The bill arrived as expected, on time, enclosed in a standard envelope. But it wasn’t the expense Athena Stine had budgeted for.
When Athena Stine, 25, and her fiancé moved into their Williston home three years ago, their first utility bill was $167. Last month, it was $458, nearly triple the previous amount. Williston Public Works, the city’s municipally owned utility company, bundles services into one bill and itemizes specific charges.
Stine, a human resources assistant, has seen her electricity bill climb nearly 17% since the beginning of summer, when energy usage typically peaks. She paid $255 in July and $223 in June, according to her expense tracker. The average American household spent a monthly $186 for electricity this summer, The Wall Street Journal recently reported.
It’s becoming increasingly challenging to manage her finances as the expenses pile up, Stine said.
“My fiancé was out of work for six months, so we were really struggling,” Stine said.
Retail electricity prices have risen faster than the rate of inflation since 2022, according to federal data. They’re poised to climb more rapidly amid growing demand for energy — in part, to power the rise of artificial intelligence.
That may soon put an extra squeeze on rate payers in Florida.
How hungry is AI?
Leading American tech companies are plowing billions of dollars into AI research. Much of their capital is flowing into the buildout of data centers, facilities that house the technology’s computing infrastructure.
Leading chatbots are trained in massive data centers that require, on the upper end, 100 megawatts of power annually, comparable to the energy needed to power tens of thousands of homes, according to CNBC.
Once trained, chatbots generate answers to user queries, drawing upon learned patterns, requiring more processing power than a traditional Google search.
Surging energy demands will stress power lines, threatening energy resilience, said Mark A. Jamison, who directs UF’s Public Utility Research Center.
“You just will have brownouts, blackouts, whatever it might be, from overloading the system,” Jamison said. “It's kind of like in some sense if you overload a highway, everything grinds to a halt.”
Mounting pressure on the grid will require costly infrastructure updates, he added.
As in other parts of the country, Florida is contending with an aging grid system. Extreme heat and recent hurricanes have exacerbated the problem.
Florida Power & Light, the state’s largest utility provider, recently justified a proposed rate hike on utilities by citing plans to invest in grid updates and other modernization projects.
FP&L filed a petition in February to raise its electricity rates by nearly $10 billion over a four-year period. The initial figure has since been scaled back about 30%. The original proposal would have marked the largest utility rate hike in U.S. history, according to the nonprofit Food & Water Watch.
The case is still under review with state regulators.
It’s challenging to assess the direct impact of data center expansion on energy prices — especially because tech companies disclose little about their consumption.
But several consumer advocacy groups in Florida have linked FP&L’s proposal to the cost of servicing the company’s tech customers.
Stakeholders push back
Florida’s AARP maintains that consumers, especially those on fixed incomes, should not be expected to shoulder the costs associated with data center demands.
“When we have a public service commission that does not look out for the best interest of the customers but rather the utility company itself, there is a problem,” Zayne Smith, senior director of advocacy with AARP Florida, recently told a utility trade publication.
Floridians have already absorbed fuel surcharges and storm recovery costs in recent years.
Ted Kury, who directs energy studies at UF’s Public Utility Research Center, said the FP&L proposal would affect customers at the same rate, regardless of their proximity to a data center.
Duke Energy Florida just filed a petition with the state’s public service commission to establish a new rate for data centers, he added. If approved, the plan would not change its current tariffs for any other customers.
Kury also pointed to issues with transparency in data center construction.
“Data centers have been signing contracts directly with electricity service providers, and in some cases, it's been very difficult for the regulators to evaluate those contracts in the name of trade secrets,” Kury said.
The closest data center hub to north central Florida is in Jacksonville. UF owns and operates HiPerGator, an AI supercomputer.
As data centers move closer to Gainesville, the city’s regional utility provider will likely need to pass along the costs of accommodating the energy demands to residential customers.
Municipal utilities often offer customers lower rates than investor-owned utilities, like FP&L. But the effects of rate hikes are felt more locally.
“If GRU incurs additional costs to serve a data center in Gainesville, then those costs will be borne by the people in Gainesville,” Kury said.
Calls for sustainable solutions
American AI companies may have overestimated the computing power and electricity their chatbots require. The Chinese company DeepSeek raised questions earlier this year about the amount of energy tools like ChatGPT demand.
DeepSeek claimed it had developed its chatbot, with capabilities comparable to those of leading U.S. rivals, using fewer power-hungry graphics processing units, or GPUs.
Training chatbots currently requires far more energy than servicing individual queries.
The global investment bank Goldman Sachs estimates the U.S. will need to generate approximately 47 extra gigawatts of power by 2030 to support data center growth. One gigawatt is equivalent to about 100 million LED light bulbs, according to the U.S. Department of Energy.
So far, the DeepSeek disruption has not meaningfully curbed energy consumption among U.S.-based AI heavyweights, according to the Brookings Institution, a non-profit research organization.
But it has spurred more interest in energy-efficient AI development, said Juan Claudio Nino, a UF professor in the materials science and engineering department.
The big AI players are still jockeying to develop larger, more efficient hardware and algorithm implementation, Nino said. Yet they’re increasingly embracing technologies that exhibit high performance while consuming less energy.
The DOE published a road map last year for scaling AI with efficiency targets in mind.
For now, people like 37-year-old Johnathon Hayes doubt they’ll be able to afford any extra costs.
Last month, his Duke Electricity bill was $442.
Hayes, a medically retired veteran in Homosassa, said he must stretch his monthly government checks to cover his expenses.
“I can’t work,” Hayes said. “I definitely have to account for every dollar that comes in and goes out.”