Prepare yourself: U.S. electricity prices are likely headed upwards. When a PBS reporter quizzed an energy consultant about electricity costs, the goal was clearly to present both sides of the story. The initial question hinted that AI centers in an area might lower prices. The consultant confirmed this, explaining that because the electric industry's costs are largely fixed, spreading those costs over more customers (like AI facilities) could indeed reduce costs for everyone. Economics 101, right? But here's where it gets controversial... The consultant then added that where there isn't extra capacity, new demand would require expensive new equipment, driving prices up. This makes sense, but the electric grid is struggling after a period of low growth, and spare capacity is scarce. The interview questions should have been flipped. Shouldn't the question be: Isn't it true that AI will generally push up electricity prices? Yes, but in some cases, it might not. Back in the COVID era, we estimated the cost of modernizing the grid and replacing old equipment, concluding that decarbonization would cost consumers more than just modernization, largely due to the high capital cost of storage batteries, but either way (fossil-fired or renewable), the price of electricity would rise in real terms. (An executive summary can be found in our book, Financing the Future of Electricity.) Given that storage battery costs have fallen 30-50% since that analysis, we suspect that decarbonization would now no longer cost more than conventional power-generating fleet modernization. Our analysis, however, assumed flat sales, which was the picture at the time.
Let's break down the situation with a simplified analysis. Electricity costs can be divided into variable and fixed costs, roughly 50-50. Variable costs (mainly fuel) will increase with demand. Fixed costs (long-lived assets like power plants, including their cost of capital) will increase proportionally to the capital invested. This capital increase is for replacing old plants and building new ones to meet growing demand.
Now, here's where it gets tricky. The average U.S. electric plant is over 30 years old. Assuming a 3% load growth (potentially on the high end if all proposed demands materialize) and that old plants retire after 40 years. Considering that utility and construction costs have tripled in the last three decades, every 1% increase in demand requires a 3% increase in capital investment (because a new plant costs at least three times the depreciated value of the old one). The replacement situation is even worse because the replaced plant has largely been written off through depreciation, meaning utilities no longer record a big expense item and customers no longer pay for it. We believe the cost of the new plant easily exceeds 5 times the book value of the replaced plant. While this might be an underestimation, it's a conservative approach to get a reasonable estimate.
Okay, what's the bottom line? With a 3% growth in domestic electric demand, the revenue needed to cover all costs will have to rise at least 10% per year. Since sales will grow 3%, the revenue can be spread over 3% more kWh, so the price of electricity will increase 7% per year in real terms. Add inflation, and the number goes up. No fancy models needed. Think in terms of one significant figure. The real number is more likely to be higher than lower.
And this is the part most people miss... We're not even calculating the collateral damage to consumers caused by a sharp increase in electricity demand, such as increased demand for natural gas and water. That only adds to inflationary pressure and, in the case of natural gas, potential commodity price volatility.
In summary, our old model showed that power prices would have to rise faster than inflation, even with flat demand. The conclusion of our simple model, reiterated, shows that if the new demand for electricity materializes, electricity prices, uncontrolled, with power produced in the manner contemplated by the current administration and favored by the electric industry (mostly natural gas, nuclear, and some coal), will rise at a rate materially faster than the rate of inflation, for years to come. And the faster the rate of demand growth, the higher the electricity price hikes. So, from an electricity consumer's perspective, these new AI data centers are not your friends.
What do you think? Do you agree with this analysis, or do you see other factors at play? Share your thoughts in the comments!
