Electricity has been so widely available and so reliable for so long that we take it for granted. That is the problem. The reason it is problematic is because the way we produce, distribute and consume electricity is changing. And as I have pointed out in numerous previous posts, change is often difficult and expensive – two things no one likes.
Electrical service is so useful and so relatively inexpensive that virtually every household not only has it, but uses it without thinking twice. It is so reliable (more than 99.9% up-time in the U.S.) that it is a major news event when widespread outages occur. And despite the potential for death every time we use it, we are absolutely shocked (figuratively) whenever anyone is seriously shocked (literally).
But what if all of that changed? What if you had to cross your fingers and hope for the best whenever you came home and flipped on a light switch? Or what if you had to ration your use of the washing machine, oven or air conditioner because you couldn’t let your electric bill get out of hand? I’m being intentionally over-dramatic – very few households in the U.S. are likely to be impacted in that way. The future does contain changes, however, that will be unpopular and uncomfortable enough that people will rethink how they use electricity in their life and complain to their elected representatives that “something needs to be done.” My goal with this article is to figure out what that “something” is likely to be and how it will impact both households and cities.
How Electrical Systems Work
The more you learn about how electricity goes from generation to distribution to eventual consumption the more amazing the entire process seems. It is like someone juggling a thousand balls at the same time. The fact that the system works at all is due to some seriously clever human inventions, some well crafted industry standards, and to some government regulations that keep various stakeholder interests in balance.
To begin with, electrical system operators have to constantly keep power supply in sync with power demand. Electrical equipment and appliances are built to operate within a specific voltage and frequency range. In North America, the reference power frequency is 60 cycles per second (or 60 Hertz). Voltage and amperage can vary depending upon the needs of each particular user, but frequency is constant and bad things happen if the frequency varies by more than one percent or so. In fact, frequency is a primary indicator of the balance between supply and demand. If the frequency drops it means there is more demand than supply and more power needs to be generated. If the frequency increases, then the opposite is true.
The demand for electricity, of course, varies constantly throughout the day. The basic pattern is relatively predictable, but weather fluctuations, equipment malfunctions, and a myriad of random events and consumer decisions can alter the basic pattern minute-by-minute. Consequently, power system operators (and their computer systems) are making constant adjustments in real time. In addition, system operators are expected to balance supply and demand in the most cost efficient manner possible, which means buying as much power as possible from low-cost providers and using higher-cost providers only when absolutely necessary.
To understand how all this plays out, a little background information will be helpful. Some power plants, for example, are best at providing a relatively fixed amount of power more or less continuously. Nuclear, coal and hydro-power plants fall into this category, and are generally referred to as baseload generators because they are used to supply the base threshold of power that is required 24/7. The downside of these generators is that they take a long time to start up and they aren’t very good at responding to sudden spikes in demand.
Intermediate, or load-following generators, are used to supply power to match the ups and downs that are relatively predictable on a day-to-day basis. These plants are generally powered by natural gas turbines that can ramp up and down relatively economically. The final category of power producers are known as peaking plants, which are used for the relatively short-duration peaks of the daily demand cycle and for unexpected surges in demand. These plants typically use quick-response gas turbines that can respond in minutes to changing demand, or – increasingly in recent years – large scale battery storage facilities which can supply additional electricity in seconds. Peaking plants are an essential element of the system but the power they supply is expensive.
Utility-scale solar arrays and large wind farms don’t fit neatly into any of the traditional categories. They are generally used as baseload generators because the cost of the power they produce is generally lower than both nuclear and coal plants, but their fluctuating output makes integrating them with other power sources a bit challenging. Pairing wind and solar with battery storage can smooth out the fluctuations considerably which makes life easier for system operators, particularly as renewables like wind and solar provide a larger and larger share of total power supply.
To make things even more complicated, it is often cheaper for the utility company (and “greener” from an environmental standpoint) to manage demand during peak periods rather than buy expensive power from peaking plants. This “demand management” approach uses rates that fluctuate during the day (time-of-use rates) to make using electricity during peak periods expensive, thus incentivizing consumers to shift their consumption to off-peak times. Unfortunately, most people don’t like doing their cooking and laundry in the middle of the night so time-of-use rates can increase bills for some residential households.
Defining the Problem
Exploding demand. From 2005 to 2020, demand for electricity was almost flat (average increase of 0.1% per year). Improvements in energy efficiency were largely offsetting economic and demographic growth. [1] That has changed in recent years, however, as electrical demand set a new record in 2024 and again in 2025, and is now growing rapidly enough in many areas that electrical utilities are struggling to keep up.
Nearly four years ago I posted an article to this blog that talked about the “electrification of everything.” That trend is still continuing as battery powered electric vehicles replace vehicles with internal combustion engines, heat pumps replace natural gas furnaces and water heaters, induction stoves replace gas stoves, and on and on. The things I talked about in that post primarily affected residential demand, and while they still are trending upward they represent a relatively small portion of recent demand increases. While modest in terms of total demand, the trend is still significant because it means that demand is increasing everywhere, which means that the distribution grid will eventually need to be upgraded everywhere rather than in just a few places.
More relevant in recent years has been the proliferation of data centers, each containing thousands and thousands of computer servers. Data center construction is booming in order to meet the insatiable computing needs of artificial intelligence and cloud data storage. It is estimated that in 2024 U.S. data centers consumed approximately 4 percent of total electrical demand, an amount roughly equal to the demand for the entire country of Pakistan. Data center demand is expected to grow by 133% over the next five years [2] but some analysts consider those forecasts to be too low. The mania over artificial intelligence is leading to announcements for data center investments that are enormous in scale. If it all comes true, and that is a big if, the combined impact might bring the power grid to its knees.
Take for example, the $11 billion data campus that Amazon is building in northern Indiana known as Project Rainier. Currently seven data center buildings are operational, but a total of 30 buildings are planned. The full site will eventually draw roughly 2.2 gigawatts of electricity, equivalent to the power consumed by 1.6 million homes. [3] Every major player in the AI industry has similar projects either under construction or on the drawing board.
While there are data centers in every state, a third of the roughly 4,000 data centers in the U.S. are clustered in three states: Virginia (643), Texas (395) and California (319). In addition, there are significant clusters in Chicago, Phoenix, Atlanta, Columbus and Des Moines. Typically, an AI-focused “hyperscaler” data center contains at least 5,000 servers and consumes electricity equivalent to the demand from 100,000 homes. Newer designs currently under construction are expected to increase that demand by at least a factor of ten. Although there are advantages to clustering data centers together (e.g. network access), power demands are stretching the capabilities of local utilities to the point where proposed data centers are having difficulties getting power supply commitments and consequently are postponing construction or are looking at locations where power is more available.
Supply constraints. If demand is growing, won’t the market respond by building more power plants in order to increase supply? The answer is yes, but there is a significant timeline mismatch. Utility scale power plants, wind farms or solar panel arrays can’t simply be added to the grid willy-nilly without upsetting the delicate balance that keeps the grid functioning smoothly. Transmission lines, substations, transformers, switching gear, and grid control systems all need to be upgraded as the new power source comes online. All of that takes five years or more to get planned, permitted and built. The process is so time consuming that there is currently a queue of power generation projects waiting to get approval to connect to the grid. That sounds like a good problem to have, but many proposed projects never get built because of problems that are uncovered during the approval process. The current system is a roadblock to new power capacity, but it is a necessary evil.
The reality is that AI data center demand has ramped up quickly in just the past few years, but growing the power capacity of the grid is a much slower process. A new data center can be planned and built in a couple of years, but the matching power supply might take twice as long – or in some cases three or four times as long. [4] This is particularly true for some of the more exotic solutions that have been proposed such as geothermal or modular nuclear power plants.
Things have gotten desperate enough that old nuclear power plants are being brought out of mothballs. Microsoft recently signed a 20-year power purchase agreement with Constellation Energy to restart a reactor at Three Mile Island to power planned AI data centers. Three Mile Island, of course, was the site of America’s worst nuclear accident back in 1979. A second, undamaged reactor was eventually restarted and ran until 2019 when it was shut down because it was too costly. [5] The needs of AI have apparently changed the economic equation because costly power is better than no power at all.
Shaky infrastructure. The period of flat demand growth from 2005 to 2020 lulled utility companies into thinking that infrastructure upgrades were a relatively low priority. That has changed recently for three reasons. First, the combination of aging infrastructure, high winds (or other extreme weather) and drought conditions have caused enormous liability issues. Pacific Gas & Electric in California filed for bankruptcy in 2019 and pled guilty to 84 counts of involuntary manslaughter following the 2018 Camp fire. Hawaiian Electric has been hit by a dozen lawsuits over its alleged role in the wildfire that killed more than 100 people and burned the town of Lahaina to the ground. The Fitch rating agency warned that the company faces more than $3 Billion in potential liability claims. PacifiCorp and Xcel Energy are facing similar problems for the 2020 Labor Day fires in Oregon and the 2021 Marshall fire in Colorado. [6]
The average age of the country’s transmission lines is 40 years old, and roughly a quarter of the lines are more than 50 years old which is the typical intended lifespan. More than half of US transformers will reach their intended lifespans in the next five to ten years. Utilities are spending billions of dollars annually to remedy the problem, but supply line disruptions and tariffs make the work more expensive, and the scale of the problem makes the solution a decades long affair. Utility companies are finally taking extreme weather seriously but there is a lot of catching up to do.
Second, the electrical grid is having to supply power to the above-mentioned data centers which are 24/7 power hogs. Take, for example, the 2,250-acre data center campus currently being built by Meta in Louisiana. The project, named Hyperion, is being built in Richland Parish which is pretty much in the middle of nowhere – more than 100 miles from either Shreveport or Jackson, the only towns of any size in the area. Upon completion, it is estimated that the campus will consume three times as much electricity as the City of New Orleans. Consequently, the local utility company (Entergy) is spending $1.2 Billion to build a 100-mile, 500kV transmission line, along with eight substations, and eight 230kV transmission lines. In theory, Meta is covering the cost, but projects like this distract utility companies from addressing the underlying fragility of their electrical networks.
Finally, utility companies are being pushed to harden the electrical grid against possible cyber attacks. The grid has long been considered a soft target for terrorists or enemy nations because of outdated (and easily hacked) control systems and the distributed nature of the grid itself. Steps are being taken, but again, this is likely a decades-long effort.
Rising prices. So let’s quickly recap. Electrical demand is rising, supply is constrained by long approval times, and the distribution network is aging and fragile. Should we be surprised that prices are rising? Unless you completely slept through your economics class, the answer, of course, is no. According to the St. Louis Federal Reserve, during the six years from 2014 through early 2020, the average US City electricity price per kilowatt hour was basically flat (actually declining on an inflation adjusted basis). During the five years prior to that the average price rose a total of just over six percent. But in the five and a half years since early 2020 (through August of 2025) the average price has gone up over 40 percent.
To be fair, this increase is just a third more than the increase in the Consumer Price Index, so it is bad but not disastrous. Coming on the heels of a ten year period of hardly any increase, however, it seems much worse. Now in addition to the rapidly rising costs of household essentials like food, medical care, and housing, working class families have to worry about the electrical bill being beyond their budget.
Interestingly, in the states with the highest average increases, the cost of power generation has declined slightly. It is the cost of electrical transmission (long distance) and distribution (short distance) that has been rising faster than inflation. [7] The decline in power generation costs coincides with a shift away from relatively expensive coal power plants to less expensive natural gas and renewable sources. Coal produces just a third of the US electric power that it did twenty years ago and our pocketbooks and lungs are better off as a result.
However, skyrocketing demand from data centers may change all of that. Power generation costs have ticked back up over the past two years and are likely to continue rising. As noted earlier, power providers are so desperate for new capacity that old nuclear reactors are being brought back to life and coal plants slated for retirement are being kept operational. These are not low cost sources for electricity, so overall rates will increase. Fortunately, the vast majority of new power generating capacity will come from solar and wind – often supplemented by large scale battery storage – since renewables produce relatively cheap power and can be brought online faster than other sources.
The final point to understand is that electricity demand and price increases are not evenly distributed across the country, nor is there even a clear pattern. The ten states with the lowest average residential rates (starting with the lowest) are Nevada (11.95 ¢/kwh), Louisiana, Idaho, Tennessee, Kentucky, North Dakota, Arkansas, Washington, Nebraska and Mississippi (13.97 ¢/kwh). The ten states where residential rates have gone up the most (Sept 2024 - Sept 2025, starting with the worst) are New Jersey (21.1%), Illinois, Indiana, Pennsylvania, Maryland, Florida, Georgia, Ohio, New Hampshire, and Washington (11.3%). [8]
Helpful Trends
The gloom-and-doom statistics above may lead you to think that we are going to suffer major grid disruptions and pay more each month for the privilege. While that is a possibility, it isn’t very likely for a variety of reasons. To begin with, the electrical grid is run by people who are smart (and cautious), and the system has a lot of built-in redundancy (even if much of it is getting outdated). Yes, we are likely to pay more and there may be an increase in localized outages or brown-outs, but there are several countervailing factors that I think will save us from anything catastrophic.
Bursting the AI bubble. The list of data center construction projects that have been announced by various tech companies is very long and their potential power demand is enormous. Oddly, however, that potential impact isn’t showing up in the energy futures market. For some reason, the “smart money” isn’t buying the story that electrical demand is going to outstrip supply because of the data centers powering artificial intelligence.
Every tech company, it seems, is jumping on the AI bandwagon and to be taken seriously you apparently have to announce expensive initiatives like billion dollar data centers. It is possible that many of these announcements are from AI unicorns that will be bankrupt before any ground is broken, or are gross exaggerations designed to garner press coverage. It is too soon to know for sure, but demand increases might be modest enough to be handled without threatening grid stability.
Soaring solar. Nearly all forms of electrical power production have gotten less costly over time due to technological improvements and manufacturing efficiencies. But there is one form of electrical production that has outshone the others, and that is photo-voltaic solar. Using data from industry analyst Lazard for the Levelized Cost of Energy – total construction, maintenance and fuel costs divided by total energy production over the life of the facility – shows that the cost per megawatt-hour for utility scale solar has fallen by a remarkable 88 percent over the past 15 years. [9] The energy produced by each solar cell has nearly doubled during that time and production costs have plummeted as manufacturing volume has increased.
.png)
Solar power has been the great “democratizer” of energy – simple and cost-effective enough that individual households and companies can generate their own power and store it for use at any time when paired with batteries. This “behind-the-meter” electricity saves the individual user a significant amount of money but it disrupts the economic model behind the electric utility industry. Utility companies generally spread operational costs across their customers based on total consumption, so if private solar causes demand to drop it can result in rates going up for everyone else.
Not every location is ideal for solar facilities, but in the best locations solar now has the second lowest cost behind on-shore wind farms. Ongoing innovations are likely to keep the cost per watt falling faster for solar than for other competing options. Even less than ideal locations are generally cost competitive now and may have outright cost advantages within the next five years.
Many solar projects are now paired with large battery storage installations to make it easier for grid operators to dispatch electricity around the clock as needed. As with solar cells, battery efficiency and the cost of manufacturing have improved dramatically over the past ten years. New battery configurations and chemical compositions are being announced almost weekly which will improve performance, reduce costs and lower environmental impacts. Battery storage facilities may eventually account for 20 to 30 percent of electricity supply during periods of high demand.
Decentralizing the grid. One of the advantages of solar power production and battery storage is that moderately sized facilities can be distributed across the grid as opposed to traditional power plants which are centralized in just a few locations. This characteristic is not only reducing transmission costs, but it is allowing the overall electrical grid to be subdivided into microgrids which can be isolated from the main grid during times of instability or weather-related disaster. Microgrids with their own battery storage and/or solar arrays (or other generating source) are not likely to be self-sufficient for more than a day or two, but that might be enough to significantly lessen the spread of major outages.
Microgrids are currently being created primarily for commercial and industrial facilities where uninterrupted 24/7 operations are critical (e.g. medical campuses, airports, logistics hubs, military bases), but they are likely to spread gradually to more mundane parts of the city. Distributed power generation and storage will eventually simplify transmission and distribution needs, and reduce demand for expensive peaking plants.
The Bottom Line
To summarize, the electrical grid is likely to be stressed over the next few years in several different ways:
Total electrical demand is likely to continue rising for the foreseeable future, for both residential and commercial customers;
Utility companies are likely to continue investing heavily in infrastructure upgrades for capacity reasons and to replace aging equipment;
Extreme weather events and changing climate patterns will continue to test grid resiliency; and
Power generation decisions are likely to be influenced by not only spiking demand near data centers, but also by shifts in political policies at the federal level.
All of this, in my opinion, is likely to cause rates to continue to rise over much of the country. Here are a couple of ways this is playing out. First, the Colorado River basin – which is lined with hydroelectric dams – is in the midst of an extended drought. The “full pool” elevation for Lake Mead, for example, is approximately 1,220 feet but it is currently at around 1,060 feet. The turbines at Hoover Dam are continuing to produce power, but at a reduced rate which means that many of the utility companies that depend on that electricity are having to purchase power from the open market which is generally more expensive. [10] If Lake Mead water levels drop to roughly 1,000 feet then electricity can no longer be generated at all, and while that is unlikely in the near term it is a possibility if the drought continues.
This may seem like an isolated issue, but extreme weather events – or even the threat of extreme weather – affects the electrical grid on a regular basis. In Asheville, North Carolina, for example, it took 10 days to restore power to the majority of customers following the torrential rains from Helene in September of 2024. Restoring power to more isolated customers in the surrounding area took weeks and even months. Xcel Energy in Colorado cut power to roughly 100,000 customers just recently due to the threat of wildfires stemming from drought and high winds. The components of the electrical grid are uniquely exposed to extreme weather which multiplies the risk associated with aging infrastructure.
Second, spiking demand and political pressures at the federal level may push some energy producers into making decisions that are questionable from a long-term economic perspective. I doubt anyone is stupid enough to build a new coal powered generating plant despite the administration’s “mine, baby mine” attitude, but there is likely to be a shift toward more natural gas power plants despite the relatively high cost of the electricity they produce. Even though natural gas prices have historically been volatile and are currently at relatively high levels, the Trump administration considers natural gas plants as “dependable” and a good match for the 24/7 demand from data centers. In my opinion, however, power from the combination of solar, wind and battery storage is the most economical long-term strategy. Unfortunately, utility company profits are based largely on a pre-set rate of return applied to all system assets, so building inefficient (and polluting) power plants is fine as long as they can be justified in some other way (e.g. “dependable”).
Finally, there is the not insignificant risk that technological change will either reduce the demand for AI data centers or will reduce the power demands of AI data centers. After all, the history of computer performance over the past several decades has shown exponential improvements in calculations per watt of energy consumed. In fact, the next-big-thing in computing is known as a quantum computer, and while it might be a decade or more until quantum computers are commonplace, when they arrive they will undoubtedly be used for things like artificial intelligence and their performance will be orders of magnitude more efficient than today’s best CPUs. This raises the possibility that the data centers which need huge amounts of electricity in the near-term will suddenly need substantially less at some point in the future – leaving us over supplied with electrical capacity. Unlike most businesses, however, utility companies can never lose money because they overestimated demand. They are guaranteed a profit which means that rates will go up because total costs will need to be spread across a smaller base of power consumption.
In short, I don’t see electric bills doing anything but going up in the foreseeable future. The long term trend of declining power production costs seems likely to continue – although it may be temporarily interrupted by the short term scramble for power at any cost – but it will be overwhelmed by increases in transmission, distribution and grid resiliency costs. We are so dependent upon electricity that we have no option other than paying the bill, but it is one more straw on the camel’s back that is the household budget. I worry about how much more working class families can take before our economic system starts to break. Changes to the electrical grid are just another example of the economic squeeze that families are facing. As a society, we continue to paint ourselves into a corner, one incremental decision at a time, and I don’t see an easy way out.
Notes:
1. “After more than a decade of little change, U.S. electrical consumption is rising again”; U.S. Energy Information Administration; May 2025; https://www.eia.gov/todayinenergy/detail.php?id=65264
2. Rebecca Leppert; “What we know about energy use at U.S. data centers amid the AI boom”; Pew Research Center; October 2025; https://www.pewresearch.org/short-reads/2025/10/24/what-we-know-about-energy-use-at-us-data-centers-amid-the-ai-boom/
3. MacKenzie Sigalos; “Amazon opens $11 billion AI data center in rural Indiana as rivals race to break ground”; CNBC; October 2025; https://www.cnbc.com/2025/10/29/amazon-opens-11-billion-ai-data-center-project-rainier-in-indiana.html
4. Martin Stansbury, et al; “Can US infrastructure keep up with the AI economy?”; Deloitte Touche; June 2025; https://www.deloitte.com/us/en/insights/industry/power-and-utilities/data-center-infrastructure-artificial-intelligence.html
5. Kris Maher and Jeanne Whalen; “Three Mile Island’s Nuclear Revival Pits Those Who Fled Against Job Seekers”; The Wall Street Journal; December 2025; https://www.wsj.com/business/energy-oil/three-mile-islands-nuclear-revival-pits-those-who-fled-against-job-seekers-2758e115?gaa_at=eafs&gaa_n=AWEtsqc5EYPgpOJS0JxWkorwIudTw0ivXgcmCYu6RyHmprKhC1UB6JVgFpP5s9qGFiw%3D&gaa_ts=692f6098&gaa_sig=ue2UXdB7U9wD2_Tz9UCgKqLwWaypCRIyF-igVa5GHxvFUjWsHHxfgxThCf_sN5wxktvjaIaWVn0BaSMgRmKQpA%3D%3D
6. Spencer Kimball and Gabriel Cortes; “Electric utilities face billions in wildfire liability with aging power lines risking another catastrophe”; CNBC; August 2023; https://www.cnbc.com/2023/08/28/wildfire-risk-electric-utilities-face-billions-in-liability-with-aging-lines.html
7. Jesse Buchsbaum and Jenya Kahn-Lang; “What’s Happening to Electricity Affordability? In Five Charts”; Resources; October 2025; https://www.resources.org/archives/whats-happening-to-electricity-affordability-in-five-charts/
8. Caitlin Ritchie; “Electricity Rates by State”; Choose Energy; Dec 2025; https://www.chooseenergy.com/electricity-rates-by-state/
9. Max Roser; “Why did renewables become so cheap so fast?”; Our World In Data; April 2025; https://ourworldindata.org/cheap-renewables-growth
10. Daniel Rothberg; “Many miles from Lake Mead, rural electric utilities struggle with Colorado River shortage”; The Nevada Independent; October 2022; https://thenevadaindependent.com/article/many-miles-from-lake-mead-rural-electric-utilities-struggle-with-colorado-river-shortage
