AI data centres are driving up power bills at America’s Rust Belt factories

Belden Brick’s electricity costs in Sugarcreek, Ohio rose 90% last year, largely attributed to the regional surge in data centers supporting the AI boom. The article frames AI-driven demand for computing infrastructure as a key driver of local power price volatility for energy-intensive manufacturers. Overall, the impact appears industry/cost-specific rather than a broad market move.

Analysis

This is less about one manufacturer and more about a localized scarcity shock in PJM. AI data-center load can absorb grid capacity faster than generation and transmission can be added, so the first beneficiaries are the toll collectors: grid equipment, transformers, switchgear, peakers, and regulated utilities that earn on incremental capex. The first losers are energy-intensive industrials with weak pricing power; their margin compression shows up before end-demand deteriorates.

Over the next 1-3 months, the key catalyst is whether utility commissions, interconnect queues, and capacity auctions confirm that this is a durable shortage rather than a one-off. If locational power prices keep rising, the trade rotates from semis/software into power infrastructure names like GEV, ETN, PWR, HUBB, CEG, and NRG. The second-order effect is inflationary: higher electricity costs can leak into construction materials, paper/packaging, and logistics, pressuring small-cap industrials that cannot surcharge quickly.

The contrarian risk is that the market may be underpricing how much AI buildout is constrained by physical power, not just chip supply. If hyperscalers self-fund substations, PPAs, and behind-the-meter generation, the cost burden shifts away from them and toward grid vendors and utilities with stronger balance sheets. Falsifier: a rapid decline in regional wholesale power prices or faster-than-expected approvals for new generation/interconnection; that would argue the current scarcity premium is temporary rather than structural.

Key Decisions for Investors

• Buy GEV or ETN on pullbacks as a 6-12 month grid-capex beneficiary; use 2026 call spreads if you want convexity. Thesis breaks if order growth or utility capex guidance rolls over.
• Pair trade: long ETN / short XLB into any AI-related rally. The short leg captures energy-cost pressure on materials and industrials; keep sizing modest because pricing pass-through could blunt the downside.
• Initiate a long CEG or NRG basket for 3-6 months if PJM capacity and forward power prices remain elevated. Hedge with a small short in a low-growth utility if you want to isolate scarcity premium; watch for reversal if power prices normalize.
• Put regional industrial margin pressure on watch: if a meaningful number of Midwest manufacturers start guiding lower gross margins, consider short exposure to energy-intensive cyclicals rather than chasing the AI beneficiaries.
• Avoid aggressively shorting data-center REITs like EQIX/DLR on this theme alone; most tenants can pass through power, so the cleaner expression is the grid and generation bottleneck, not the landlords.