“Landmark June 2026 study found DCs modestly reduced US electricity prices between 2015 and 2024.” – Love ’em or hate ’em; data centers are not the source of electricity inflation
This study was published on June 18, authored by the Electric Power Research Institute (EPRI), and entitled “Have Data Centers Raised Your Electric Bill? Causal Evidence from the United States.” A 4-page fact sheet can be found here, and the 45-page academic study here.
Statistics
“Observing the capacity expansion of data centers in particular, there does not appear to be a correlation with rising average prices at the state level from 2019 to 2024. However, these correlations do not establish causality due to omitted variable bias from endogenous selection into states, which could flow in either direction. In this paper, we estimate the price changes caused by data centers over this period by deploying instrumental variables to isolate a plausibly exogenous source of variation, thus overcoming unobserved confounding drivers of price changes and data center construction.”
In Minnesota, the average electricity price was largely flat from 1983 to 2002, growing only 4.5% from 5.55 to 5.80 cents per kWh. From 2003 to 2018, it exploded 79% from 6.01 to 10.37, and then grew a further 20% of 10.33 to 12.35 from 2019 to 2024.
“It seems plausible a priori that data center developers would be incentivized to locate in places where future electricity prices are expected to be low even with load growth, which would cause negative bias in the ordinary least squares (OLS) estimate. Alternatively, we can tell a speculative story that leads to bias upwards: utilities that clear interconnection queues faster have higher capacity to update their infrastructure; having more capacity to build is costly; and data centers prefer to locate in places where there is quick access to power, which may be correlated with higher prices. These are hypotheses; there may be other plausible stories. An IV approach is required to adjudicate between hypotheses.”
Instrumental Variables (IV) estimation is a statistical tool for trying to deduce a cause-and-effect relationship between two variables. OLS is seeking correlation, not causation as such.
“For the instrumental variable, we use the total length of the 1947 Eisenhower Interstate Highway Plan within each state, which we argue is exogenous to electricity prices conditional on GDP and population. The relevancy of the instrument comes from the observation that terrestrial fiber-optic infrastructure is typically located along interstate highway corridors.”
This seems reasonable given that, according to Yahoo!Finance in May 2026, “optical fiber cables held the largest share of 71.8% of the data center cable market in 2025.”
Economics
“How can higher demand lead to lower prices? Does the law of supply and demand not apply? A key difference between the power sector and other commodity markets is that due to the natural monopoly of distribution networks and high fixed costs throughout the system, retail prices are a function, at least in part, of average costs and not marginal costs. If the incremental cost of meeting demand is smaller than the status quo mix of average costs and marginal costs that are passed through to customers, then new demand mechanically lowers average costs.”
By definition, a natural monopoly has decreasing average costs at the current, and foreseeable future, demand levels. And total costs lower than two or more firms combined.
“Consider first a setting in which the physical assets of the power system are held fixed. As demand rises, grid operators dispatch electricity from increasingly expensive generators, which increases wholesale electricity prices as the law of supply and demand suggests. These short-run variable costs are passed through to retail customers via wholesale energy prices or similar mechanisms, but unlike in most commodity markets, retail tariffs must also recover the fixed costs of generation, transmission, and distribution capital. A bump in demand can allow for higher utilization of the existing generation and distribution network. Provided the bump does not violate capacity constraints, this leads to fixed costs that are spread across more kilowatt-hours (kWh), working in the opposite direction as the dispatch effect and pushing average costs down.”
The three major electricity companies in Minnesota, whether natural or not, are franchise monopolies regulated by the Public Utilities Commission (PUC). The PUC process determines whether lower costs translate into lower prices or not.
“If, instead, we allow the power system to change, and if higher demand is not within the expected distribution of demand but a new long-lived demand shift—such as new data centers, electric cars, and heat pumps—then the system will invest in new capital equipment to service this new load, shifting the dispatch curve outward and lowering marginal variable cost. Due to secular technology cost declines and efficiency gains, the incremental mix of new generation assets tends to have a lower average cost than the embedded incumbent asset mix, thus lowering the amortized capital cost of generation per kWh.“
Natural monopolies, in contestable markets, have incentives for cost efficiency and technology investment. Franchise monopolies, in regulated markets, do not. The latter is why, for example Xcel Energy, have fully embraced expensive and unreliable wind and solar.
Findings
“While ratepayers fear new load will raise prices, the historical record is consistent with the opposite. From 2014 to 2020 demand fell by 0.3% year over year while average prices grew at a rate of 0.2%, well below inflation during this period. From 2021 to 2024 demand grew on average a modest 1.5% year over year while electricity rates grew at 5.2%, in line with the average annual increase in core consumer prices of 5.8% meaning inflation adjusted electricity rates were flat on average. These recent rate increases have not been uniform across states but concentrated in the Northeast and West Coast.”
Minnesota saw annual electricity inflation from 2014 to 2020 of 1.4%, and then 4.7% from 2021 to 2024, compared to general inflation of 1.3% and 3.3% respectively for those two periods.
“We estimate for every 10% increase in data center capacity average residential retail prices fell by approximately 0.4% on average. The result is interpreted thus: from 2019 to 2024 the average residential customer lived in a state where data center capacity grew 160%, which caused their rates to fall by 6%.“
From 2019 to 2024, average annual residential electricity rates in Minnesota inflated by 3.1%, whilst data center capacity only grew by 13% from 38 to 43 MW. It has since jumped up 616% to 308 MW in 2026.
“A natural concern is that data centers, classified as either industrial or commercial customers depending on the jurisdiction, might lower overall system costs while shifting costs to residential ratepayers. Industrial and commercial customers typically see lower rates compared to residential households, so it is plausible that residential customers are subsidizing the incremental fixed costs of these lower-rate customers. The estimate can be interpreted thus: a doubling of data center capacity causes residential retail prices to fall by 3.5% for a fixed level of residential demand. These results suggest that data centers did not increase residential rates on average.”
In contrast, Minnesota commercial and industrial electricity rates increased 2.9% and 3.6% per annum respectively from 2019 to 2024, compared to 3.1% for residential.
Conclusion
“This study estimates a modest causal impact of data center load growth on decreasing residential prices. This is supported by descriptive statistics and a theoretic model that indicate economies of scale of power systems. Our findings are not without caveats, which we place into three categories: (1) supply constraints, (2) durability of demand growth, and (3) fuel costs.”
This EPRI study is a very important contribution to debunking the myth that data centers are the drivers of higher electricity prices. In addition, data centers are providing a public service by both exposing real causes, like renewables, and promoting real solutions, like nukes.