You’re in the middle of watching your favorite show, or playing through a recent game and, without warning, the power flickers or goes out completely. It happens, and it’s a bummer. Worse yet, you never really know how long the power is going to be out. It could be quick — just a minor (if annoying) inconvenience — or it could take hours. The obvious culprit for these outages is a storm in your area, infrastructure concerns, seasonal temperature extremes, or local energy spikes that put a lot of strain on the grid. But with that last scenario, it turns out there might be a very specific cause behind those blackouts. In particular, an increase in AI usage could be to blame, and it might make your power bill go up, even if you don’t use it.
According to data from the U.S. Energy Information Administration, the average U.S. household paid 17.47 cents per kilowatt-hour in May 2025, compared to 16.41 cents in May 2024. That’s a 6.5% cost increase in just a year, which is, historically, a pretty big spike. It doesn’t just highlight rising costs, though; it shows that energy consumption is higher than ever, too. Part of the reason those costs are increasing is because everyone is using more energy. Of course, the next question is whether or not AI is truly to blame for higher consumption. After all, you don’t necessarily know from this information if AI is triggering electrical surges and outages, right? But spoiler alert: Not only is AI causing outages, it’s even making residential power worse, according to a report from Bloomberg.
How do we know AI is a problem?
According to experts on the topic such as Whisker Labs and DC Byte, data centers which also support AI usage and applications are putting a big strain on the electrical grid and its aging infrastructure. That is known thanks to information gathered using Ting devices, which are smart sensors that can monitor electrical quality in residential homes. They’re primarily meant to reduce or prevent electrical fires, but the sensors inside also measure something referred to as “bad harmonics.” Or, perhaps more accurately, it’s called total harmonic distortion, which causes reduced energy efficiency and can damage appliances and home electronics.
Bad harmonics occur when the normal flow of electricity is interrupted, resulting in electrical spikes — increases in voltage — and, conversely, dips. For reference, a good harmonic is essentially a continued and balanced wave of electrical voltage. Bloomberg’s report shows that the worst distortions happen within close proximity to major data centers: Over 75% are often within 50 miles, and that affects nearly 3.7 million Americans in total. Moreover, the worst distortions are happening in Northern Virginia, Chicago, and areas where there is a high concentration of data centers. Unlike the steady, predictable increase in demand caused by factors like population growth, this creation of new data centers can put near-instant pressure on the local infrastructure. Those power demands are suddenly exponentially higher. (All that said, the tech isn’t entirely bad; AI-powered topology optimization could be a cheap, easy way to supercharge America’s power grid.)
Why is power quality important?
In addition to damaging appliances and electronics, poor power quality can also cause malfunctions, overheating, and spark electrical fires, which is certainly a scary prospect. In other words, there is a serious concern about these outages, aside from people being without power — and for a lot of us, that’s already bad enough. It’s not just about the “haves” and “have nots,” but the potential consequences of decreasing power quality for everyone. An uptick in electrical fires, for example, is a very real possibility, and it’s a particularly concerning one at a time when wildfires are a huge concern in various areas across the country.
While devices like Ting do exist to help, imagine having to monitor the power quality in your home or apartment and limiting what can be plugged in, from appliances and media devices to lighting and beyond. Power rationing could also become a very real concern, and it’s something that there really is no precedent for, at least not at the moment. Power companies might have to do this to preserve the grid. But how would you decide what power goes where and how much?
Fortunately, there are some other solutions in play. In Virginia, for instance, data centers are being required to build dedicated substations and transformers to handle the increase in power and protect residential circuits. And startups like Exowatt are working to make modular systems dedicated to producing clean energy to run these data centers. Depending on how this all plays out, however, other long-term solutions may soon be necessary. It might be time to start tapping into unexpected energy sources that can be used as fuel, like seawater, urine, or plastic waste.
