How a 141-Year-Old Brick Plant Reveals Shifting Real Asset Prices in the AI Era
In Sugarcreek, Ohio, there is a company called Belden Brick. It is a 141-year-old brick plant with no connection whatsoever to artificial intelligence. This summer, Reuters reported on the company's electricity bill. A "capacity charge" line item that had been $1,600 a month had ballooned to $12,000. The power bill rose nearly 90% in a single year. The company raised brick prices by 4%, but profit fell anyway. Brad Belden, the company's president, said the capacity charge "jumped off the bill" at him.
This did not happen because the company used more electricity. What rose was the price of the right to use electricity. And that price was set by parties the company has never once done business with: data centers going up hundreds of miles away, or data centers that have not even broken ground yet.
For decades in the U.S., electricity was simply there. Build a factory, and the utility ran a line to it. The bill was a monthly expense. That is no longer the case. The right to use electricity now carries a price, and that price is being reflected in the value of land, factories and buildings. Electricity is turning from a cost into an asset.
A Power Auction Changed the Bill
What changed Belden Brick's bill was an auction. PJM, which operates the grid across 13 states in the eastern and midwestern U.S., holds a "capacity auction" each year. The auction sets in advance what power plants are paid to reserve generating capacity needed years ahead. The clearing price jumped from $28.92 per megawatt-day for the 2024/2025 delivery year to $269.92 the following year, a 9.3-fold increase in a single year. The three auctions that followed all ended at the ceiling price approved by regulators, and the most recent two failed to secure the capacity needed even at that price. That had never happened in PJM's history.
PJM's independent market monitor calculated where these costs came from. Of the $63.6 billion in capacity costs generated across the four most recent auctions, $29.4 billion, or 46%, stemmed from data center load. In the 2027/2028 auction, $6.2 billion of the $6.5 billion attributed to data centers came from projected demand at facilities that have not yet been built. Existing factories are paying for facilities that do not exist.
This is not someone else's story. On bills for sites I help operate, the capacity charge line has more than doubled over the past two years. We did not use more electricity. Contracted load stayed the same. What changed was the price attached to each kilowatt of that load.
The bluntest testimony came from an aluminum company. Molly Beerman, chief financial officer of Alcoa, told an industry event last September that the company is now competing for power with Amazon and Microsoft, which are willing to pay more than $100 per megawatt-hour.
Transmission Lines Set Land Prices
Real estate has a familiar maxim: "location, location, location." Where a property sits is what matters. But in U.S. industrial real estate, that rule is becoming "power, power, power."
According to an August analysis by real estate data firm Altus Group, the median transaction price for data center sites in Northern Virginia, the world's largest data center cluster, is $2.8 million per acre. In markets where utilities have committed little load, the figure falls to as low as $28,000, a gap of up to 100 times. Even within the same market, sites adjacent to 230-kilovolt and 500-kilovolt transmission lines sell for 2.4 times more than sites served only by lines of 138 kilovolts or less.
Last November, Amazon bought 189 acres of raw land in Prince William County, Virginia, for $700 million, or $3.7 million per acre. The land came with permits allowing construction of up to three substations.
Let me add one experience of my own. Not long ago I was looking for land to build a factory. The first two things I asked the broker were how much electricity had been secured for the site and when grid connection would be possible. A few years ago, those questions would have come only after price and location had been weighed. I was not trying to build a data center. These days, even someone building a factory asks about electricity first.
Even if Power Forecasts Are Wrong, the "Electricity First" Rule Holds
Whether forecasts for AI power demand will prove accurate cannot be known precisely. Not even the utilities know. But the rules have already changed.
As of September, 25 U.S. states had approved separate rate schedules for large loads, according to Edison Electric Institute tallies. AEP Ohio's tariff requires data centers of 25 megawatts or more to pay for at least 85% of their contracted capacity for 12 years, whether they use it or not. PJM's capacity prices are already set by auction through the May 2029 delivery year. Even if only half of the planned data centers get built, these rules and prices will remain, and they will land on the bills of plants like Belden Brick.
Korea stands at both ends of this story. It sells the transformers, and it is also the party queuing up in front of the same grid as the data centers when it builds or expands factories in the U.S. When Korean companies negotiate with U.S. state governments, what they need to confirm before tax incentives is the timing of grid connection and the rate class.
The first project under the $350 billion Korea-U.S. investment package announced on Sept. 30 was not a factory either. It is a 6,472-megawatt gas-fired power plant in Texas, built to supply electricity to nearby AI data centers.
So the question now has to change. In valuing companies and real estate, it is no longer enough to ask where they are. One has to confirm how much electricity they can secure. The core issue now is who holds the right to use electricity.
He is...
· B.A. in Economics, University of Chicago; M.S. in Financial Engineering, New York University
· Current: Vice President, Chafia Asset Management (2024–); CFA, CAIA
· Former: Director, PE/Real Asset Investment Team, Industry Capital (2021–2024)
· Former: Director, Investment Team, iCap Advisors (2016–2021)
· Former: Head of Investment Team, WNW Capital (2010–2016)
· Former: Analyst, Morgan Stanley (2007–2008)
