 17 billion gallons used by all U.S. data centers. Researchers forecast that consumption could climb as high as 33 billion gallons of water by 2028. Residents near some facilities in Georgia report lower water pressure, brown or contaminated tap water, and dry wells, sparking even more local opposition and, in a number of cases, legal action. Another concern for proponents of data centers is that water demands will put a strain on supplies in drought-prone regions like Arizona, Georgia, and Texas.

DeSantis’ ‘AI Bill of Rights’

​​Florida, with its 127 data centers fueling the AI boom, is now ground zero for pushback.

On Dec. 4, 2025, Governor Ron DeSantis proposed an “AI Bill of Rights” targeting hyperscale facilities to shield residents from hidden costs.

His stated impetus: the common practice of utilities negotiating discounted electricity and water rates for data centers and then shifting some operating costs to residential and small business ratepayers.

The DeSantis plan would, among other things, prohibit electric, gas, or water utility surcharges to subsidize hyperscale data centers. In proposing this consumer protection, DeSantis is directly addressing reports of preferential rates that shift part of the financial burden to residential communities and small businesses in favor of Big Tech.

The proposed legislation also bars tax subsidies for data center projects and would give local governments the power to outright reject AI data centers, including veto power over sites on agricultural land or greenbelt exemptions to protect farmland and water resources.

​The Florida governor says the new legislation would mandate Florida Department of Transportation (FDOT) reviews for noise, setbacks, and vegetative buffers near proposed facilities in response to complaints of constant mechanical hums disrupting neighborhoods (complaints that echo lawsuits in both Virginia and Minnesota).

“We’re not going to let a rush to build data centers harm our people,” DeSantis said at a news conference in The Villages, framing the rules as essential guardrails against what he called an “age of darkness and deceit.”

Responding to Trump’s executive order announcement on Monday, DeSantis pointed out on X.com that only Congress can preempt the states through legislation, not the president.

“The problem is that Congress hasn’t proposed any coherent regulatory scheme but instead just wanted to block states from doing anything for 10 years, which would be an AI amnesty,” DeSantis said, referring to an attempt to block state and local AI legislation as part of the One Big Beautiful Bill Act.

The Senate eventually stripped that provision in a near-unanimous vote before final passage.

​The proposal is expected to be put forth in the next legislative session and signals Florida’s pivot from economic incentives to stricter local control, potentially slowing the Sunshine State’s AI expansion amid national debates over jobs versus environmental strain.

Although the DeSantis proposal could become a national model for other states to follow, is it too little, too late?

Again, at well over a hundred data centers and more on the way, ​only three other states have more server farms than the Sunshine State. Is this good politics to truly protect the constituency, or another dead-on-arrival example of closing the barn door after the horses have already escaped?