On data centers and energy “affordability”

Start with how a regulated price is set. The price per kilowatt-hour is essentially the total cost of generating and delivering power, including a return on capital that regulators deem reasonable, divided by the total kilowatt-hours delivered. A data center is an enormous addition to the denominator. Whether it raises anyone’s price therefore depends on whether the numerator rises by a larger percentage, and that turns on whether the system has slack. Where there is spare capacity to generate and deliver, adding load should spread costs and lower them for everyone else. Where the grid is already at capacity at peak, which he says describes many regions today, it should not, because new infrastructure costs far more than the existing, partly depreciated infrastructure it adds to. He supports the cost half of that with material reproduced from his own recent post rather than argued here: transformer, switchgear and aluminum conductor prices spiked during the pandemic and have stayed well above general inflation; combined cycle gas plants cost materially more than a few years ago; and solar and wind costs have been rising and look likely to keep rising through the decade.

That is the theory, and he then argues it does not describe practice, because the pricing formula admits exceptions and data centers are exceptional customers. They are exceptionally large, they have an exceptionally high willingness to pay, and the policymakers courting them for economic development are exceptionally motivated not to burden households, all the more so as opposition to data centers rises. What he says he has actually watched is utilities and regulators going to great lengths in the other direction, through individually negotiated agreements that make the operator cover all the incremental cost of serving the facility and then some. That reduces the numerator while the denominator still grows. Entergy’s published “Fair Share Plus” principles are his example, and an outside study is his evidence: a recent paper found essentially no correlation between data center development and household prices, and a modest downward causal effect over 2019 to 2024.

Then comes the turn the post is built around, and it is a distinction between levels rather than a reversal. There are over two hundred investor-owned utilities in the United States plus hundreds of municipal and cooperative ones, and every one of them can be handling its local data centers well while the aggregate still drives prices up. His diagnosis is that the problem is not really data centers at all but that the supply side of the power market is inelastic, meaning it cannot expand quickly whatever the price signal. The reasons he lists are specialized supply chains and labor, the exhaustion of the best solar and wind sites alongside tax credits expiring over the following four years, utilities in North America and Europe grown accustomed to practically zero growth since the Great Recession and short of the institutional capacity for it, and the large share of capital spending that is non-negotiable replacement and storm hardening rather than added capacity. So the individual project is exonerated and the category is not, and he expects bills to rise further. He notes that financial markets appear to be reaching the same conclusion, quoting a ratings agency at length.

On affordability he reframes rather than answers directly. Measured against what people would be willing to pay, electricity is one of the best bargains available: he walks through the kilowatt-hour as roughly the work of an oversized horse for an hour, except pointable at anything from an air conditioner to a laptop, and says it may be second only to water in consumer surplus. Its share of the average household budget has fallen substantially over fifteen years and remains low even as it ticks back up. He then refuses to rest on that, saying he does not want to be glib: in a society this unequal the average is the wrong statistic, and for lower-middle-income families electricity is already a meaningful expense. He adds two further reasons to want cheap power, one of which he immediately qualifies. Electricity-intensive industry matters for competitiveness and security, though he grants that power cost does not tend to be a meaningful driver of competitiveness in high- precision manufacturing. The other he does not qualify: high prices deter electrification, which he takes to be the most promising decarbonization path across ground transport, heating and cooling, industrial heat and more.