AI Data Centers Drove a 1,038% Grid Price Jump — Factories Pay First

A 141-year-old Ohio brickmaker watched its monthly capacity charge climb from $1,600 to $12,000, and the company's leadership blames AI data centers. The Belden Brick Company, based in Sugarcreek, Ohio, saw its electricity costs surge 90% last year, largely because of data center demand, Reuters reported. Belden raised brick prices 4%, and profits still shrank.

That brick goes into houses, schools and storefronts. Families feel the AI buildout through two channels at once: the monthly utility bill, and the price of goods made by factories paying the same higher power rates. That double exposure is the mechanism behind the story, and it's worth taking apart piece by piece.

What happened to grid prices?

PJM Interconnection runs the mid-Atlantic and Midwestern grid, the region with the nation's highest concentration of data centers. Its capacity prices — the payments that keep generators available at peak demand — rose from $28.92 per megawatt-day in 2024 to $329.17. That is a 1,038% jump in two years.

PJM's independent market monitor didn't mince words about the cause. The monitor concluded that "data center load growth is the primary reason" for the tight supply and high prices.

Factories absorb more of that shock than homes do. Capacity charges make up about 10% of residential bills but can run up to three times that share for manufacturers. As of December 2025, industrial electricity prices had risen 31% in Pennsylvania and 26% in Ohio, against 7% nationwide. The Rust Belt, where the data center boom concentrates, is paying well above the national pace.

The takeaway: the grid market repriced dramatically in two years, and manufacturers carry the heaviest share of that increase.

Why are manufacturers squeezed from both sides?

Demand keeps building. The U.S. Energy Information Administration expects American power use to beat record highs in 2026 and 2027 as AI use surges. Its May outlook projected industrial electricity consumption rising 1.0% in 2026 and 4.0% in 2027, with residential prices averaging 18.2 cents per kilowatthour this year — up nearly 5%.

PJM forecasts peak summer demand climbing from 161,000MW to 241,000MW over 15 years. All of that load has to be generated, moved and cooled by somebody.

Regulators are struggling to protect households, and factories are becoming collateral damage. The problem is structural: manufacturers sit in the same rate class as data centers, so rules aimed at hyperscalers land on them too. The Federal Energy Regulatory Commission (FERC) has proposed transmission charges for onsite generation, which would penalize plants that build their own power.

Paul Cicio of the Industrial Energy Consumers of America put the industry's frustration directly to Reuters: "Manufacturers are not data centers. We should not be impacted by their effort to manage data centers."

The stakes extend beyond individual firms. Steel alone supports more than 56,000 jobs and roughly $94 billion in annual activity inside PJM's region, according to House Energy and Commerce testimony.

The takeaway: demand is accelerating while the rules for who pays remain unfinished, leaving factories exposed from both directions.

How does the second bill reach your kitchen table?

The macro backdrop matters here. Consumer prices rose about 3.7% in the year through September, based on the Consumer Price Index. Retail sales still hit $745.0 billion that month, a firm demand background. But shoppers are already stressed: University of Michigan consumer sentiment sat at 48.1 in September.

A Cornell study adds a distributional dimension: the energy costs from AI data centers land hardest on lower-income, rural households.

Belden's own warning applies well beyond brickmaking. "There are going to be some companies that are on the razor's edge." Companies at that edge face a limited menu of options: raise prices, slow hiring, or move.

Each of those choices pushes the cost downstream. When a factory pays more per megawatt-day, the goods it produces carry that cost to retailers, and retailers pass it to shoppers. The household that never touches an AI model still pays for the compute boom — once at the meter, once at the checkout.

The takeaway: the AI power bill doesn't stay inside the data center industry; it travels through the supply chain into ordinary purchases.

What decides whether you pay once or twice?

The deciding factor is rate design. States are exploring standardized large-load tariffs to replace confidential data center contracts. The design of those tariffs determines everything:

• If the tariffs put hyperscalers in their own rate class, the cost stays with the companies driving it.
• If manufacturers stay lumped in with data centers, households pay at the meter and again at the checkout.

That is the fork in the road, and regulators haven't chosen a path yet. The next two quarters hold three concrete markers to watch:

1. FERC's final action on onsite-generation charges — the proposal that would penalize factories generating their own power.
2. PJM's next capacity auction, which will show whether the $329.17 per megawatt-day level holds or climbs further.
3. Whether Pennsylvania and Ohio regulators separate data centers from factories in their rate classes.

Each of those decisions will either insulate manufacturers from hyperscaler-driven costs or extend the double bill to households. For a 141-year-old company like Belden, and for the thousands of factories in PJM's footprint, the regulatory calendar now matters as much as the electricity price itself.

The takeaway: tariff design, not technology, will decide who ultimately pays for the AI power buildout.