U.S. President Donald Trump famously likes to build big things — or at least have others build big things to which he can attach his name. True to form, his latest solution to the gathering crisis on the largest U.S. power grid involves constructing what would be the biggest power plant in the country and among the biggest gas-fired projects in the world. The associated White House fact sheet led with the gargantuan aspect.

As a sheet of facts, it was light on them. Along with the plant being pitched as part of Japan’s supposed $550 billion investment pledge to avoid higher tariffs, which themselves just got kiboshed by the Supreme Court, that suggests this project may be more imagined than real. Also, neither the regional grid operator nor regulators in Ohio were seemingly aware of plans for a 9.2-gigawatt plant that alone would boost the state’s power output by more than a third, which adds to the general blurriness.

But there was one useful factoid: a $33 billion price tag. Useful in the sense of demonstrating the inflation problem embedded in power bills.

The headline figure implies a cost of nearly $3,600 per kilowatt of capacity (Note: There is no breakdown of the $33 billion price tag). Combined-cycle gas turbines cost an average of just over $1,000 per kW in 2023, according to Bloomberg NEF, and just over $2,000 in 2025. Demand for turbines has surged alongside demand forecasts for electricity, linked to the proliferation of data centers chasing artificial intelligence. The backlog for new turbines has stretched to four years or more and an indicative price above $3,000 would represent a new level in surge pricing.

As it was, the cost advantage that natural gas enjoyed relative to other types of generation had been eroding. The levelized cost of electricity — which measures the cost per unit of energy of building and running a new power plant across its lifetime — for gas-fired power in the U.S. jumped by 48% in the second half of 2025, year over year, according to Bloomberg NEF’s latest analysis.

Using my own assumptions for fuel costs and the discount rate, among other things, the new Ohio plant’s levelized cost comes out at about $75 to $80 per megawatt-hour. That is well above prevailing average futures for the PJM grid of less than $60. Moreover, add in transmission fees of around $15 to $20 per MWh, and the all-in cost of delivered power approaches $100, similar to the estimated prices that Big Tech has paid for supply contracts with several nuclear plants in the region. (Of course, unlike the nuclear reactors, a gas plant also emits carbon dioxide, which in this case would equate to another $17.50 per MWh if it were priced at $50 per ton.)

Leaving aside the likelihood of this apparently Softbank-led, Japanese-funded gas-fired behemoth actually springing up in Ohio, those economics are a clanging alarm bell for Trump, Midwestern residents and Big Tech alike. To date, the majority of inflation in utility bills relates to the capitalized costs of building distribution networks, not generation. A big pickup in the latter would compound the problem.

If the demand for powering artificial intelligence really is set to spiral, then building more generation, including higher-priced gas-plants and even new nuclear reactors, is unavoidable. But the underlying message of Trump’s big beautiful burner is that the old model of just building giant power plants to meet whatever new peak demand forecast we dream up is unsustainable — economically, environmentally and politically.

While hyperscalers clearly are willing to pay up for power, locking in large slugs of expensive, carbon-emitting new supply for decades presents a risk to both them and ordinary ratepayers, especially if AI demand turns out to be less than advertised. As it is, new supply commitments reported by utilities are running far ahead of consensus data center demand forecasts, especially through the end of this decade, according to Andy DeVries of CreditSights.

This inherent uncertainty suggests the smarter option would be to hedge with a more modular approach favoring smaller gas plants as well as faster-to-build renewables and batteries; build smaller and then see, rather than big and just hope.

Moreover, when gas power is starting to be priced like nuclear power, data center developers and utilities should be thinking more creatively about demand management — avoiding using grid power for a handful of hours per year in order to shave the peaks in power consumption down rather than simply chasing those peaks with ever more expensive and underutilized capacity. "Uberizing" the grid in this way means using existing infrastructure more productively, as opposed to simply building, building, building — whatever the president’s proclivities.

© The Japan Times