Follow the Money: Electric Bills, Tax Breaks, and Who Really Pays

An Alderman friend recently asked what I thought about the data center question.

"If a data center goes anywhere near us," I said, "our power rates will probably increase anyway because of how the grid works. So why not welcome one here and at least get the tax benefits?"

That was my initial angle and question - the kind that sounds like common sense until you dig into how electricity pricing and tax incentives actually work. The answer turns out to be more complicated than "yes" or "no."

Let's follow the money.

Before we can understand how data centers affect your electric bill, we need to understand how that bill gets calculated in the first place.

If you're a Granite City resident, your electricity comes from Ameren Illinois. But Ameren doesn't just generate power and send it to your house. It's part of a much larger system.

Ameren Illinois belongs to something called MISO - the Midcontinent Independent System Operator. Think of MISO as the traffic controller for the electrical grid across 15 states and part of Canada. It manages the flow of electricity, runs the markets where power is bought and sold, and makes sure there's enough supply to meet demand.

Northern Illinois - Chicago and the ComEd service territory - belongs to a different grid operator called PJM Interconnection. PJM covers 13 states from Illinois to the East Coast and is the largest grid operator in the country.

This matters because electricity prices aren't just determined by how much power your local utility generates. They're affected by supply and demand across the entire regional grid - and by the infrastructure needed to move that power around.

When your electric bill arrives, it includes several components: the cost of generating the electricity, the cost of transmitting it over long-distance power lines, the cost of distributing it to your neighborhood, and various other charges. Many of these costs are "socialized" - spread across all customers in a service territory rather than charged specifically to the customers who cause them.

That's been the system for a century, and it generally works. But data centers are breaking the model.

Here's the issue: data centers use electricity on a scale that's hard to comprehend.

A single large data center can consume as much power as a small city - the equivalent of 800,000 homes or more. And there aren't just a few of these facilities. Virginia alone has over 500 data centers, with more under construction.

When demand increases that dramatically, two things happen:

First, the cost of keeping power available goes up. Grid operators like MISO and PJM run "capacity auctions" - essentially paying power plants to be ready to generate electricity when demand spikes. As data center demand has exploded, so have the prices in these auctions.

In the PJM territory, an independent watchdog found that data center demand accounted for 63% of the total capacity bill for 2025-2026 - some $9.3 billion. Those costs get passed down to utilities and ultimately to customers through their supply charges.

MISO has seen similar dynamics. Last summer, capacity prices spiked dramatically, and Ameren estimated that the average residential customer saw their bill increase by 18 to 22 percent - roughly $38 to $46 per month - as a result.

Second, new infrastructure has to be built. Data centers need high-voltage transmission lines, substations, transformers, and other equipment to connect them to the grid. Somebody has to pay for all that.

Under the traditional model, those costs get spread across all customers of the utility. A data center locates in your region, needs $100 million in new transmission infrastructure, and everyone's rates go up a little to cover it. The data center gets the power; you get the bill.

How much are we talking about? Let's look at what's happening in states with high data center concentrations.

Electricity bills rose 6% on average nationwide in August 2025 compared to the year before. But in states with lots of data centers, the increases were much steeper: 13% in Virginia, 16% in Illinois, 12% in Ohio.

In Virginia - home to the largest concentration of data centers in the world - a recent analysis found that household power bills could increase by more than 25% by 2030 due to data center demand.

A Bloomberg investigation found that wholesale electricity costs have increased by as much as 267% over the past five years in areas located near significant data center activity. Those wholesale costs flow through to retail rates.

One Union of Concerned Scientists researcher reviewed utility planning documents and found $4.3 billion in data center connection costs in 2024 alone - across just seven states in the PJM territory. Those costs are being folded into the rates paid by all customers.

"The wealthiest companies are building extraordinarily expensive data centers that you and I are subsidizing," the researcher wrote. "Our policy brief walks through this problem and provides recommendations for several ways to assign the new costs to the customers that cause those costs. That is a responsibility of the rate-setting utility commissions."

Electricity isn't the only resource data centers consume at scale. Water is the other appetite that communities need to understand.

Traditional data centers use evaporative cooling - essentially giant air conditioners that spray water across cooling pads, letting evaporation carry heat away. It's effective and relatively cheap to build. It's also extraordinarily thirsty.

A large data center using evaporative cooling can consume 3 to 5 million gallons of water per day - equivalent to a small city's residential water usage. About 80% of that water evaporates and is lost. The remaining 20% gets discharged as wastewater, potentially overwhelming local treatment facilities that weren't designed for that volume.

In water-stressed regions like Phoenix and the Southwest, this has become a major flashpoint. Google faced criticism for negotiating water rates of $6.08 per 1,000 gallons in Mesa, Arizona - while residents paid $10.80. The perception that tech giants receive preferential treatment at the community's expense doesn't help public relations.

But here's where it gets more nuanced: newer cooling technologies can dramatically reduce or even eliminate water consumption.

Closed-loop liquid cooling systems recirculate coolant through sealed pipes, transferring heat without evaporating water. These systems can reduce water use by up to 70% compared to traditional evaporative cooling.

Immersion cooling goes further - submerging servers directly in non-conductive dielectric fluid. The liquid absorbs heat directly from the components, and because it's sealed and recirculated, essentially no water is consumed for cooling at all.

Microsoft announced in August 2024 that all new datacenter designs would use zero-water evaporative cooling technology. These systems fill once during construction and recirculate continuously without requiring fresh water supply.

The catch: these newer technologies cost more upfront. Most data centers still use traditional evaporative cooling because it's cheaper to build, even if it consumes more water and energy over time.

This creates an important negotiating point for communities. A company can promise to use water-efficient technology, but unless that commitment is binding and enforceable, there's nothing stopping them from deploying cheaper, thirstier systems.

Any serious data center agreement should specify:

Water might seem like a secondary concern compared to electricity costs and jobs. But for a community with finite water resources - or one downstream from facilities that could affect water quality - it matters enormously.

So here's where my original logic breaks down - or at least gets more nuanced.

Regional grid dynamics mean Granite City residents might see rate increases regardless of whether a data center locates here specifically. If MISO capacity prices spike because of demand elsewhere, Ameren customers feel it. If PJM costs rise and affect Illinois generally, we're not immune.

But I was wrong to assume that hosting a data center wouldn't add additional costs on top of that.

Local infrastructure costs are different from regional grid costs. If a data center locates in Granite City, the transmission lines and substations needed to connect it to the grid would likely be paid for, at least in part, by local ratepayers. That's a cost Granite City specifically would bear, not just the region generally.

The deal structure matters enormously. Some states and utilities have started requiring data centers to pay their own connection costs rather than socializing them. Oregon recently passed a law requiring data centers to "pay for the actual strain they place on Oregon's electrical grid." Ohio now requires data centers to pay for at least 85% of the power they've contracted, even if they use less.

If Granite City were to host a data center, the question isn't just "will we get tax benefits?" It's "will we get enough tax benefits to offset the costs we'll bear - including potentially higher electric rates for every household and business in town?"

Speaking of tax benefits, let's talk about how those actually work.

When communities court data centers, they typically offer incentives: property tax abatements, sales tax exemptions, income tax credits. The pitch is that even with reduced rates, a billion-dollar facility will generate significant revenue.

But the math isn't always as favorable as it sounds.

Nearly half of state data center subsidy programs - 16 out of 36 - don't even require job creation in exchange for tax breaks. Companies can receive millions in incentives without committing to hire a single local worker.

When incentives do include job requirements, the numbers are often modest. New Jersey requires 100 jobs; most other states require 50 or fewer. Compare that to the thousands of jobs traditional manufacturing facilities might provide.

And the abatements themselves can be substantial. Some communities offer 10, 15, or even 20 years of reduced property taxes. A facility that looks like it's worth $500 million might pay taxes on a fraction of that assessed value for decades.

The net result is that tax revenue from data centers is often much lower than the headline investment numbers would suggest. A billion-dollar facility with aggressive tax abatements might generate less revenue for local schools and services than a much smaller traditional business would.

None of this means communities should automatically reject data center proposals. But it does mean they should negotiate hard.

Some provisions that can protect residents:

Cost allocation requirements. Make the data center pay for its own grid connection and infrastructure upgrades rather than socializing those costs to ratepayers.

Clawback provisions. If the company doesn't deliver promised jobs or investment, they should return some of the tax benefits they received.

Minimum job requirements. With teeth - not just promises, but binding commitments with penalties for non-compliance.

Local hiring preferences. Requirements that a certain percentage of positions go to local residents.

Environmental protections. Limits on noise, water consumption, and emissions - in the zoning code, not just verbal assurances.

Shorter abatement periods. Five years of reduced taxes is different from twenty.

The communities that have successfully managed data center development are the ones that demanded these protections before negotiations began. Once a company knows it has leverage - once a community has signaled it desperately wants the investment - the bargaining position weakens considerably.

This is why what Mayor Parkinson is doing matters.

By getting zoning protections in place before any proposal arrives, Granite City establishes its standards in advance. Any company interested in locating here knows what's required. The city isn't negotiating from desperation.

The Air Products example the Mayor cited is instructive. That facility's noise issues exist because protections weren't in place before it was built. Once something is operating on properly zoned land, the community has little recourse.

The same logic applies to utility costs. If Granite City establishes clear expectations about who pays for what - and works with Ameren and state regulators to ensure data centers bear their own infrastructure costs - residents are protected.

If those provisions aren't in place, and a data center locates here under the traditional cost-socialization model, every household and business in town could be subsidizing a tech giant's electricity bill for decades.

Regional grid dynamics do mean that data center growth elsewhere can affect Granite City rates. We're not an island.

But that's an argument for paying attention and advocating for policies that protect residential ratepayers everywhere - not an argument for giving up and hoping to benefit from a facility that might make the problem worse locally.

The questions to ask about any data center proposal:

On infrastructure costs:

On tax benefits:

On utility rates:

The answers to these questions determine whether a data center actually benefits a community - or just extracts value from it while leaving us residents with the bill.

This is the third article in a five-part series on data centers and what they could mean for Granite City. Next: "What's Happening Elsewhere: Illinois Communities Grapple With Data Center Proposals."

Have a story tip or want to get involved? Contact me at [email protected].

GraniteCitizen is just getting started. Be sure to sign up to get notified when new stories go live - business profiles, heritage features, local news, and more. No spam, just Granite City.