October 22, 2026 - Economic Insider

AI Data Center Expansion Prompts New Utility Cost Safeguards

Economic Insider Staff 2 on October 22, 2026

AI data center expansion is prompting utilities and regulators to adopt new cost protections, according to an industry analysis. Independent research associates earlier facility growth with higher electricity rates, while federal forecasts show further price increases. Decisions in Michigan, Ohio and Virginia illustrate how household bills could be affected.

Key Takeaways

A report highlighted more than 20 data center proceedings decided in 16 states during 2025 and 2026
Lawrence Berkeley National Laboratory estimates data centers could use 11.8% of U.S. electricity by 2030 in its reference scenario
Federal forecasters expect residential electricity prices of 18.2 cents per kilowatt-hour in 2026, up from 17.3 cents in 2025
A June study associated earlier data center openings with a 2.1% increase in residential electricity prices
Michigan approved a 20-year Google electricity agreement with an 80% minimum billing commitment

New contracts are turning estimates of electricity demand into enforceable payment obligations. When Michigan approved power agreements for a Google facility on October 1, Michigan Public Service Commission Chair Dan Scripps said the safeguards would ensure that “other customers aren’t subsidizing this large addition to the state’s energy grid.”

AI Data Center Expansion Brings New Cost Safeguards

Charles River Associates reviewed more than 20 regulatory proceedings decided across 16 states in 2025 and 2026. Its October report, prepared for the Edison Electric Institute, identified minimum charges, longer contracts, financial guarantees and payments for dedicated infrastructure.

These provisions address a risk: utilities might construct equipment for a data center that ultimately uses less electricity than requested or closes before the costs are recovered. The report says large customers could reduce rates for others when their payments adequately cover grid expenses.

That result is not guaranteed for every project. Because the research was prepared for a utility industry association, its conclusions warrant comparison with independent studies of historical prices.

The scale of potential demand explains the attention. Lawrence Berkeley National Laboratory estimated in June that data centers could account for 11.8% of total U.S. electricity consumption in 2030 under a reference scenario, with modeled outcomes ranging from 9.5% to 15.3%.

A previous laboratory report estimated their share at 4.4% in 2023. The newer forecasts depend on assumptions about server shipments, power use and cooling efficiency. Growing AI infrastructure build-outs therefore create planning demands before every proposed facility becomes operational.

Electricity Bill Forecasts Show Multiple Cost Pressures

The U.S. Energy Information Administration’s October 6 outlook forecasts residential electricity prices averaging 18.2 cents per kilowatt-hour in 2026 and 18.7 cents in 2027, compared with 17.3 cents in 2025.

It also projects 2026 wholesale prices averaging $52 per megawatt-hour across 11 monitored hubs, up 11% from 2025. The agency expects a 41% rise in the PJM regional market, with weather affecting prices in several areas. Its outlook also anticipates further growth in commercial-sector electricity use next year.

Wholesale costs are only part of residential bills, which also reflect generation, transmission, distribution and capacity-market expenses. Seasonal demand, illustrated by summer electricity bill forecasts, can contribute to higher costs without proving that data centers caused any particular increase.

Independent research nevertheless suggests data center development can affect rates. A June working paper published by the Massachusetts Institute of Technology’s Center for Energy and Environmental Policy Research studied new facilities from 2010 through 2024.

The authors associated data center entry with 2.7% higher average retail electricity prices, including 2.1% for households. Effects were larger among shareholder-owned utilities and in states with deregulated electricity generation.

Those historical estimates do not establish a specific charge on 2026 bills. They also address a different question from the Charles River Associates review, which examined safeguards approved in more recent proceedings.

Michigan, Ohio and Virginia Set Terms for Large Users

On October 1, the Michigan Public Service Commission conditionally approved electricity agreements between DTE Electric and Google for a data center in Van Buren Township.

The 20-year agreement requires minimum billing based on 80% of contracted demand, even when actual use is lower. Financial assurances and early termination provisions address the possibility of unrecovered costs, with Google responsible for at least 15 years of minimum payments if the agreement ends early.

Google must also pay for renewable generation and battery storage associated with the project. The Michigan commission’s authority covered electricity service terms, not whether the facility could be built.

Ohio established a separate data center tariff for AEP Ohio. The Public Utilities Commission of Ohio requires measures including collateral, minimum demand charges, lengthy contracts and early-exit fees to limit costs being shifted to other customers.

Virginia’s State Corporation Commission approved a new rate class for customers requiring at least 25 megawatts, effective January 1, 2027. Certain large users face minimum charges tied to 85% of contracted transmission and distribution demand and 60% of generation demand.

The states use different safeguards. None establishes a single nationwide bill surcharge, and broader electricity price changes remain possible.

Frequently Asked Questions

Does AI data center expansion automatically increase household bills?

No. AI data center expansion does not create a nationwide surcharge, and any impact depends on utility contracts, local costs and regulatory decisions.

How much electricity could data centers use by 2030?

Lawrence Berkeley National Laboratory estimates 11.8% of U.S. electricity use under its reference scenario, with projections ranging from 9.5% to 15.3%. These are modeled outcomes, not measured future consumption.

Which states have adopted protections for electricity customers?

Michigan, Ohio and Virginia have approved special rates or contracts for large electricity users. The measures include minimum payments and safeguards against shifting costs to existing customers.

Are data centers the only reason electricity prices are rising?

No single cause explains U.S. electricity prices. Federal forecasters also identify weather, fuel costs and expenses for electricity delivery.

Why the Simplest Products Are the Hardest to Design, Inside BrainBlink’s Engineering

Economic Insider Staff on October 22, 2026

By: Ethan Rogers

BrainBlink looks simple: a pocket-sized piece with buttons. The effortless experience is exactly what the design team wanted you to feel. BrainBlink is an arcade device supporting eight games that you can pick up, play with and put down at any moment.

And the problem is that effortless products are almost never simple to build.

“The products that feel the simplest are often the hardest to design well,” says Josh Fairbairn, founder of Morpho, the manufacturing company behind BrainBlink. “You look at it and think ‘it’s just a game.’ And then you realize the iteration behind that simplicity was enormous.”

BrainBlink launched on Kickstarter in 2025 and hit 100 percent funding in under eight hours. The campaign finished at over 1,500 percent of its goal. The numbers are impressive. But the real story is what it took to get there.

Backers From 20+ Countries

The campaign attracted backers from more than 20 countries. That scale of interest reflects something deeper than just product curiosity. It reflects a hunger for something that actually works, a device designed by people who understand both the problem and the manufacturing reality required to solve it.

BrainBlink is a collaboration between Morpho and LastObject. Morpho has shipped more than 2,000 products over 14 years. They know what it takes to get a product from an idea to a factory to a doorstep. LastObject built the world’s first reusable cotton swab. It sold in millions of units across dozens of countries and won a Red Dot Design Award in 2024. They know how to make something people actually use every day.

That combination of proven manufacturing expertise and proven product design is why backers from 20+ countries trusted BrainBlink before it was even built.

The Engineering Behind Simplicity

The engineering behind BrainBlink’s simplicity started with a single principle: every component had to serve the experience. Nothing decorative. Nothing that looked good in a render but created problems on the assembly line.

The tactile feedback. The button response. The way the screen illuminates. Each of these required iteration. Not because they were difficult in isolation, but because they had to work together at scale, across thousands of units, without variation.

Morpho’s approach to manufacturing begins long before production starts. Each project receives a comprehensive quality standard document that establishes tolerances, inspection methods, and defect categories before tooling begins. For BrainBlink, that meant defining exactly how the buttons should feel, how much resistance they should have, and what constituted an acceptable variance.

Critical components are sourced directly from qualified suppliers. These relationships are built over years, which means Morpho knows the quality of work, the reliability of timelines, and the pricing structure before a project starts. Components are individually inspected before they reach the assembly line.

The company places particular emphasis on Design for Manufacturability, or DFM. This is the discipline that identifies production problems before they become expensive ones. For BrainBlink, DFM happened before tooling was committed. Once a mold is cut, design changes become structural and expensive to revisit.

The AI Makes Sure It Stays Hard

Most games have a fixed difficulty. Once you learn the pattern, you stop improving. BrainBlink’s AI watches how you perform on every run: your speed, your error rate, where you hesitate, and recalibrates before the next one.

Players who have been using it for six months are playing at a level that would be impossible for a beginner to follow. That is the point. It should always feel like the edge of what you can do.

The AI is one of the reasons the product works so well and why backers from 20+ countries trusted it enough to fund it at 1,500 percent of goal. It is why people keep playing.

Iteration Until Success

The iteration behind BrainBlink was not visible in the final product and that was the point.

The team tested different button materials. They tested different screen technologies. They tested different ways to house the battery. Each test revealed constraints that the previous iteration had not accounted for. Each constraint led to a new design decision.

“That’s true of almost every product we’ve built,” Fairbairn says. “The effortless ones are the most work.”

That consistency did not happen by accident. It happened because someone spent months thinking about what could go wrong, and then built systems to prevent it from happening.